Analytics

Data Science, Data Analysis and

Predictive Analytics for Business
4th Edition
By Daniel Covington

2
© Copyright 2016 - All rights reserved.
In no way is it legal to reproduce, duplicate, or transmit any part of this
document in either electronic means or in printed format. Recording of this
publication is strictly prohibited and any storage of this document is not allowed
unless with written permission from the publisher. All rights reserved.
The information provided herein is stated to be truthful and consistent, in that
any liability, in terms of inattention or otherwise, by any usage or abuse of any
policies, processes, or directions contained within is the solitary and utter
responsibility of the recipient reader. Under no circumstances will any legal
responsibility or blame be held against the publisher for any reparation,
damages, or monetary loss due to the information herein, either directly or
indirectly.
Respective authors own all copyrights not held by the publisher.

Legal Notice:
This book is copyright protected. This is only for personal use. You cannot
amend, distribute, sell, use, quote or paraphrase any part or the content within
this book without the consent of the author or copyright owner. Legal action
will be pursued if this is breached.

Disclaimer Notice:
Please note the information contained within this document is for educational
and entertainment purposes only. Every attempt has been made to provide
accurate, up to date and reliable complete information. No warranties of any
kind are expressed or implied. Readers acknowledge that the author is not
engaging in the rendering of legal, financial, medical or professional advice.
By reading this document, the reader agrees that under no circumstances are we
responsible for any losses, direct or indirect, which are incurred as a result of
the use of information contained within this document, including, but not
limited to, - errors, omissions, or inaccuracies.

3
Table of Contents
Introduction
Chapter 1: The Importance of Data in Business
What Is The Source Of Data?
How Can Data Improve Your Business?
Chapter 2: Big Data
Big Data – The New Advantage
Big Data Creates Value
Big Data = Big Deal
Chapter 3: Benefits of Big Data to Small Businesses
Chapter 4: Key Training for Big Data Handling
Chapter 5: Process of Data Analysis
What Is Data Analysis?
Steps Involved In Data Analysis
Chapter 6: Predictive Analytics
What Is Predictive Analytics?
What Are The Types Of Predictive Analytics?
Chapter 7: Predictive Analysis Techniques
Regression Techniques
Machine Learning Techniques
Chapter 8: Why Predictive Analytics?
Analytical Customer Relationship Management (CRM)
Clinical Decision Support Systems
Collection Analysis
Cross Sell
Customer Retention
Direct Marketing
Fraud Detection
Operations
Underwriting
Chapter 9: What is Data Science?
Data Science Skill Set:
What Is A Data Scientist?
How Analytics And Machine Learning Are Linked To Data Science

4
 Data Munging
Chapter 10: Further Analysis of a Data Scientist’s Skills
Further Skills Recommend For A Data Scientist
Further Demystification Of Data Science
Chapter 11: Big Data Impact Envisaged by 2020
How Does Moving Online Impact Data?
What’s The Market Like For Big Data?
How Bad Is The Security Situation Today?
Chapter 12: Benefits of Data Science in the Field of Finance
Chapter 13: How Data Science Benefits Retail
Chapter 14: Data Science Improving Travel
Chapter 15: Big Data and Law Enforcement
Chapter 16: What about Prescriptive Analytics?
What is Prescriptive Analytics?
What Are The Benefits Of Prescriptive Analytics?
Data Analysis And Big Data Glossary
Conclusion

5
6
Introduction

What defines the success of a business? Is it the number of people employed by

the firm? Is it the sales turnover of the business? Or is it the strength of the
customer base of the firm? Does employee satisfaction play a role in the success
of business operations? How does management factor into the overall
operational success? How critical is the role of the data scientist in this process?
Finally does fiscal responsibility play any role in the success of any business?
To answer any of the aforesaid questions, it is important that you have the
required data in hand. For instance, you need to know how many staff you have
employed first to assess the value contributed by them to the growth of your
business. Similarly, you need to have a repository of all the customers along
with details of their transactions to understand if your customer base is
contributing to the success of your firm.
Why is data important? Data is important for the sustenance of a business. The
most preliminary reason why it is important is because you need information to
be aware of the state of affairs. For instance, if you don’t know how many units
your company sells, in a month, you will never be able to state if your business
is doing well or not. There are several other reasons as to why data is important.
I have dealt in detail in the upcoming chapters of this book regarding the
importance of data.
Just collecting data is not enough. Analyzing and putting it to use is important.
Of course, if you belong to the class of people who are not worried if their
business lost a customer or not, you don’t have to spend time on analyzing data.
However, if this attitude persists, you will soon see the end of your business,
with the growing competitors who care about the expectations of their
customers. Hence, this is where predictive analytics comes into play. How you
employ predictive analytics in your firm is what distinguishes your firm from
the others in the market. Predictive analytics has the capacity to change the way
you play your game in the market. It is capable of giving you that little edge
over your competitor.
In the first chapter of this book, I have highlighted the importance of data in
business. I have also highlighted how data plays an important role in increasing
the efficiency of a business. In the second chapter of this book, I have
mentioned the different steps involved in the process of data analysis. In the
third chapter of this book, I will be taking you through the basics of predictive
analytics and the various methods involved. I have mentioned the different
techniques used for conducting predictive analytics in the fourth chapter of this
book. In the final chapter of this book, you will see how predictive analytics is
being employed in different fields. You will truly appreciate its flexibility,

7
seeing how it can be used in finance as well as in medicine, in operations as
well as in marketing.
The fields of data analysis and predictive analytics are so vast and there are so
many sub branches to these fields, which are even more extensive. One of
which is prescriptive analysis, which, I will briefly cover in the last chapter. I
have covered only the fundamentals of these fields in this book. It is being used
by large industries to determine what will happen in future and using that data
to prevent or make things happen in the future.
I hope you truly enjoy this book. I am sure you will be motivated to manage
your data better and employ predictive analytics in your business and reap the
maximum benefits. Thank you for purchasing this book.

8
9
Chapter 1: The Importance of Data in Business

Were you ever involved or interested in the study of the classic languages?
Languages that are otherwise known as “dead” languages? While the sheer
complexity of these languages is fascinating, the best part about them, the part
that will simply blow your mind is that the ancients used some extremely
tedious methods to preserve sacred texts that range from just a few hundred
years old to several thousand. To preserve these texts, scribes would
painstakingly copy them, not just once but several times and this could take
years to complete. They copied the texts onto papyrus paper and used ink made
from oil, burnt wood and water. Another common method was to chisel the
characters onto shards of pottery or onto stone. While these were exceedingly
time intensive and, I have no doubt extremely mind numbing at times, the
information that has been preserved was deemed to be so valuable to the ancient
civilizations, and certain people dedicated their whole lives to doing it.
So, I can hear you asking, what on earth do ancient texts and dead languages
have to do with data analysis and business analytics? Data, everything is about
the data; it is something that is all around us and we simply can’t get enough of
it. Think about today; think about social media platforms, platforms that are
revolutionizing the landscape for marketing because they are providing
companies with a whole set of analytics that allow them to measure how
successful, or not as the case may be, their company content is and many of
these platforms provide these analytic tools for free.
On the other hand, there are platforms that charge a high price to provide you
with high quality data, telling you what does and doesn’t work on your website.
In the world of business, market and product data gives a business the edge over
their competitors and that makes it worth its weight in gold. Important data
includes historical events, weather, products, trends, customer tendencies,
outliers and anything else that is relevant to business verticals.
What has changed is the way that data is stored. It is no longer a long and
cumbersome task; it is automatic, requires little in the way of human
intervention and it is done on a huge scale. Today’s modern day scribes are
connected sensors.
If you think about it, if you were to collect every single piece of information
that you could, every piece of data that society generated, it would be nothing
compared to what you will face in the next few years. This is the Internet of
Things. This is because most of our devices are connected; they record
performance and usage data, they transmit that data. Sensors record

10
environmental data. Cities are completely connected to ensure traffic and
infrastructures are running at top level. Delivery vehicles are connected so their
location is monitored, their efficiency, even to the extent that a problem with the
vehicle can be identified early. Buildings and homes are connected to improve
the cost and control of energy, while manufacturing premises are connected so
that the communication of critical data is automatic. This is the present and ever
more so, it is the future. We are connected; everything is connected to the
internet.
The fact that data is important really isn’t anything new. It’s just that we have
moved past a scribe and a chisel to using microprocessors. The way we capture
data and the type of data we capture is ever hanging and it is vital that you stay
up to date and ahead of the game if you want to be in the game to win.
Even though it is the service provided or the goods manufactured by a company,
which helps it in establishing a niche for itself in the market, data plays a crucial
role in sustaining the success. In today’s technology driven world, information
can make or break a business. For instance, there are businesses that have
disappeared in such a short time, because they failed to gauge their progress or
customer base. On the other hand, we also have start ups that have been doing
extremely well because of the increased importance they show towards numbers
and the expectations of their customer base.

11
What Is The Source Of Data?
By data, it could refer to the sales figures or the feedback from the customers or
the demand for the product or service. Some of the sources of data for a
company are as follows:
Transactional data:
This could be pulled out from your ledger, sales reports and web payment
transactions. If you have a customer relationship management system in
place, you will also be able to take stock of how your customers are
spending on your products.

Online engagement reporting:

This is data pulled out based on the interaction of customers on your
website. There are tools available such as Crazy egg and Google
analytics, which can help you to collect data from your website.

Social media:
Social networking sites such as Twitter, Facebook and LinkedIn also
provide insights on the customer traffic on your page. You can also use
these platforms to conduct a cost effective survey on the tastes and
preferences of the customers and use it to improve their products or
services.

12
How Can Data Improve Your Business?
Data can improve the efficiency of your business in many ways. Here is a taste
of how data can play an important role in upping your game.
Improving your marketing strategies:
Based on the data collected, it is easier for the company to come up with
innovative and attractive marketing strategies. It is easier for the company
to alter existing marketing strategies and policies in such a fashion that it
is in line with the current trends and customer expectations.

Identifying pain points

If your business is driven by predetermined processes and patterns, then
data can help you identify any deviations from the usual. These small
deviations could be the reason behind the sudden decrease in sales or
increase in customer complaints or decrease in productivity. With the help
of data, you will be able to catch these little mishaps early and take
corrective actions.

Detecting fraud
When you have the numbers in hand, it will be easier for you to detect
any fraud that is being committed. For instance, when you have the
purchase invoice of 100 units of tins and when you see, from your sales
reports, that only 90 tins have been sold out and you are missing ten tins
from your inventory, you know where to look. Most companies are being
silent victims of fraud because they are not aware of the fraud being
committed in the first place. One important reason for that is the absence
of proper data management, which could have helped them detect fraud
easily in the early stages.

Identify Data Breaches

The explosion of complex data streams in the past few years has brought
on a new set of problems where fraudulent practices are concerned. They
have become more subtle and comprehensive. Their negative effects can
be widespread and impact your company’s retail, accounting, payroll and
other business systems adversely. In other words, data hackers will
become more devious in their attack on company data systems.
Using data analytics and triggers, your company can prevent fraudulent
data compromises in your systems, which can severely cripple your

13
business. Data analytics tools allow your company to develop data testing
processes to detect early signs of fraudulent activity in your data systems.
Standard fraud testing may not be feasible in certain circumstances. If this
is the case with your company, then special, tailored tests can be
developed and used to trace any possible fraudulent activity in your data
processes.
Traditionally companies have waited until their operations have been
impacted financially to investigate fraud and breach prevention strategies.
This is no longer feasible in today’s rapidly changing data-saturated
world. With information being disseminated worldwide so quickly,
undetected fraudulent activity can cripple a company and its subsidiaries
in no time globally.
Conversely, data analytics testing can equally stop potential fraudulent
data destruction by revealing indicators that fraud has begun to seep into
data systems. Fraud can be stopped quickly for a company and its partners
worldwide if these data analytic tests are put into application periodically.

Improving customer experience

As I mentioned before, data also includes the feedback provided by
customers. Based on their feedback, you will be able to work on areas,
which can help you improve the quality of your product or service and
thereby satisfy the customer. Similarly, when you have a repository of
customer feedback, you will be able to customize your product or service
in a better fashion. For instance, there are companies, which send out
customized personal emails to the customers. This just sends out a
message that the company genuinely cares about its customers and would
like to satisfy them. This is possible solely because of the effective data
management.

Decision making
Data is very crucial for making important business decisions. For
instance, if you want to launch a new product in the market, it is
important that you first collect data about the current trends in the market,
the size of the consumer base, the pricing of the competitors etc. If the
decisions taken by the company are not driven by data, then it could cost
the company a lot. For instance, if your company decides to launch a
product, without taking into consideration the price of the competitor’s
product, then there is a possibility that your product might be overpriced.
As is the case with most overpriced products, the company would have
trouble in increasing the sales figures.

14
When I say decisions, I don’t really just refer to the decisions pertaining
to the product or service offered by the company. Data can also be useful
in taking decisions with respect to the functioning of the departments,
manpower management etc. For instance, data can help you assess how
many personnel are required for the effective functioning of a department,
in line with the business requirements. This information can help you to
decide if a certain department is overstaffed or understaffed.

Hiring Process
Using data in selecting the right personnel seems to be a neglected
practice in corporate life. It’s critical to position the most qualified person
in the right job in your company. You want your business to be highly
successful in every facet of operation. Using data to hire the right person
is a sure way to put the best person in the job. What kind of data would
you use to hire a professional?
Big companies, with astronomical budgets, use big data to locate and
select the most skilled individuals for the right jobs. Start-ups and small
companies would benefit immensely from using big data to hire the right
group of people to make their recruiting successful from the start. This
avenue for gathering data for hiring purposes has proved to be a
successful avenue for hiring the right fit for organizations of all sizes.
Again companies can use their data scientists to extract and interpret the
precise data for human resource departments.

Using Social Media platforms to recruit

Social Media platforms (Facebook, Twitter, and LinkedIn to name some)
are hidden gold mines as data sources for finding high profile candidates
for the right positions within companies. Take Twitter, for instance,
company job recruiters can follow people who tweet specifically about
their industry. Through this process, a company can find out and recruit
the ideal candidates by their knowledge of a particular industry or job
within that industry. Do their “tweets” inspire new thoughts, and possibly
new innovations for their industry? If so, you have a whole pool of
potential job applicants.
Facebook is another option for data gathering for potential job candidates.
Remember, these avenues are virtually free which corporations could use
as major cost effective strategies. Facebook is all about collecting social
networking data for companies looking to expand their workforce or
replace an existing open position(s). Company recruiters can join industry
niche groups or niche job groups.” Liking” and following group

15
member’s comments will establish the company’s presence within the
group allowing highly focused job ads to be posted within the group. The
company can increase views widening the pool of potential job
candidates.
It is easy to establish a timeline to brand the company as an innovative
and cutting-edge place to work. You establish your presence by engaging
with friends/followers who are in the same industry as your company. For
a minimal advertising fee promote your job ad. By doing this you
geometrically increase your reach to potential job seekers. If your
company puts out highly effective job data posts you will reel in a higher
yield of highly skilled job searchers, therefore greatly increasing the
percentages of getting the perfect fit for your job.

Niche social groups

Niche socials groups are specialized groups that you can join on social
and web platforms that can help you find specific skillsets. For example,
if you are looking to hire a Human Resources manager. What better place
to find a potential recruit than by finding a specific Human Resources
social group? Locate social connections within that group and post
descriptive but alluring job data posts and you may find the right person
to fit into your Human Resources position. Even if your company doesn’t
find the right person there, those group members surely have referrals.
Again, approaching these groups is a very cost effective way to advertise
your job data posts.

Innovative data gathering methods for the hiring process

Why not think outside the hiring process box and try new methods of data
collection to hire the right professional? Use social collecting data sites
that collect data from sites like Facebook, Google+, LinkedIn and Twitter.
Your company can search on these sites extracting pertinent data from
posts made by potential job candidates. This data can be used to help your
company connect with highly efficient job applicants.
An overlooked, but very good data pool to use is keywords. Keywords are
used on the internet for every type of search imaginable Why not use the
most visible keywords in your online job descriptions? By doing this your
company can widely increase the number of views your job posting will
attract.
You can also use computers and software to find the right job candidate
for your company. Traditionally, these data sources have been used to

16
either terminate a company’s employee or analyze if an existing employee
is a right fit for another job within the company.
Why not try a whole new data collecting system? This system would be
rooted in a different set of standards than the usual IQ, tests, skill testing
or physical exams. These are still valuable tools to measure candidates by
but they are limiting. Another focus can be the strong personality traits a
candidate may possess. Is the person negative? Is the person
argumentative? Is the person an isolationist that doesn’t get along with
other people? These types of people can be identified in this personality
trait database and be filtered out as possible team members. This type of
data properly extrapolated will save the company time, resources and
training materials. By eliminating a mismatch between the expectations
the company requires for the job and the person who would potentially
fill the job.
Another benefit of this data gathering system is the results will not only
identify skilled people for the right jobs, but people with the right
personality to fit in with the current company culture. It’s imperative that
a person is sociable and will be able to engage other employees to
produce the most effective working relationships. The healthier the
working environment the more effective company production is over all.

Gamification
This is a unique data tool that isn’t currently in widespread use. It does
motivate candidates to press in and put forth their best effort in the job
selection process. You provide the person with “badges” and other virtual
goods that will motivate them to persevere through the process. In turn
their skills in being able to perform the job requirements will be readily
obvious. This also makes the job application a fun experience instead of a
typical tedious task.

Job Previews
Pre-planning the job hiring process with accurate data about what the job
requirements are will prepare the job seeker to know what to expect if he
is hired for the position. It is said a lot of learning on the job is by trial
and error, which prolongs the learning process. This takes more time to
get the employee up to speed to efficiently function as a valuable resource
within the company. Incorporating the job preview data into the hiring
process reduces the learning curve and helps the employee become
efficient that much quicker.
These are some innovative data gathering methods companies can use to

17
 streamline the hiring process. They also help Human Resource
departments pick the most skilled people to fill their employment needs.

Hence, data is very crucial in aiding businesses to take effective decisions.

These are some of the reasons why data is crucial for the effective functioning
of a business. Now that we have had a glance at the importance of data, let us
get into the other aspects of data analysis in the upcoming chapters.

18
19
Chapter 2: Big Data

Data is firmly woven into the fabric of society, across the entire globe and, like
every other important production factor, like human capital and hard assets,
much of our modern economic activity could not happen without it. Big data is,
in a nutshell, large amounts of data that can be gathered up and analyzed to see
if any patterns emerge and to make better decisions. In the very near future, big
data will become the base on which companies compete and grow, the base on
which productivity will be strongly enhanced and significant value will be
created for the global economy by increasing the quality of services and
products and reducing the amount of waste.
Until now, the river of data that has been flooding the world was something that
most likely only grabbed the excitement of some data geeks. Now, we are all
excited by it simply because the sheer amount of data that is generated, mined
and stored has become one of the most relevant factors, in an economic sense,
for consumers, governments and businesses alike.
Looking back, we can see now that trends in IT innovation and investment, and
their impact on productivity and competitiveness suggest that big data has the
ability to make sweeping changes to our lives. Big data has the same
preconditions that allowed previous IT-enabled innovation to power
productivity (for example, innovations in technology, followed closed by
complementary management innovations being adopted). We expect that big
data technology suppliers and analytic capabilities that are now far advanced
will have as much, if not more impact on productivity as do suppliers of
different technologies.
Every business in the world needs to take big data very seriously as it has a
huge potential to create real value. Some retail companies that are wholly
embracing big data are seeing the potential for a 0% increase in operating
margins

20
Big Data – The New Advantage
Big data is fast becoming the most important way for those companies at the top
to seriously outperform their competition. In most industries, both new entrants
to the market and those that are firmly established will leverage strategies that
are driven by data to compete, to innovate and to capture real value. Examples
of this exist everywhere. Take the healthcare industry; data pioneers are
examining the outcomes of certain pharmaceuticals that are widely prescribed.
During their analysis, they discovered that there were risks and benefits that
were not seen during limited trials.
Other industries that have taken on big data and run with it are using sensors
that are embedded into products to gain data. These products range from a
child’s toy to large-scale industrial goods and the data tells them how the
products are used in the real world. This kind of knowledge allows for better
design of these products in the future.
Big data will assist with creating new opportunities for growth and it will also
help to create entirely new categories of companies, i.e. those that aggregate
industry data and analyze it. A good proportion of these companies will be
situated in the center of large flows of information, where the data that come
from products, buyers, services and suppliers can be analyzed. Now is the time
for forward thinking managers and company leaders to build up their company
capabilities for big data and they need to be aggressive about it.
The scale of big data isn’t the only important factor; we must also take into
account the high frequency and the real-time nature of the data as well. Take
“nowcasting” for example. This is the process of estimating metrics, like
consumer confidence, straight away – something that at one time could only be
done after the fact. This is being used more extensively, adding a considerable
potential to prediction. In a similar way, high frequency allows users to analyze
and test theories to a level that have before been possible.
Studies of major industries have shown that there are a few ways that big data
can be leveraged:
1. Big data has the potential to unlock some serious value for industries
by making all information transparent. There is still a lot of data that
has not yet been captured and stored in digital form or that cannot
easily be found when searched for. Knowledge workers are
expending up to 25% of their time and effort in searching for specific
data and then transferring it to another location, sometimes a virtual
one. This percentage represents a vast amount of inefficiency.
2. As more and more companies store their transactional data in a
digital format, they are able to collect detailed and highly accurate
performance information on just about everything – from inventory

21
 right down to the amount of ick days being taken – thus giving
themselves the ability to boost performance and root out variabilities.
Some of the leading companies use this ability to collect big data and
analyze it to carry out experiments to see how they can make more
informed management decisions.
3. Big data allows companies to divide their customers down into
smaller segments, allowing them to better and more precisely tailor
the services and products that they offer.
4. More sophisticated analytic allows for much better decision making.
It can also cut down the risks, and bring to light some very valuable
insights that might otherwise never see the light of day.
5. We can use big data to create the next generation of services and
products. For example, manufacturers are already using the data
that they get from their embedded sensors to come up with more
innovative after-sales service.

22
Big Data Creates Value
Let’s use the US healthcare system as an example here. If they were to make
effective and creative use of big data to improve quality and efficiency, they
could actually create in excess of $300 billion of value every single year.
Around 70% of that would be from a cut in healthcare expenditure of around
8%.
Moving on to the European developed economies; if government administrates
used big data in the right way, they could create improvements in operational
efficiency of around €100 billion every year. And that is just in one area; we
haven’t looked at what they could achieve if they used advanced analytic tools
to boost tax revenue collection and cut down on errors and fraud.
It isn’t just organizations or companies that and gain benefit from using big
data. The consumer can benefit significantly, i.e. those who use services that are
enabled by location data can realize consumer surplus of up to $600 billion.
Take smart routing that uses real-time traffic information, for example. This is
one of the most used of all the applications that use location data and, as more
and more people use smartphones and as more and more of those people take
advantage of the free maps apps provided, smart routing use is likely to grow
significantly. By the year 2020, it is expected that more than 70% of all mobile
phones will have GPS capability built in, way more than the 20% recorded in
2010. As such, we can estimate that, by 2020, the global value of smart routing
has the potential to create savings of around $500 billion in fuel and time
savings. This is the equivalent of cutting 20 billion driving hours or saving a
driver around 15 hours a year on the road and savings of around $150 billion in
fuel.
The most potential for value from big data is going to come from the
combination of data pools. Again, the US healthcare system has four significant
data pools – activity and cost, clinical, medical and pharmaceutical products R
& D and patient data. Each of these data pools is captured and then managed by
a different part of the healthcare system but if big data were used to its full
potential, annual productivity could be increased by around 0.7%. This would
require the combination of data from all the different sources, including from
those organizations who don’t share data at scale. Data sets would need to be
integrated, for example clinical claims and patient records.
Doing this would realize benefits for everyone, from the industry payers to the
patients. The patients themselves would have better access to more healthcare
information and they would be able to compare process of physicians, treatment
and drugs. They could compare effectiveness, allowing them to pick the
medication and treatment that suited them the best. To be able to take advantage
of these benefits though, they would have to accept a tradeoff between the

23
benefits and their privacy.
Data security and privacy are two of the biggest hurdles that must be faced if the
benefits of big data are to be truly realized. The most pressing challenge is the
huge shortage of people who have the skills needed to be able to analyze big
data properly. By 2018, the US is facing a shortage of between 140,000 to
190,000 people with the right training in deep analysis and another shortage of
around 1.5 million people with the quantative and managerial skills needed to
interpret the analyses correctly so that they can base their decisions on them.
There are also a whole slew of technological issues, which will have to be
resolved. Incompatible formats and standards as well as legacy systems often
stand in the way of data integration and stop sophisticated analytical tools being
used. Ultimately, in order to make full use of the larger digital datasets is going
to require a technology stack being assembled from computing and storage right
through to the application of visualization and analytical software.
Above all, to take true advantage of big data, access to all data has to be
widened. More and more, organizations will need to have access to data stored
with third parties, for example with customers or business partners, and then
integrate that data with theirs. One of the most important competencies in the
future for data-driven companies will be the ability to come up with compelling
value propositions for other parties, like suppliers, consumers and possibly even
competitors to a certain extent.
For as long as the true power of big data is understood by governments and
companies, the power it has to deliver better productivity, more value for the
consumer and the power it has to fuel the net wave of global economy growth,
there should be some incentive for them to take the necessary actions to
overcome the barrier that stand in their way. By doing this, they will open up
new avenues of competitiveness among industries and individual companies.
They will create a much higher level of efficiency in public sectors that will
allow for better services, even when money is short and they will enable more
productivity across the board.

Big Data Brings Value To Businesses Worldwide

The value Big Data has brought to businesses worldwide and will is
immeasurable..
Here is just a brief summary of ways that it has impacted our world:
Created a whole new career field-Data Science
Big data has revolutionized the way data interpretation is applied
The healthcare industry has been improved considerably by the

24
application of predictive analytics
Laser scanning technology has changed the way Law Enforcement
reconstructs crime scenes
Predictive analytics is changing the roles of caregivers and patients
Data models can now be built to investigate and solve many business
problems
Predictive analytics is changing the way the Real Estate industry is
conducting business

25
Big Data = Big Deal
One more thing that big data could do is open the way for new management
principles. In the early days of professional management, corporate leaders
found that one of the key determining factors of competitive success was a
minimum scale of efficiency. In the same way, the competitive benefits if the
futures are likely to build up with companies that can capture more data that is
of higher quality and use that data at scale with more efficiency.
The following five questions are designed to help company executives
determine and recognize just how big data can benefit them and their
organizations:

What will happen in a “transparent” world, with data available so

widely?
Information is becoming more and more accessible across all sectors and, as
such, it has the potential to threaten those organizations that rely heavily on data
as a competitive asset. Take the real-estate industry, for example. They trade on
access to transaction data and a secretive knowledge of bids and buyer
behaviors. To gain both requires a great deal of expense and a lot of effort.
However, recent years have shown that online specialists have begun bypassing
the agents and allowing buyers and sellers to exchange their own perspectives
on property values and have created a parallel resource for real estate data to be
garnered from.
Pricing and cost data have also become more widely accessible across a whole
slew of industries. Some companies are now assembling satellite imagery that is
readily available; imagery that, when it’s processed and then analyzed can
provide clues about the physical facilities of their competitors. This gives them
ideas about the expansion plans or any constraints that their competitors are
coming up against.
One of the biggest challenges is that much of the data that is being amassed is
being kept in departmental “silos”, such as engineering, R & D, service
operations or manufacturing. This stop the data being exploited in a timely
manner and can also cause other problems. Financial institutions, for example,
suffer because they don’t share data among the diverse business lines, such as
money management, financial markets or lending. This can stop these
companies from coming up with a coherent view of their customers, on an
individual basis, or of having an understanding of the links between the
financial markets.
Some manufacturing companies are trying their hardest to get into these
enclaves. They integrate data from a number of different systems, inviting

26
formerly closed off units to collaborate and looking for information from
external parties, such as customers and suppliers, to help co-create future
products. In the automotive industry, global suppliers make hundreds of
thousands of different components. Integrating their data better would allow the
companies, along with their supply chain partners, to collaborate at the design
stage, something that is crucial to the final cost of manufacturing.

Would being able to test your decisions change the way a company
competes?
Big data brings in the possibility of a different style of decision-making.
Through the use of controlled experiments, organizations will be able to test out
hypotheses and then analyze the results. This should give them results that they
can use to guide their decisions about operational changes and investments.
Effectively, experimentation is going to allow managers to distinguish between
correlation and causation, cutting down on the wide variations in outcomes
while boosting products and financial performance.
These experiments can take a number of different forms. Some of the top
companies online are always testing and running experiments in some cases,
they will set aside a specific part of their web page views in order to test which
factors drive sales or higher usage. Companies who sell actual physical products
use tests to help them make decisions but using big data can take these
experiments to a whole new level. McDonalds has fitted devices into some of its
stores to track customer interaction, the amount of traffic and patterns in
ordering. Using the data gained, they can make decisions on changes to their
menus, changes to the design of their restaurants and training on sales and
productivity, among other things.
Where it isn’t possible to carry out controlled experiments, a company could use
a natural experiment to find where the variables in performance are. One
government sector collected data on a number of different groups of employees
who were all working at different sites but doing similar jobs. Just by making
that data available, the workers who were lagging behind were pushed to
improve performance.

If big data were used for real-time customization, what effect

would it have on the business?
Companies who face the public, as it were, have been using data to divide and
target specific customers for a long time. Big data takes that much further that
what always used to be considered as top of the range targeting, by making it
possible to use real-time personalization. In the future, retailers will be able to
keep track of individual customers and their behaviors, by monitoring internet

27
click streams, and will be able to make changes to preferences and model the
potential behavior in real-time. By doing this, it will be easier for them to know
when a customer is heading towards making a decision on a purchase; they will
be able to “nudge” the decision and take the transaction through to completion
by bundling together products, and offering benefits and reward programs. This
s called real-time targeting and it will also bring in data from loyalty programs
and this, in turn, can help to increase the potential for higher-end purchases by
the most valuable customers.
Retailing is probably the most likely of all industries to be driven by data. They
have vast amounts of data at their fingertips – from purchases made on the
internet, conversations on social network sites and from location data from
smartphone interactions, alongside the birth of new, better analytical tools that
can be used to divide customers down into even smaller segments for better
targeting.

How will big data help management, or even replace it?

Big data opens up more avenues for the application of algorithms and analysis
that is mediated by machines. Some manufacturers use algorithms to analyze
data garnered from sensors on the production line. This helps them to regulate
their processes, reducing waste, avoiding human intervention that can be
expensive and, in some cases, dangerous, and increasing their output. In the
more advanced “digital oilfields”, sensors are used to constantly monitor the
conditions of the wellheads, the pipelines and all the mechanical systems. That
data is then analyzed by computers and the results are fed to operation centers
where, in real-time, they adjust the oil flows to boost production and reduce
downtime. One of the biggest oil companies in the world has managed to
increase their oil production by 5% while reducing staffing and operating cost
by between 10% and 25%.
Products that run from a simple photocopier to a complex jet engine are now
able to generate streams of data that track usage. Manufacturers are able to
analyze the data and, in a few cases, fix glitches in software or send out repair
representatives automatically. In some cases, the data is being used to preempt
failure by scheduling repairs to take place before systems can go down.
The bottom line is big data can be responsible for huge improvements in
performance, much better risk management and uncover insights that would
most likely never have been found. On top of that, the price of analytic
software, communications devices and sensors are falling fast and that means
more companies will be able to afford to join in.

Can big data be used to create a brand new business model?

28
Big data is responsible for coming up with brand new company categories that
embrace business models driven by data. A good many of these companies are
intermediates in a value chain where they can generate valuable “exhaust” data
that is produced by transactions. A major transport company saw that, while
they were going about their transport business, they were gathering in large
amounts of data on product shipments across the globe Sensing a top
opportunity, they came up with a nit that now sells the data gained to
supplement economic and business forecasts.
Another major global company learned a great deal by analyzing their own data
and decided that they should create a new business to the same sort of work for
other organizations. They now aggregate supply chain and shop floor data for a
large number of manufacturers and sells the relevant software tools needed to
improve performance. This part of the business is now outperforming their main
manufacturing business and that is all thanks to big data.
Big Data has created a whole new support model for existing markets. With all
the new data needs for companies the increased demand for qualified people to
support that data is increasing.
As a result your business may need an outside firm to analyze and interpret the
data for you. These are specialized firms that that can help interpret and analyze
the data. These companies would specialize in assimilating large amounts of
data both in structured and unstructured models. These companies will exist
solely for the purpose of supporting leading companies in whatever industry.
Their employee base would be trained in looking for and gathering data in
systems and processes. These data analysis companies support your business in
doing this function for you and charge for their expertise.
They would have the resources and training to assimilate, analyze and interpret
new trends in the data and be required to report back to you any alerts.
A viable alternative would be for existing companies to employ data support
departments within their existing infrastructure. This method would be much
more cost effective then using an outside firm, but will require specialized
skillsets within your company. This could be the next step from your current
information technology group. The data support analysts only focus on data in
the existing information systems. Their focus would be deciphering data flow
by analyzing, interpreting and actually finding new applications for this data.
These new applications and support team would monitor existing data for any
fraud, triggers or issues that it might present.
Big data has created a whole new field of study for colleges and higher
institutions of learning to offer in their degree programs. People are being
trained in the latest methods of big data gathering, analyzing and interpreting.
This career path will lead them to critical positions in the newly trending data
support companies. Big data has not only created a new industry but a whole

29
new field(s) of study. Along educational lines, big data will change the way
teachers are hired. Big data recruiting processes combined with predictive
analytics can stimulate interactive models between the traits the most effective
teachers should have to maximize the learning experience for students.

30
31
Chapter 3: Benefits of Big Data to Small Businesses

One reason small businesses watch from a distance as certain things happen, is
that often the cost involved is prohibitive. As for big data, a good number of big
size companies have embraced it, and that has created the notion that small
businesses are yet to qualify for big data. Nothing could be further from the
reality.
With data analytics, there is no need for complex systems that come with huge
monetary demands if the enterprise in question is small. Much that is needed is
based on proper organization, human attention to detail as well as how well
people in the organization are equipped in the skills of data analytics. Any
organization that can gather ample data that is relevant to its operations, and
analyze it critically, has great chances of seizing business opportunities that they
would otherwise have missed.
A small business can, therefore, utilize data analytics to improve the quality of
its products or services; modify or fully alter its marketing strategies; and even
improve on customer relations. All those areas have a great impact on the
overall productivity of the business and hence its sustainability and profitability.
Yet such improvement does not necessarily come with a massive price tag.

Examples Of Cost-Effective-Techniques Of Data Analytics:

Why is cost an issue anyway? The reason cost is always factored in before a
crucial business decision is made is that cost has a direct impact on profitability.
If, for instance, your small business increases sales revenue by 30% at a cost
that has risen at 35%, the net profit will not have increased and the newly
introduced techniques will have backfired. Still, you cannot stay put when your
competitors are reaping increased profits from improved systems, whether those
systems have anything to do with data analytics or not.
If you realize your competitors are using data analytics and before you know it
they are eating into your market share, it would be suicidal not to consider the
use of data analytics too. Since at the end of the day big data brings to the fore
any business opportunities that exist; and those opportunities come with reduced
costs and increased revenues; it is wise for small entrepreneurs to consider using
it. After all, small businesses grow by making use of new business opportunities
just the same way big businesses do.
This is what small businesses ordinarily rely on for growth:

32
 Personal intuition
Commitment and ability to provide high quality service

Why Small Businesses Do Not Often Give Priority To Big Data

Well, all businesses, big or small, have, for a long time run their businesses
traditionally. However, whenever developers come up with innovative ideas,
they first of all target big businesses as those are likely to jump at new ideas
simply because they can afford to implement them, cost notwithstanding. The
scenario is not much different for big data. Many vendors who market
enterprise-software-solution put a lot of emphasis on the advantages of
economies of scale. Now, how can a small business even think of economies of
scale, when it’s actually struggling to build capacity?
With such kind of marketing, therefore, small enterprises consider themselves
below par; like big data is not for them. This may have made business sense
when the idea of big data had not become very popular, especially because a
relatively big capital outlay was required upfront in preparation to roll out the
application to the thousands of end users involved. And this is, definitely, where
economies of scale come in.
However, in today’s heightened level of business competition, small businesses
need to borrow a leaf from the world leading companies, so that if big data is
working for huge enterprises, small businesses may try it too. Luckily,
innovators have come up with data solutions that are suitable for small
businesses – solutions that equip the small businesses with the appropriate tools
to be used by the type of end users small businesses engage. These solutions are
thus designed to help the small business accomplish its work faster and more
efficiently.
Unless small businesses begin to adopt the concept of big data, big competitors
are likely to expand their market and run the small entrepreneurs out of town.
After all, when big companies use big data, they increase their efficiency, and
that gives them the drive and ability to embrace a bigger market share at the
expense of small companies. As a consequence, it becomes increasingly
difficult for the small business to break even after resigning themselves to the
fate of continually declining profits.

Areas Where Big Data Can Help A Small Business Cost Effectively:
Social media

Here the small business could analyze the relationship that exists between the

33
use of social media – like Twitter; Pinterest; Facebook; and others – and users’
tendency to consume your product or service. With this kind of analysis, you
can then proceed to capitalize on emerging sales opportunities. By setting up a
business presence on these social platforms and targeting potential
clients/customers using the site’s consumer analytics and ad platforms you can
grow your client base faster than using traditional marketing strategies. Setting
special targeting ad strategies can place your ads in front of the clients and
customers that are the most interested in your products and services. This is a
cost effective marketing strategy when done correctly, however some
specialized knowledge of each platform is needed to make this effective.

Launch of an online service

You can take the opportunity to analyze people’s tendency to visit your site. Is
the visitor traffic high? If so, how high is it? While at it, you could study the
users’ habits; like which pages they click next after the landing page. This is the
time you analyze the details that seem to capture the attention of these users. In
the same breath, you can tell the details that turn off the users.
Aren’t such details great at helping you identify the best pages to place your
promotional content? Don’t they also point at the pages and sites where cross
selling opportunities abound? In short, there are cost effective ways of utilizing
big data and small businesses can benefit just as much as big ones in this regard.

Factors To Consider When Preparing For A Big Data Solution:

1) A customized solution
When you are a small business, you don’t go acquiring a big data solution that
has been mass produced, or one that is suitable for a big business. You need one
that your internal users will find applicable and helpful. If your big data solution
is not tailor made for your business, then it needs, at least, to have capabilities
that the users can take as options, without being encumbered by useless
capabilities that have nothing to do with the needs of the business. It also needs
to leverage the solutions your business has already put in place as well as your
entire system.
Ultimately, what you want for your small business is a solution that has
everything you need, integrated and packaged conveniently for use. Beware of
vendors who may wish to have you overhaul your system, removing capabilities
you already have in place; capabilities you have already adopted and even
implemented. Remember to justify every cost you bring upon your business.
In case the recommendations for big businesses seem different from those of
small ones, it is only because small businesses do not go for a massive

34
integrated solution for the entire organization. Rather, they work with smaller
cost centers such as departments. For example, the marketing department could
requisition for a marketing automation system without relying heavily on the
views of the IT department. All that the marketing department is called upon to
do is lay out its unique requirements, like its system requirements; spell out its
cost justifications; and then do appropriate research to identify the solution that
suits the department best. At the end of the day, you could have your small
business with a varied range of solutions, each of them befitting its relevant
department. Each department collects the type of data it deems relevant to its
operations, analyzes it, and acts on it to increase the department’s efficiency.

2) Ease of deployment
You need to acquire a big data solution that is simple to deploy. You also need
to ensure that you are paying for a solution that the users find easy to use. In
short, in terms of putting the solution in place, it need not take more than a
couple of days of weeks, including testing. Any big data solution that is bound
to take you months or even years to put in place in preparation for use is not
suitable for your small business. For starters, time is money and a complex
solution that will cause a department heavy downtime is not worth it. In any
case, as you struggle to deploy your new solution, business opportunities might
be passing you by.
As you ascertain simplicity, you also need to ensure that the big data solution
you are contemplating acquiring can work with your other applications. A
solution will be worth it only if it can work with your existing applications and
systems seamlessly. Otherwise, you could be introducing a liability into your
business. You need a solution that users will be able to use without pushing the
organization to outsource highly paid specialists to help out.
By the same token, the big data solution you identify for your small business
needs to be one that your staffers can use without the need to undergo expensive
and time consuming training. In fact, you need to look for a solution with self
service capabilities, so that even when you change users, new ones can still
make use of it without a hitch; and without the need to always call on the IT
department to help out.

3) The cost involved

Nothing in business makes sense until you know the cost involved. In the case
of big data solutions, it is a good idea to buy one that is versatile so that you can
increase the use of its capabilities as your business grows. When the solution is
new, you may need to use only a few capabilities; and your solution should
allow for that. There should also be room for you to pay for only the capabilities

35
you are prepared to use. However, as the business grows, you may need to use
more of the big data solution’s capabilities.
So, not only should the solution be priced reasonably, it should also come with a
licensing strategy that allows the business to progressively increase the
capabilities used as its need for data analytics goes up. Business owners always
look forward to their small businesses growing fast, and so the big data
capabilities you are able to bring on board should match the business rate of
expansion as well as growth.
Clearly, it is possible and practical to switch from being intuition driven, even
when yours is a small business, to being analytics driven. All you need to do is
identify an IT solution that is suitable for the size and nature of your business.
And from then on, your business will be in a position to enjoy the benefits that
come with big data, mainly, the chance to identify viable business opportunities.

36
37
Chapter 4: Key Training for Big Data Handling

For the continued success of an organization, management needs to focus on the

impact new systems are likely to have on overall operations, and ultimately the
bottom line. Taking a country as an example, the reason that some countries
choose to have small armies while others choose to have big ones is that some
countries consider the existing threat to their security high while others consider
it low. In short, countries do not enroll their youth into the military just to have
them draw a salary. That is the same thinking that runs through the minds of
management as they seek to establish the people in their organization that
require training in big data.
Irrespective of whether an organization is big or small, the ability to employ big
data is a worthwhile challenge. Even if various departments may work
independently when deploying their applications, big data enables the
departments to relate better as far as sharing of information is concerned,
especially because each of them is able to feed the rest with credible
information and relatively faster at that. Still, it is important that training in big
data is limited to people who can actually use it productively in the
organization. In short, there needs to be some eligibility criteria when it comes
to training.
The highest priority needs to go to employees who handle massive data, and
who may need to make use of Hadoop. As has been explained elsewhere in this
book, Hadoop is a software library that is in open-source form, and which helps
in distributed massive data processing. It is the role of the Chief Training
Officer (CTO) to determine who in the organization needs training, if big data is
to benefit the organization. The officer may consult with the various department
heads because they understand what each employee under them does, and they
know the individuals who handle data in heavy traffic. It is the same way a
country’s Army General liaises with government to determine the caliber of
citizens that qualifies for training and recruitment into the army.
However, the CTO is ultimately responsible for enlisting those persons for big
data training, being, as is expected, the topmost person as far as the IT
infrastructure in the organization is concerned. The CTO needs to have in mind
that whatever recommendations are taken, they will help to establish legacy
systems to drive the new technology, which will help the employees to become
more efficient in their respective roles.
Traditionally, organizations have tended to put a lot of emphasis on acquisition
of data. Other times is just thrown around especially now that the web has a
whole world of data from every imaginable quarter. However, it has become
increasingly clear that massive data is only helpful if there is a way of sorting it

38
out so that specific data is classified and utilized where it is most relevant.
Unfortunately, this sorting has been left to be done mainly as per individual’s
intuition and judgment, without any structured system. This means there is
deficiency in people’s ability to utilize data to its optimum.

Current Level Of Competence In Data Management

A report provided by CIO, a popular site that deals with issues of technology,
indicates that the Computing Technology Industry Association (CompTIA) has
done a survey that shows how poorly equipped staff in organizations leveraging
data are, in the aspects of data management. A good number of them are also
deficient in the skills of data analysis. This survey, which was based on
information derived from 500 executives from the business as well as IT
sectors, indicates that half of the organizations in the lead as far as leveraging
data goes have staff that is not sufficiently skilled in data management and
analysis. When it comes to organizations that only leverage data moderately,
71% of them see their staff as poorly equipped or just moderately skilled in the
area of data management and analysis.
What these results point to is the need to make a conscious effort to train staff
when it comes to big data, so that the availability of data makes business sense
to the organization.

Where, Exactly, Is Big Data Training Required?

1. The IT Department
This is a no-brainer. If there are people who are bugged, raveled and nagged, by
the rest of the organization when it comes to matters of IT, it is members of the
IT department. Even when liaison of data between departments fails to work
properly, these are the people who are called upon to help streamline the
systems so that information flow is better coordinated. It is imperative,
therefore, that they are well versed in matters of big data, to be able to give
appropriate support to members of other departments. In fact, when the IT
department is well trained, they save the organization the expenses it would
have incurred seeking outside help in supporting the rolled out systems. The
role of IT Security has become a new player given all of the new rules and
regulations on how data should be treated and stored. This department also
needs to be aware of the latest hacking and breaching applications being used
and get in front of those before they happen. There are also departments being
built around Data Privacy and Compliance within the IT Department that will
need to work heavily with the Legal and Compliance departments to make sure
that the data isn’t compromised and cause significant losses for the company.

39
 2. The Department of Product Development.
Not only is this department tasked with creating new products, it is also tasked
with re-engineering existing products. Everyone involved is called upon to have
a fresh way of thinking, what can be aptly termed, re-thinking innovation.
Everyone in the department is deeply involved in every step of the Research and
Development process.

For re-engineering, there is a massive responsibility on the part of the staff, and
it involves a full understanding of the product as it currently is, and the product
as it is envisaged to be after the innovative process of re-engineering. For this
understanding to be comprehensive and helpful, massive data is involved, and it
needs to be well analyzed, critically evaluated, and properly utilized. That is
why the staff in the department of product development needs to be well
trained. They should be in a position to capitalize on the advantages provided by
big data, particularly when it comes to data integration across existing touch
points in the process of product development. Some of those important touch
points include product design; software applications; diagnostics;
manufacturing; quality determination; among others.

3. The Department of Finance

Ever heard something like show me the money? Nobody wants to say that more
than the staff in the department of finance. And if you cannot show them the
money, they will want to find out independently if you are really worth funding
as a department or you are actually a liability to the organization. It is, therefore,
clear why the staff in this department needs to be well trained in handling big
data. They need to be able to say in monetary terms if there is value in
deploying certain resources or not.
In fact, the department of finance is almost as central to the organization as that
of IT, and its employees need to be well trained to make use of big data
platforms as far as financial modeling is concerned. Only then can a business be
sustained because without conservative spending and optimum funding of
projects, money, sometimes referred to as the actual blood of the business, is
likely to dry out; meaning the business goes under. The essence of training
personnel in the finance department is to prepare them to make use of big data
in their core roles of business planning; auditing; accounting; and overall
controlling of the company’s finances. Conversely, when the department does
these roles well, it ends up creating massive finances for the organization.
Some of the ways members of the finance staff contribute to the success of the
organization when they are well trained in big data is by managing to generate
reliable cash flow statements; compliance in finance standards; cost modeling;

40
prize realization; and so on. With the influx of data, the finance function has
become more complex, especially when the staff wants to get insights into the
future of the organization. As such, training of finance staff in big data is not a
luxury but a necessity.

4. The Human Resource Department (HR)

Though it may not appear obvious, the Department of HR can utilize skills in
application of big data to improve on the quality of personnel hired.
Competence in big data analysis also comes in handy in assessing the
capabilities of existing employees and in determining the capabilities required
in future engagements.
As such, members of the HR department need to be well trained too so that they
can analyze data according to relevance, with a view to engaging in a more
strategic manner when it comes to its functions. With that demand, it becomes
imperative that this staff is able to use available tools for data analysis, because
such are the competencies that will ultimately help in solving issues of staff
retention; staff-customer relations that affect sales; talent gaps within the
organization; quality of candidates to shortlist; and so on. In short, HR is no
longer about the number of employees engaged, but also about predictive
analysis of matters that have to do with the human resource.

5. The Department of Supply and Logistics

No business can thrive when customers perennially complain about late
deliveries; breakages on transit; and other factors that are irritating to customers.
It is, therefore, important that the staff in the department of supply and logistics
be well trained to handle big data, so that they can utilize it in implementing the
department’s strategies and in achieving its goals. Once the members of staff in
this department are trained in big data, they will be in a position to improve
performance and save on cost, primarily by being faster and more agile in their
activities. On the overall, their training would greatly improve on the
department’s efficiency in service delivery.
What is the import of that? Well, it is massive improvement in customer
experience as well as emergence of a fresh business model. Needless to say, this
achievement comes with significant conservation of resources, in terms of
reduced downtime; reduced overtime; reduced waste; and so on. Ultimately, the
smooth supply chain and the great logistics put in place end up promoting your
brand name markedly. And that is precisely how a business pulls in more
customers, increasing the business market share, without planned and costly
marketing.

41
 6. The Department of Operations
In many organizations, this department also encompasses customer support. As
such, it is important to train the employees in this department, so that whatever
an employee does in the course of duty, there is customer satisfaction in
consideration. Besides, it is important that the employees understand the impact
of providing great customer support even after they have made a sale. Big data
training enlightens the employees very well in this regard. Being able to analyze
big data shows the staff clearly that great customer service and support in key to
the success of the organization because it improves customer retention;
expansion of customer base; and such other success.

7. Department of Marketing
Do you realize how critical numbers are in marketing? Whether it is the number
of customers you have brought on board in the last one year; the percentage of
market share you have captured within a given time; the sales revenue your
product has earned from a certain demographic; this and more keep the
marketing department busy. There is massive data in the market, some positive
and some potentially damaging to your brand, and it is imperative that members
of staff in the marketing department are able to amass the data that appears
relevant; analyze it and sieve it to remain with what can be organized and acted
upon to the benefit of your brand. Without the skills to handle big data, it is
difficult to make head or tail of the mass of data making the traffic through
social media, in this era when digital marketing is the order of the day.
However, with proper training in big data, the department of marketing can
measure with relative accuracy the response your advertisements have on the
market; the volume and impact of click-through-rate; impressions; Return on
Investment (ROI); and such other factors. While plenty of such data from social
media may look irrelevant to the general web visitors, it is as valuable as gold
when it comes to the trained eye of a marketer. Besides, the trained marketer is
able to make good use of the large volume of data generated through all stages
of customer interaction; social media; as well as during the sales process. For
the marketing team, platforms such as Hadoop are very helpful.

Big data training helps in retrospection too where the marketing staff can gauge
how well their brand has been doing compared to competing brands; what it is
that other similar brands offer; what varying features competing brands have;
and such other information that the business can use to improve on its brand.
There is even a function in big data’s domain that the marketing team can use to
crawl competitors’ websites and do a bit of helpful text mining.

8. Department of Data Integrity, Integration and Data Warehouse

42
With the increased amount of data there is to monitor, it is crucial that you have
a team of specialists that can take in data from your various company systems.
This team will need to be current and trained on the various data in the systems,
as well as potential risks associated with them. There also needs to be folks
who know how to warehouse all this data and make structured sense of it. From
customer protection and privacy laws, the teams that work with and interpret the
data will need to know the appropriate rules for the handling of it.

9. Department of Legal and Compliance

With the new legal and compliance rules surrounding data in large
organizations, it is necessary for the Legal Department to stay aware of new
privacy and retention policies as it relates to certain pieces of data. The Legal
team and Compliance departments should work together to interpret data
privacy laws to make sure that the data is protected, stored and treated
appropriately. Businesses that don’t monitor and report their data can suffer
significant legal issues and should have policies in place to protect themselves
against potential lawsuits and breaches.

43
44
Chapter 5: Process of Data Analysis

In this chapter, I have highlighted the steps involved in the process of data
analysis. Let us look at the steps one at a time.

45
What Is Data Analysis?
Before we get on with the steps involved in the process of data analysis, let us
look at what data analysis is first. Data analysis is the process by which raw data
is collected from various sources and gets converted into meaningful
information, which can be used by various stakeholders for making better
decisions.
To put it in the words of John Tukey, a famous Statistician, data analysis is
defined as follows:
“Procedures for analyzing data, techniques for interpreting the results of such
procedures, ways of planning the gathering of data to make its analysis easier,
more precise or more accurate, and all the machinery and results of
(mathematical) statistics which apply to analyzing data.”

46
Steps Involved In Data Analysis
Even though the data requirements may not be the same for every company,
most of the below steps are common for all companies.

Step 1: Decide on the objectives

The first step in the data analysis process is the setting of objectives. It is
important that you set clear, measurable and concise objectives. These
objectives can be in the form of questions. For instance, your company’s
products are finding it difficult to get off the shelves because of a competitor’s
products. The questions that you might ask are, “Is my product overpriced?”
“What is unique about the competitor’s product?” “What is the target audience
for the competitor’s product?” “Is my process or technology redundant?”
Why is asking these questions upfront important? This is because your data
collection depends on the kind of questions you ask. For instance, if your
question is, “What is unique about the competitor’s product?” you will have to
collect feedback from the consumers about what they like in the product as well
as do an analysis on the specifications of the product. On the other hand, if your
question is, “Is my process or technology redundant?” you will have to do an
audit of the existing processes and technologies used at your establishment as
well as do a survey about the technology used by the others in the same
industry. As you can see from this, the nature of data collected differs vastly
based on the kind of questions you ask. Since data analysis is a tedious process,
it is necessary that you do not waste the time of your data science team in
collecting useless data. Ask your questions right!

Step 2: Set measurement priorities

Now that you have decided your objectives, you need to establish measurement
priorities next. This is done through the following two stages:

Decide what to measure

This is when you have to decide what kind of data you need to answer your
question. For example, if your question pertains to reducing the number of jobs,
without compromising the quality of your product or service, then the data that
you need in hand right now are:
- The number of staff employed by the company.
- The cost of employing the present number of staff.
- The percentage of time and efforts spent by the current staff on the

47
 existing processes.

Once you have the above data, you will have to ask other questions ancillary to
the primary question such as, “Are my staff not being utilized to their fullest
potential?” “Is there any process that can be altered to improve the productivity
of the staff?” “Will the company be in a position to meet increased demands in
the future despite the downsizing of manpower?”
These ancillary questions are as important as the primary objective/question.
The data collected in connection with these ancillary questions will help you in
taking better decisions.

Decide how to measure

It is highly important that you decide the parameters to measure your data
before you begin collecting it. This is because how you measure your data plays
an important role in analyzing the collected data in the later stages. Some of the
questions you need to ask yourself at this stage are:
- What is the time frame within which I should complete the analysis?
- What is the unit of measure? For instance, if your product has
international markets and you are required to determine the pricing of
your product, you need to arrive at the base price using a certain currency
and extrapolate it accordingly. In this case, choosing that base currency is
the solution.
- What are the factors that you need to include? This could again
depend on the question you have asked in stage 1. In the case of staff
downsizing question, you need to decide on what factors you need to take
into consideration with respect to the cost of employment. You need to
decide whether you will be taking the gross salary package into
consideration or the net annual salary drawn by the employee.

Step 3: Collection of data

The next step in the data analysis process is the collection of data. Now that you
have already set your priorities and measurement parameters, it will be easier
for you to collect data in a phased manner. Here are a few pointers that you need
to bear in mind before you collect data:
We already saw the different sources of data in the previous chapter.
Before you collect data, take stock of the data available. For example, in
the case of the downsizing of the staff case, to know the number of

48
 employees available, you can just look at the payroll and get the numbers.
This could save you the time of collecting this particular data again.
Similarly, collate all available information.
If you intend to collect information from external sources in the form
of a questionnaire, then spend a good amount of time in deciding the
questions that you want to ask the others. Only when you are satisfied that
the questionnaire looks satisfactory and serves your primary objective,
should you circulate it. If you keep circulating different questionnaires,
then you will have heterogeneous data in hand, which will not be possible
to compare.
Ensure that you have proper logs as and when you enter the data
collected. This could help you analyze the trends in the market. For
instance, let us assume you are conducting a survey regarding your
product over a period of two months. You will note that the shopping
habits of people change drastically during holiday seasons than any other
period of the year. When you don’t include the date and time in your data
log, you will end up with superfluous figures and this will affect your
decisions in a grave fashion.
Check how much budget is allocated for the purpose of data collection.
Based on the available budget, you will be able to identify the methods of
data collection that are cost effective. For instance, if you have a tight
budget and you still have to do a survey to gauge the preferences of your
customers, you can opt for free online survey tools as opposed to printed
questionnaires that are included in the packages. Similarly, you can make
the best use of social networking sites to conduct mini surveys and collect
the data required. On the other hand, if you have enough budget, you can
go for printed and attractive questionnaires that can be circulated along
with the package or can be distributed at retail outlets. You can set up
drop boxes at nearby cafes and malls for the customers to drop these filled
in questionnaires. You can also organize contests to collect data as well as
market your product in one go.

Step 4: Data cleaning

Now, it is not necessary that the data you have collected will be readily usable.
This is why data cleaning is very crucial in this process to ensure that
meaningless data does not find its way into the analysis stage. For example,
when you correct the spelling mistakes in the collected questionnaires and feed
it into your system, it is nothing but data cleaning. When you have junk data in
the system, it will affect the quality of your decision. For instance, let us assume
50 out of 100 people responded to your questionnaires. However, you get ten
incomplete questionnaire forms. You cannot count these ten forms for the

49
purpose of analysis. In reality, you have gotten only 40% response to your
questions and not 50%. These numbers make a big difference for the
management to make decisions.
Similarly, if you are doing a region wise survey, you need to be extra careful at
this stage. This is because most people have a tendency to not fill in their
correct addresses in the questionnaires. Hence, unless you have a fair idea about
the population of each region, you will never be able to catch these slip-ups.
Why is it important to catch these mistakes? Let’s assume that your survey
results show that a majority (say 70%) of your customer base is from X region.
In reality, the population of the region is not even close to 30% of your
customer base. Now, let us assume that you solely decide to make a decision
based on your survey results. You decide to launch an exclusive marketing drive
in this region. Of course, the marketing drive will not improve your sales
because even if all the citizens in the region buy your product, they have
contributed to only 30% of your customer base and not 70% as you imagined.
Hence, these little numbers play an important role when it comes to big and
expensive decisions.
As you can see, improving the quality of data is highly important for taking
better decisions. Since this involves a lot of time, you should automate this
process. For instance, to detect fake addresses, you can get the computer to
detect those entries that have incorrect or incomplete Zip code. This could be
easily done if you are using an Excel to store your data. Alternatively, if you
have customized software for feeding and storing data, you can get in touch
with the software designer to put certain algorithms in place to take care of such
data.

Step 5: Analysis of data

Now that you have collected the requisite data, it is time to process it. You may
resort to different techniques to analyze your data. Some of the techniques are
as follows:
Exploratory data analysis:
This is a method by which data sets are analyzed with a view to
summarize their distinct characteristics. This method was developed by
John W. Tukey. According to him, there was too much importance that
was being shown towards statistical hypothesis testing, which is nothing
but confirmatory data analysis. He felt the need to use data for the
purpose of testing hypotheses. The key objectives of exploratory data
analysis are as follows:
(i) Suggestion of hypotheses in connection with the causes of
the phenomena under question.

50
 (ii) Assessing the assumptions on which the statistical
inference will be based.
(iii) Supporting the choosing of appropriate statistical
techniques and tools.
(iv) Providing a basis for further data collection through modes
such as surveys or experiments.
Several techniques prescribed by the exploratory data analysis have been put to
extensive use in the fields of data mining and data analysis. These techniques
also form part of certain curriculum to induce statistical thinking in students. As
you perform explorative data analysis, you will be required to clean up more
data. In some cases, you will be required to collect more data to complete the
analysis. This is to ensure that the analysis is backed by meaningful and
complete data.

Descriptive statistics:
This is another method of data analysis. By this method, data is analyzed to
identify and describe the main features or characteristics of the data collected.
This is different from inferential statistics. Under inferential statistics, the data
collected is analyzed to learn more about the sample. These findings are then
extrapolated to the general population based on the sample. On the other hand,
descriptive statistics aims to only summarize and describe the data collected.
These observations about the data collected can be either quantitative or visual.
These summaries could just be the beginning of your data analysis process.
These could form the basis on which further analysis is done to process the data.
To understand better, let us look at an example. The shooting percentage in the
game of basketball is nothing but a descriptive statistic. This shooting
percentage indicates the performance of the team. It is calculated by dividing
the number of shots made, by the number of shots taken. For instance, if a
basketball player’s shooting percentage is 50%, it means that he makes one shot
in every two. Other tools used under descriptive statistics include mean, median,
mode, range, variance, standard deviation etc.

Data visualization:
As the name suggests, data visualization is nothing but the representation of
data in a visual form. This can be done with the help of tools such as plots,
informational graphics, statistical graphics, charts and tables. The objective of
data visualization is to communicate the data in an effective fashion. When you
are able to represent data effectively in a visual form, it helps in the analysis of
data and also to reason about data and evidence. Even complex data can be

51
understood and analyzed by people, when put in visual form. These visual
representations also facilitate easy comparison. For instance, if you are vested
with the job of reviewing the performance of your product and your
competitor’s product, you will be able to do so easily if all the related data are
represented in visual form. All your data team needs to do is data pertaining to
the parameters such as price; number of units sold, specifications etc, and put it
in pictorial form. This way, you will be able to assess the raw data easily. You
will also be able to establish correlation between the different parameters and
make decisions accordingly. For instance, if you notice that your price is higher
than your competitor’s and your sales are lower than your competitor’s, then
you know where the problem lies. The decreased sales can be attributed to the
increase in price. This can be set aside easily by reworking your prices.
Apart from these three major methods, you can also take the help of software
available in the market. Some of the prominent software available in the market
for the purpose of data analysis are Minitab, Stata and Visio. Let us not forget
the multipurpose Excel.

Step 6: Interpreting the results

Once you have analyzed your data, it is now time to interpret your results. Once
you have analyzed the data, here are a few questions that you need to ask:
Does the analyzed data answer your key question? If yes, how so?
If there were any objections to begin with, did your data help you
defend them? If yes, how so?
Do you think there are any limitations to your results? Are there any
angles that you haven’t considered while setting priorities?
Do you have trained people to properly interpret data?
If the analyzed data satisfies all the above questions, then your analyzed data is
final. This information can now be used for the purpose of decision-making.

Importance Of Interpreting The Data Effectively & Accurately

The importance of accurately interpreting the data cannot be emphasized
enough. Your company, website, etc. must have experienced professionals who
know how to take the organic data and interpret the results properly. For
example, let’s say your company finds it necessary to analyze data from the two
of the most popular social media platforms, Facebook and Twitter.
Your company cannot depend on an untrained professional to effectively
respond to your “likes” or “tweets” on a minute-by- minute basis. Most

52
companies today employ a Social Media Manager to manage their social
platforms. These individuals are trained to know the ins and outs of each social
platform and effectively be able to respond to your customers in a way that
represents your brand.
At the core of every successful business is accurate interpretation of vital data.
It’s necessary to hire trained professionals who have the training to take the
unstructured and otherwise random data and put it into an understandable
structure. This will change the dynamics of how your company operates and
what decisions need to be made based on the data.
People who can take the all of your customers “likes” on your corporate
Facebook page and trace the consumer’s behavior regarding the use of your
product. Follow the decision-making process of these consumers. The
consumers like your product, then what? Do they read the product description?
Do they reap the benefits of using your product? Is your product reasonably
priced in comparison to your competitor’s prices? What makes your company’s
product better than the rest of the competition?
Trained data analysts will be able to trace these questions and analyze the
pattern your consumers will take. They will follow the trace the consumers take.
They can analyze the data from the original “like” from your consumer all the
way to the purchase on your website.
The right people who have the training to follow and analyze this process can
help your company generate increased product sales by taking this information
and disseminate throughout the company to the appropriate team members.
Actually having properly interpreted meaningful data may be the difference
between your company expanding its influence or shutting down because of
misinterpreted data.
An example of how you can analyze tweets is interpreting historical tweets,
knowing the substantial “tweet” from the casual “tweet.”
The data interpreters are able to analyze historical data from previous company
“tweets” and their influence on consumer buying habits. These experts can
translate which “tweets” are substantial and which “tweets” are just social.
From the initial root message texted on twitter the analyst is able to trace the
impact on the consumer’s initial mindset as to whether they will follow the
company’s core goal to buy their product or not. Which text is more convincing
than the other one? Why is it more successful? Do images with the “tweets”
tend to convince your consumer base to buy your product? Which “tweets”
work best with what regions in the world? Which “tweets” work best with what
age group?
These are important questions that can be answered by the data and why it is
important to have analysts review and show you what marketing strategies are

53
working the best on each platform. Analysts can interpret large amounts of data
with the use of visual graphs of numerical data statistics. These can be given to
the appropriate departments to make decisions to improve the overall sales
experience for your customers.

54
55
Chapter 6: Predictive Analytics

We have seen now how data and data analysis is crucial for the effective
functioning of a business. Let us look at another wing of data mining, which
plays a vital role in the growth of a business. In this chapter, I will be taking you
through the various aspects of Predictive analytics and help you appreciate the
role played by it in facilitating the effective functioning of a business.

56
What Is Predictive Analytics?
In simple terms, predictive analytics is nothing but the art of obtaining
information from the collected data and utilizing it for predicting behavior
patterns and trends. With the help of predictive analytics, you can predict
unknown factors not just in the future but also the present and past. For
example, predictive analytics can be used to even identify the suspects of a
crime that has already been committed. It can also be used to detect a
fraud as it is being committed.

57
What Are The Types Of Predictive Analytics?
Predictive analytics can be referred to as predictive modeling. In simpler terms,
it is pairing data with predictive models and arriving at a conclusion. Let us look
at the three models of predictive analytics.

Predictive models
Predictive models are nothing but the models of the relationship between the
specific performance of certain element in a sample and few known attributes of
the sample. This model aims at assessing the likelihood that a similar element
from a different sample might exhibit the specific performance. It is widely used
in marketing. In marketing, predictive models are used to identify subtle
patterns, which are then used to identify the customers’ preference. These
models are capable of performing calculations as and when a certain transaction
is happening i.e., live transactions. For instance, it is capable of evaluating the
opportunity or risk associated with a certain transaction for a given customer,
thereby helping the customer decide if he wants to enter into the transaction or
not. Given the advancements in the speed of computing, individual agent
modeling systems have been designed to simulate human reactions or behavior
for certain scenarios.
Now, let us look at some more aspects of these models in detail. The
terminology ‘training sample’ refers to the sample units/elements that are
available, whose attributes and performances are known. The units/elements
that are present in other samples, whose attributes are known but whose
performances are unknown, are referred to as ‘out of training sample’. There is
no chronological relation between the training sample and the out of training
sample. For example, the blood splatters in a simulated environment are training
samples. On the other hand, the blood splatter from an actual crime scene is the
out of training sample. Predictive models can help in identifying the probable
suspects and modus operandi of the murder based on these samples. As
mentioned earlier, these samples are not required to be from the same time
frame. Either of the samples could be from a different time frame.

Predictive Models in relation to Crime Scenes

3D technology has brought big data to crime scene investigations to help police
departments, and criminal investigators reconstruct crime scenes without
violating the integrity of the evidence. There are two types of laser scanners
used by crime scene experts:
Time-of–flight laser scanner

58
The scanner shoots out a beam of light. It bounces off the targeted object, and
different data points are measured as the light returns to the sensor. It’s capable
of measuring 50,000 points per second.

Phase shift 3D laser scanners

These scanners are much more expensive but also much more effective. They
measure 976,000 data points per second. These scanners use infrared laser
technology.
These data laser scanners make crime scene reconstruction much easier.
Needless to say, the process takes a lot less time than traditional crime scene
reconstruction took. The advantage of 3D technology is that the investigators
can re-visit the crime scene anywhere. Investigators can now do this while they
are at home, in their offices or out in the field. This makes their job more
mobile, and they can visit the crime scenes virtually anywhere. They no longer
have to depend on notes or their memories in calling up the details of the crime
scene. Also, they visit the crime scene once, and that’s it. They have all the data
images recorded on their scanners.
Investigators are able to re-visit the crime scenes viewing the images on
computers or iPads. Distance between objects will be reviewed (like weapons.)
The beauty of this technology is crime experts don’t have to second guess their
information gathered from crime scenes. The original crime scene is constructed
right there on the scanner images. It’s as if the crime was committed right there
on the scanned images. The images tell the story about who the perpetrators
were and how they carried out the crime.
Investigators can look at the crime scenes long after the crime scenes are
released. Nothing in the crime scene will be disturbed or compromised. Any
evidence that is compromised is inadmissible in court and cannot be used. All
evidence must be left in its original state and not tampered with. This is no
problem when the evidence is recorded in the data scanners.
Law Enforcement engineers are able to reconstruct the whole crime scene in the
courtroom. Forensic evidence is untouched and left intact guaranteeing a higher
rate of convictions.

Forensic Mapping in Crime Scene Reconstruction

The purpose of 3D forensic mapping is to reconstruct every detail of the crime
scene holistically. This is a very pure way of reconstructing crime scene
evidence. None of the evidence is touched or accidentally thrown away.
Investigators don’t have to walk in or around the evidence avoiding the
possibility of anything being accidentally dropped or kicked.

59
3D data mapping allows for greater understanding and insight into the motive
and the method of the crime committed. This helps law officials to present a
convincing understanding in court so evidence will convince the jury beyond a
reasonable doubt that the crime was or wasn’t committed.
3D forensic mapping is also invaluable in training new investigators in the
process of criminal reconstruction.

Descriptive models:
Descriptive models are used to ascertain relationships to the data collected. This
is similar to how a company would classify and group its customers or its
products into certain categories. When it comes to the predictive models, it
focuses on predicting only a single customer behavior, as in the case of the
computation of the credit risk. On the other hand, descriptive models focus on
identifying different relationships between products or customers. Descriptive
models do not seek to rank customers based on their attributes or based on the
actions taken by them, as in the case of predictive models. Descriptive models
tend to categorize customers based on their product preferences. These models
can be used to build more models, which can be used to make more predictions.

Decision models:
Decision model is nothing but a system, which contains at least one action
axiom. An action axiom is nothing but the action that follows the satisfaction of
a certain condition. A model action axiom is as follows:
If <a certain fact> is true, then do <this certain action>.
In simpler words, the action axiom is used to test a certain condition. The
fulfillment of the certain condition necessitates the completion of a certain
action.
The decision model also describes the relationship between all the elements that
form part of a decision, such as the decision, the known data and also the
forecast results associated with the decision. This model aims to predict the
results of those decisions, which have many variables involved in it. Decision
models are also used to achieve optimization and maximize certain outcomes,
while minimizing certain other outcomes. These models can be used to come up
with a set of rules for a business, which will be capable of producing the
expected and desired action for every customer who avails the service of the
business.

60
61
Chapter 7: Predictive Analysis Techniques

In this chapter, let us look at the different techniques used for the purpose of
conducting predictive analytics. The two major categories into which these
techniques can be grouped into are machine learning techniques and regression
techniques. Let us look at these techniques in detail now.

62
Regression Techniques
These techniques form the foundation of predictive analytics. These techniques
aim to establish a mathematical equation, which will in turn serve as a model for
representing the interactions among the different variables in question. Based
on the circumstance, different models can be applied for performing predictive
analysis. Let us look at some of them in detail now:

Linear regression model:

This model assesses the relationship between the dependent variable in a given
situation and the set of independent variables associated with it. This is usually
expressed in the form of an equation. The dependent variable is expressed as a
linear function of the different parameters. These parameters can be adjusted in
such a way that it leads to the optimization of measure of fit. Model fitting is
required to minimize the size of residual. This importance shown towards model
fitting is required to ensure that it is distributed randomly in connection with the
model predictions.
The objective of this model is the selection of the parameters of the
model, with a view to minimize the sum of the squared residuals. This is known
as the ordinary least squares estimation. Once you have estimated the model,
you are now required to check the statistical significance of the different
coefficients used in the model. This is where the t-statistic comes into play,
whereby you test if the coefficient is different from zero. The ability of the
model to predict the dependent variable depending on the value of the other
independent variables involved can be tested by using the R2 statistic.

Discrete choice models:

Linear regression models can be used in those cases, where the dependent
variable has an unbounded range and is continuous. But there are certain cases
where the dependent variable is not continuous. In these cases, the dependent
variable is discrete. Since the assumptions related to the linear regression model
do not hold good completely in the case of discrete variables, you will have to
go for another model to conduct predictive analytics.

Logistic regression:
This model is used in those cases where the dependent variable is categorical. A
categorical variable is one, which has a fixed number of values. For instance, if
a variable can take two values at a time, it is called as a binary variable.
Categorical variables that have more than two values are referred to as

63
polytomous variables. For example, blood type of a person is an example of
polytomous variable.
Logistic regression is used to determine and measure the relationship between
the categorical variable in the equation and the other independent variables
associated with the model. This is done so by utilizing the logistic function to
estimate the probabilities. It is similar to the linear regression model. However,
it has different assumptions associated with it. There are two major differences
between the two models. They are as follows:
The Linear regression model uses a Gaussian distribution as the
conditional distribution whereas the logistic regression uses a Bernoulli
distribution as the conditional distribution.
The predicted values arrived at in the logistic regression model are
probabilities and are restricted to 0 and 1. This is because the logistic
regression model is capable of predicting the probability of certain
outcomes.

Probit regression
Probit models are used in place of logistic regression for coming up with models
for categorical variables. This is used in the cases of binary variables i.e.
categorical variables, which can take only two values. This model is popular in
economics. This method is used in economics to predict those models, which
use variables that are not only continuous but are also binary in nature. Two
important reasons why the probit regression method is often opted over logistic
regression method are as follows:
1. This is because the underlying distribution in probit regression is normal.
2. If the actual event is a binary proportion and not a binary outcome, then
the probit method is more accurate.

64
Machine Learning Techniques
Machine learning is nothing but a field in artificial intelligence, which was used
initially for the purpose of developing techniques that will facilitate computers
to learn. Since, machine learning consists of an array of statistical methods for
classification and regression, it is now being employed in different fields such
as credit card fraud detection, medical diagnostics, speech recognition, face
recognition and for analyzing the stock market. In certain applications, these
techniques can be used to predict the dependent variable, without taking into
consideration the independent variables or the underlying relationships between
these variables. In other words, machine-learning techniques can be readily
employed in those cases where the underlying relationships between the
dependent and independent variables is complex or unknown. Some of the
techniques used under this method are as follows:

Neural networks
Neural networks are nothing but sophisticated and non-linear modeling
techniques, which are capable of modeling complex functions. These techniques
are widely used in the fields of cognitive psychology, neuroscience, finance,
physics, engineering and medicine.
Neural networks are generally employed when you are not aware of the exact
relationship that exists between the inputs and output. However, they learn the
underlying relationship by means of training. There are three kinds of trainings
that form part of neural networks namely supervised training, unsupervised
training and reinforcement learning. Of the three, supervised training is the most
commonly used.
A few examples of training techniques used in neural networks are as follows:
- Quick propagation
- Conjugate gradient descent
- Backpropagation
- Projection operator
- Delta bar delta

Multilayer perceptron
The multilayer perceptron is made of an input layer and output layer. Over and
above these, there are one or more hidden layers made up of sigmoid nodes or
nonlinearly activating nodes. The weight vector plays an important role in
adjusting the weights of the network. The backpropagation technique is

65
employed to minimize the squared error, which usually arises between the
network output values and the expected values for the output.

Radial basis functions

A radial basis function is a function that has an inbuilt distance criterion in
connection to a center. These functions are typically used for the interpolation
and smoothing of data. These functions have also been used as part of the neural
networks instead of the sigmoid functions. In such cases, the network has three
layers namely, the input layer, the hidden layer with the radial basis functions
and the output layer.

Support vector machines

Support vector machines are employed for the purpose of detecting and
exploiting complex patterns in data. This is achieved by clustering, classifying
and ranking of data. These are nothing but learning machines, which can be
used for the purpose of performing regression estimations and binary
classifications. There are several types of support vector machines such as
polynomial, linear, sigmoid, to name a few.

Naïve bayes
This is based on the Bayes conditional probability rule. This is an easy
technique using which classifiers are constructed. In other words, they are used
for the purpose of classifying various tasks. This technique is based on the
assumption that the predictors are statistically independent. It is this
independence that makes it a great tool for the purpose of classification.
Moreover, this technique is also easier to interpret. In those cases when the
number of predictors is very high, it is wise to use this method.

K-nearest neighbors
The nearest neighbor algorithm forms part of the pattern recognition statistical
methods. Under this method, there are no underlying assumptions associated
with the distribution from which the sample is drawn. It consists of a training set
and has both positive and negative values. When a new sample is drawn, it is
classified based on its distance from the nearest neighboring training set. The
sign associated with that point also plays an important role in classifying the
new sample. While using the k nearest neighbor classifier, all the k- nearest
points are considered. The sign of majority of these points are then used for the
purpose of classifying the new sample.

66
The performance of this algorithm is determined by the following factors:
1. The distance measure that is used for the purpose of locating the nearest
neighbors.
2. The decision rule that is used for the purpose of deriving a classification
from the k- nearest neighbors.
3. The number of neighbors, which are being considered for the purpose of
classifying the new sample.

Geospatial predictive modeling

The underlying principle of this technique is the assumption that the
occurrences of events, which are being modeled, are limited in terms of
distribution. In other words, the occurrences of events are neither random nor
uniform in distribution. Instead, there are other spatial environment factors
involved, such as socio cultural, infrastructure, topographic etc, which are
capable of constraining and influencing the locations of the occurrences of these
events. This is a process for assessing events using a geographic filter, with a
view to come up with statements of likelihood for the occurrence of a certain
event.
There are two kinds of geospatial predictive models. They are as follows:
1. Deductive method
2. Inductive method

Let us look at these methods in detail now.

1. Deductive method:
This method is based on subject matter expert or qualitative data. This
data is then used for the purpose of describing the relationship that exists
between the occurrence of an event and the factors associated with the
environment. In other words, this method relies majorly on subjective
information. The limitation of this model lies in the fact that the number
of factors that are being keyed in is completely dependent on the modeler.

2. Inductive method:
This method is based on the spatial relationship (which is empirically
calculated) that exists between the occurrences of events and the factors
that are associated with the environment. Every occurrence of an event is
first plotted in the geographic space. Following the plotting, a quantitative

67
 relationship is then defined between the occurrence and the factors
associated with the environment. An important reason why this method is
highly effective is that you can develop software to empirically discover.
This is crucial when you have hundreds of factors involved and there are
both known and unknown correlations existing between events and
factors. The quantitative relationship values, as defined before, are then
processed, with the help of a statistical function. These values are
processed to figure out spatial patterns, which are capable of defining
areas that are not only highly probable for the occurrence of an event but
also those areas which are less probable locations for the occurrence of an
event.

Hitachi’s Predictive Analytic System

How would you like to take a look into the future? This is what Hitachi’s
Predictive Analytic System (PCA) will achieve if it’s successful in its
unveiling. Their technology will revolutionize the way police departments
handle crimes. The present method is labeled the “break/fix model” when
police arrive on the scene after a crime has been committed. A death has
taken place, rape, domestic violence or a burglary. Then the police have
to administer damage control. Sometimes the police can get to the scene
before the crime is committed or while the crime is being committed.
Then they can stop it.

What if they could arrive on the scene before even the crime is
committed? How many lives would be saved? People could be spared
bodily pain from being abused or beaten up. Costly damages from
property loss could be reduced or stopped. There are so many benefits
that would be derived if the PCA is successful in its intended purpose.

The Pac would be able to predict crimes before they happened. This
would enable police departments to arrive on the scene before the crime is
committed. Yes, the police would arrive on the scene before the crime is
committed. Instead of taking the quirks of traditional crime predicting
methods where investigators use their own experiences (the product of
their personal variables) to project the committing of crimes. PCA will
eliminate countless variables to accurately analyze a multitude of factors
that will affect crime. Here are the components PAC analyzes and will
predict in real time the committing of a crime before it happens:
Weather Patterns

68
 Public Transit Movements
Social Media Actions
Gun Shot Sensors
Many other Factors

The above data sets are fed into the system and over a two-week period
the Pac will determine if there is a correlation between the said datasets.
Hitachi will allow different Law Enforcement agencies to test their
systems in the real world working environments. These “experiments”
will be held in different agency locations in unknown cities. The agencies
will participate in what is known as “Double-blind trial.” The predictive
system will be running in the background but criminal experts will not be
able to see the predictive results as they happen.

Summation of the testing period will be when Hitachi will compare and
analyze the results of the predictions in relationship to the actual daily
police activities over the same period.

Benefits of PCA

The benefits of the PCA being successful would be countless:

Obviously, countless lives would be saved
Law Enforcement Agencies could still work efficiently in the light of
continual manpower cuts.
Officers can be placed strategically in locations to stop potential
crimes
Officers would exponentially improve their crime-fighting time
exponentially
Law Enforcement Training costs would dramatically drop
Law enforcement experts would in less life-threatening circumstances
Homicides, rapes, domestic violence cases would drop
Stolen property losses would be noticeably reduced
Crime prevention success would dramatically increase

69
 Predictive Analytics influencing many Insurance industries
It has been researched that across the many parts of the insurance industry
they have switched to applying predictive analytics to their daily business
practices. 49% of the Personal Auto Insurance industry now uses
predictive analytics to:
Profit increase
Risk reduction
Revenue growth
Improve overall company operational efficiency

Big data and specifically predictive analytics are changing the way insurance
companies conduct business and relate to their customers. The trend now is to
give their customers the product they want at a price they want. Also, the
insurance companies have become more sensitive to their customer’s core
issues. Competition has increased in this key industry and companies can no
longer say, “You don’t like our prices? Too bad.” Insurance companies can no
longer ignore the wishes of their clientele.
If the insured doesn’t like the prices, their company is offering them they will
go to the competition who will offer them a better insurance rate. How does
predictive analytics help in these causes? Well, insurance companies are now
offering telemetry-based packages to consumers. These packages are based on
predictive analytics. The insurance company uses predictive modeling to form a
risk management profile on the customer. They take all the risk factors
connected with this particular person:
Likelihood of the person getting in an accident
Will their car be stolen?

By comparing this behavioral information with thousands of other profiles the

insurance company can compile an accurate assessment on the risk factor of this
individual. In turn, the insurer can offer the insured a reasonably priced
premium. The data is transmitted from a box located in the car or from the
person’s smart phone app. This process helps insurance companies to bring
value to their customers and stay in touch with their needs. The whole point is
for the companies to offer better services to their customer base.
One insurance company, US insurers Progressive, is an example of a company
using data to enhance their customer services. They have developed what they
call “Business Innovation Garage” where their technicians will road test these

70
latest innovations. One example is they render images of damaged vehicles
from computer graphics. The images are sent from cameras and constructed into
3D models showing the condition and extent of damage to the vehicle. This will
speed up the claim process. It will enable the insurance company to process the
paperwork quickly. They will have the condition of the car in real time. The
customer will get the estimate quicker, and he will be able to get his car into the
body shop that much quicker. The customer’s car will back on the road in good
working condition in no time. This will definitely help this insurance company
increase their customer base.
There are still other methods of conducting predictive analysis, which I have not
mentioned in this chapter. With these techniques, it is possible to reap the
multiple benefits of predictive analytics.

71
72
Chapter 8: Why Predictive Analytics?

Now that you have a fair idea about predictive analytics, let us now understand
why it has gained increased importance over the years. Predictive analytics has
found its way into several fields and applications and has had a positive impact
so far. In this chapter, I have highlighted the key applications, which employ
Predictive analytics.

73
Analytical Customer Relationship Management (CRM)
This is one of the famous applications that employ predictive analytics.
Different methods of predictive analytics are used on the customer data that is
available with the company, with a view to achieve the CRM objectives. The
CRM aims to create a holistic view of the customer, irrespective of where the
information about the customer lies. Predictive analytics is also used in CRM
for sales, customer services and marketing campaigns. The different methods of
predictive analytics help the company in meeting the requirements of its huge
customer base and making them feel satisfied. Some areas how predictive
analytics is used in CRM are as below:
- Analyzing and identifying those products of the company that have
the maximum demand.
- Analyzing and identifying those products of the company, which will
have increased demand in the future.
- Predicting the buying habits of customers. This will help the
company in the promotion of several of its other products across multiple
areas.
- Proactive identification of pain points, which may result in the loss of
a customer and mitigating such instances.

CRM can be used throughout the lifecycle of a customer, starting from

acquisition and leading to relationship growth, retention and win back.

74
Clinical Decision Support Systems
Predictive analytics is extensively used in health care for calculating the risk of
contracting certain diseases or disorders. It is primarily used to determine the
patient’s risk of developing health conditions such as heart diseases, asthma,
diabetes and other chronic diseases. If you were under the impression that
predictive analytics is used only for the purpose of diagnostic reasons, then you
are wrong. Predictive analytics is also employed by doctors for making
decisions regarding a certain patient who is under medical care at a certain point
of time. How does it work? Clinical decision support systems aim at linking the
observations of a certain patient’s health with the knowledge about health. This
relationship will aid clinicians in taking the right decisions with respect to the
patient’s health.
Here are some other ways predictive analysis is helping the healthcare industry:
Predictive analytics is influencing the healthcare industry. It could revolutionize
Healthcare worldwide. Here are seven ways of how it will impact healthcare:
Predictive analytics increases medical diagnosis accuracy
Predictive analytics will support public health and preventive care.
Predictive analytics gives doctor’s solutions for their patients.
Predictive analytics can provide predictions for hospitals and employers
about insurance product costs.
Predictive analytics lets researchers make prediction models without
studying thousands of patient cases. The models will get more accurate
over time.
Predictive analytics help pharmaceutical companies develop the best
medicines for the welfare of the public.
Predictive analytics provide patients with potentially better health
outcomes.

These seven points can have industry changing practices for the healthcare
industry. With the ability to predict medical conditions with greater accuracy
could save lives and medical expenses for patients. Malpractice suits will
decrease, and potentially healthcare costs could decrease. Doctors being able to
accurately diagnose illnesses can save patients and insurance companies
thousands of dollars in individual cases. Potentially insurance companies could
save untold millions in medical claims worldwide.
Imagine doctor’s interpreting patient’s sickness right the first time. There would
be a domino effect. This should drop healthcare costs dramatically. People will

75
be able to afford healthcare insurance. Doctors can lower their rates and live in
peace knowing they won’t be sued for malpractice. Patients would save money
because they wouldn’t be spending money needlessly on prescriptions that
weren’t connected to their illnesses. Healthcare would be streamlined in the
patient care process if doctors have the ability to use predictive analytics to give
patients better care based on accurately researched information.
Elderly patients would be able to afford their medications and in some cases
they are administered many prescription that could cost them thousands of
dollars monthly. Pharmaceutical companies would use predictive analytics to
research and manufacture medicines to better meet the needs of the public. It
could be possible to reduce the inventory of existing drugs worldwide. How?
Pharmaceutical companies can produce necessary medicines that would directly
deal with particular conditions patients may have. By using predictive analytics
to find accurate data about patient’s health issues the companies can produce the
exact medicines that will help cure those issues. If this practice were done on a
wide-scale, eventually, unnecessary medicine thought to help treat conditions
but didn’t, could be removed from the market. Pharmacies, in turn, would be
left with only medicines that were needed to treat various diseases, etc.
Back on the accurate diagnosis trend, this could be huge in medical care.
Imagine a patient has a family history of having heart attacks, and their
ancestors were prone to having heart conditions. The patient’s doctor uses
predictive analytics to predict the chances of the patient having a heart attack in
the same age range as other family members have had. More to the point what if
the patient is at the age where the risk of heart attack is imminent.
The patient’s physician is aware of all these facts. The doctor can use predictive
analytics to predict when the event of a heart attack could occur. Armed with
this life-threatening knowledge the doctor shares the information with the
patient. The doctor and the patient develop a health plan to reduce the risk of the
heart attack or eliminate the risk all together. The patient may have extended
their life for several years because of this health prediction. Better yet the
patient may have saved their life in the here and now. These are the kind of
radical changes predictive analytics can bring to the health care industry.
It is reasonable to assume the roles of the doctor and patient will change based
on the new health data predictive analytics has/will bring to medicine. The
patient being more aware of their care options will increasingly make health
care decisions for themselves. Doctors will become more of a consultant in
advising the patient on the best course of action regarding their health choices.
It seems as if that trend is already taking place.
Today more than ever patients are aware of their health issues and options more
than ever before. They are getting more involved in the direct decision-making
process about their personal healthcare. Physicians seem to rely more on the

76
patient’s awareness about health care plans than before.
These are just the beginning of wide sweeping changes that have to occur in the
healthcare industry because of the innovation of predictive analytics.

77
Collection Analysis
There are several industries that have the risk of the customers not paying the
payments on time. A classic example of this case would be banks and other
financial institutions. In such cases, the financial institutions have no choice but
to engage the services of collection teams to recover the payments from the
defaulting customers. It is not surprising that there will be certain customers
who will never pay despite the efforts of the collection teams. This is nothing
but a waste of collection efforts for the financial institutions. Where does
predictive analytics come into this? Predictive analytics will help the institution
in optimizing the collection efforts. Predictive analytics play an important role
in collection activities in the following manner:
- Optimizing the allocation of resources for the purpose of collection.
- Identification of collection agencies, which are highly effective.
- Formulation of collection strategies.
- Identifying those customers, against whom legal action is required to
be taken.
- Formulation of customized collection strategies.

With the help of predictive analytics, it is easier for the financial institutions to
collect their dues in an effective and efficient manner. This also reduces the
collection costs for the financial institutions.

78
Cross Sell
This is applicable for those organizations, which sell multiple products. As I
mentioned before, it is important that every organization has details of its
customer base. When an organization has a comprehensive database of the
customers, predictive analytics will help in the promotion of the other products
of the organization. Predictive analytics helps in determining the spending
capacity of each customer, their usage and other behavior, which makes it easy
for the organization to promote other related products. This cannot only just
lead to the increase in sales for the organization but also helps in building
customer relationship.

79
Customer Retention
Customer satisfaction and customer loyalty are two of the major objectives of
any organization. These objectives have gained more importance in the recent
past, given the increase in the competition. Apart from these two major goals,
another goal has emerged in the light of such heavy competition in the market,
which is nothing but reducing customer attrition. It is not important if new
customers purchase an organization’s product or services. What is more
important is that existing customers continue buying their products or services.
When you are capable of retaining your customers, you are bound to increase
your profits without much difficulty. Hence, it is highly important that each
organization pays attention to the needs of the existing customers.
Most businesses tend to react to the customers’ needs and not proactively
address them. This attitude could easily break them, should their competitor be
more empathetic towards the needs of the customers. Moreover, it is too late to
change a customer’s decision once he or she has decided to discontinue the
service of an organization. No matter what the organization says or does, may
not have that much of an impact on the customer. Ultimately, the organization
will have to bend backwards to retain the customer, which is not only an added
cost to the company but also an embarrassment.
With predictive analytics, it will be possible for organizations to be more
proactive as opposed to being reactive. How does it work? With just a quick
glance at the customer’s service usage history and spending habits, on a
frequent basis, it will be possible for the organizations to come up with
predictive models. These predictive models will help in the identification of
those customers who are most likely to discontinue the service of the
organization. This gives the organization an opportunity to act proactively and
figure out the reason behind the possible termination of service. It can also
decrease the number of customers terminating its service in this fashion. Some
strategies used by organizations to retain the customers who are on the verge of
leaving include but are not limited to lucrative offers, discounts etc.
Another kind of customer attrition that is equally detrimental to the interests of a
business is silent attrition. This is when a customer reduces his purchase of the
organization’s products or services, gradually over a period of time and
discontinues eventually. Most companies are not aware of customers who have
shifted loyalties this way. However, with the help of predictive analytics, it will
be possible to pick out such customers who are exhibiting this behavior. As I
mentioned before, predictive analytics mandates the need to monitor the
customer’s service history and expenditure regularly. This will help the
companies in spotting those customers who have reduced their spending over a
period of time. With this information in hand, it will be possible for the
company to come up with a customized marketing strategy for such customers

80
and ensure that they continue availing the services of the organization.
Let’s introduce some case studies to support the claim that predictive analytics
can help increase customer retention and reduce “churning.” Churning is where
the customer stops buying products from a store or business. They begin
shopping at another retailer’s website or physical store.
Here are the case studies: Windsor Circle’s Predictive Analytical Suite produces
many predictive data fields to help vendors in their e-mail marketing
campaigns:
Predicted Order date–based on customers personal purchasing habits
Replenishment Re-Order date, when the product they purchased will run
out
Product Recommendation–related to products customers have purchased
historically
Hot Combo-Products that are typically bought together.

The goal in predictive analytics in direct e-mail marketing is to combine

Predicted Order Dates, Replenishment Re-Order Dates with product
Recommendations and Hot Combo sales to retain customers by appealing to
their predicted spending patterns. The vendors entice customers with future Hot
Combo sells by reducing prices and predicting when those Hot Combo sells will
sell out.
By using Windsor Circle’s Predictive Analytics Suite, vendors can create
models predicting which customers will buy which products in what particular
time frames. Based on historical data showing previous product purchases;
vendors can piece together effective direct e-mail marketing strategies to induce
customers to keep coming back.
These predictive models can also predict when customers will purchase certain
products together and which products they most likely would buy together,
another component of these predictive models is they can foretell the right
timeframes to keep products in inventory based on customer demand in the past.
Coffee for Less.com uses Windsor Suite’s predictive data to run their
replenishment e-mails to remind customers to stock up on their coffee
purchases. Windsor automatically pulls up a predicted order date on customers
who have made three or more purchases. They then trigger replenishment e-
mails and send it to the customer. The e-mail is based on the customer’s
historical buying patterns. In this e-mail example, they would put a smart image
to the left. This would then populate a particular product image based on the
products the customer was getting the replenishment e-mail about. Shrewdly

81
enough Coffee for Less will use the rest of the e-mail page to suggest other
products the customer will like. This e-mail is generated on predicted customer
products they will most likely buy. So here Coffee for Less is using two data
marketing strategies to retain this particular customer. They are Product
Recommendations + Predicted Order dates all based on Windsor’s predictive
analytic models.
Normally, vendors use Windsor’s predicted order date field to trigger product
recommendations and replenishment e-mails. Australia’s premiere surf and
apparel retailer Surf Stitch used the predicted order date field in an innovative
way. They use it to locate potential “churning” customers. (Defined previously
as customers who will stop buying products from a particular vendor.) The
usual timeframe for win-back e-mails to be sent out is in 60, 90 and 120-day
cycles since the last customer purchase.
Surf Stitch has access to their customer’s purchase history and their predicted
order data. By using the replenishment e-mails in an innovative way (win-
backs) they were able to reduce churning customers by a whopping 72% over a
six-month period.
The case studies above show that companies (in many industries) can use
predictive analytics in an innovative way. They can take standard data fields and
use them in new ways to retain their customer base and win-back lost
customers. Predictive analytics are so flexible in their data applications they can
be used in thousands, potentially millions of ways. There are applications
waiting to be discovered that no company has thought about using.
The possibilities that predictive analytics can/will have on company’s business
practices are unlimited. Sadly, though there are still companies in various
industries that are lagging behind in using predictive analysis models. They are
still using traditional predictive analytic methods that are no longer effective.
These old methods are slow, cumbersome and not cost effective. They take too
much manpower and time to facilitate all their processes. They slow down
revenue sources, production and overall operational efficiency of these
companies.
The old predictive analytical models aren’t user-friendly and can’t predict as
many different variables as the new methods can. In the long run, the newer
predictive systems will eventually be implemented in all the companies that are
currently not using them. What will motivate the change? When the companies
see that there traditional predictive analytical models are increasingly costing
them their market share to their competitors. Internal processes will not be
streamlined and run more efficiently. Their bottom line will not increase but
decrease because they aren’t using the newer predictive analytical models to
their fullest capacity. They aren’t using them at all!

82
Direct Marketing
For organizations that manufacture consumer products or provide consumer
services, marketing is always an important matter of discussion. This is because,
when marketing, the team should not just consider the pros and cons of their
organization’s products or services, they also have to take into consideration the
competitor’s marketing strategies and products.
With predictive analytics, this job is made easy. It can improve your direct
marketing strategies in the following manner:
- Identification of prospective customers.
- Identification of the combination of product versions, which are the
most effective.
- Identification of effective marketing material.
- Identification of communication channels, which are highly effective.
- Determining the timing of marketing strategies to ensure that it
reaches the maximum audience.
- Reducing the cost per purchase. Cost per purchase is calculated by
dividing the total marketing cost incurred by the company divided by the
number of orders.
Insurance companies are using predictive analytics to develop personal
marketing strategies to attract potential customers. They will analyze many
sources of data to find out what customers like. They will analyze customer
feedback data, which involves scouring their own websites and social media
sites. Algorithms scope through unstructured data in phone calls, e-mails, and
information revealed on social media about what customers like or dislike. Also,
the insurance company will analyze how much time a person will spend on the
FAQ section of their website. Even forums and message boards are scanned to
see what a customer’s preferences are. All is this done so the Insurance
Company can make a unique profile for the potential customer.
I mentioned this in another section of this book that insurance companies are
now using all data available to them. They need to focus individually on each
customer’s needs to stay competitive actually to stay in business. Insurance
companies have always focused in on whether they are meeting the customer’s
needs or not in the past. But now they are doing this more aggressively through
the new data analysis techniques available to them. Clearly they are attempting
to determine if they are having a positive impact on their customers or not.
Marketing departments of the Insurance Companies are also using predictive
analytics to track if customers will terminate their policies or not. Like other
tracing processes the company will investigate a customer’s behavior and

83
compare that behavior to other customer profiles who have actually canceled
their insurance policies. This will help determine or predict whether a customer
will cancel in the near future. One pattern data source that is analyzed is
whether a customer has made a high or low amount of calls to the helpline. If
the numbers are especially high or low, it will be flagged.
The flagging will alert the company to try and change the customers mind about
canceling his policy. They might offer lower premiums or discounts on other
services to get the customer to stay and having this data to make those calls
before they cancel is critical. More importantly, they can spend more time
solving the customer’s issues before they happen, so predictive analytics is
playing a crucial role in marketing. Insurance companies are now zeroing in on
more customized customer profiles.
Summing it up big data is helping the insurance industry to improve its
customer service, fraud losses and lower the price of its premiums. A side notes
the FBI reports on average that customers pay $400-$700 more in premium
costs because of fraud losses.
This is how predictive analytics plays an important role in direct marketing.

84
Fraud Detection
As I mentioned earlier in this book, fraud is an inherent threat for every
organization. Lack of data makes it difficult for a company to even detect fraud.
Fraud can be of the following types:
- Fraudulent transactions. This is not just restricted to online
transactions but also involves offline transactions.
- Identity thefts.
- Credit applications, which are inaccurate.
- False insurance claims.
As I said before, the size of the organization does not shield it from fraud.
Whether it is a small business or big corporate, both stand the risk of being
victims to fraud. Classic examples of organizations, which are often victims of
fraud, are retail merchants, insurance companies, suppliers, service providers
and credit card companies. Predictive analytics can help these kinds of
organizations to come up with models, which will be capable of detecting
incorrect data or applicants and thereby reduce the company’s risk of being a
victim to fraud.
Predictive analytics plays an important role in detecting fraud, even in the
private and public sector. Mark Nigrini developed a risk scoring method that is
capable of identifying audit targets. He employed this method to detect fraud in
the franchisee sales reports associated with an international food chain. Each
franchisee is scored 10 predictors at the outset. Weights are then added to these
initial scores, which will give you the overall risk score for every franchisee.
These scores will help in identifying those franchisees that are more prone to
fraud than the others.
This approach can be used to identify travel agents, who are fraudulent,
questionable vendors and other accounts. There are certain complex models,
which are developed using predictive analytics, which are capable of submitting
monthly reports on the fraud committed. In fact, the Internal Revenue Service of
the United States of America makes use of predictive analytics to keep a tab on
tax returns and identify any fraudulent entries or evasion and also catch tax
fraud early.
Another growing concern when we speak about fraud is cyber security.
Predictive analytics can help you come up with models based on behaviors,
which are capable of examining actions on a specified network, continuously,
and spot any deviations or abnormalities.
Widespread fraud is a serious challenge for the insurance industry. But
companies are fighting fraud on the front lines with profiling and predictive

85
analytics. The companies will compare current claims with claims that have
been historically found to be fraudulent. Certain variables will be picked up by
computer analysis that indicates a claim that is being filed is a fake. The fake
historical claim has certain markings to show it is a fraudulent document. The
claims are matched up, and the claim in question has the same markings as the
historical fraudulent claim then it will be investigated further. The matches
could be the behavior of the claimant. Criminals of fraud typically display
certain patterns that can be detected by computer analysis. These might not get
flagged by a human who manually these process claims, but computers can look
at a volume of claims for triggers and alert the appropriate people of the
occurrence for further review.
The other factors the company will look at are the people the claimant
associates with. These are investigated through open sources, like their activity
on social media, or they will check the other partners in the claim like an auto
body shop. The insurance company will be looking at dishonest behavior the
auto body shop may have engaged in the past. Credit reference agencies are
checked out as well.

86
Operations
Sometimes, analysis can be employed for the purpose of studying the
organization, its products or services, the portfolio and even the economy. This
is done to identify potential areas of business. Businesses use these predictive
models to even manage factory resources as well as for forecasting inventory
requirements. For example, airline companies use predictive analytics to decide
the number of tickets that they are required to sell at a specified price range for
every flight. Similarly, hotels may use predictive models to arrive at the number
of guests they may be expecting during a given night. This way, they can alter
their price range accordingly and increase their revenue.
For instance, if the Federal Reserve Board is keen on forecasting the rate of
unemployment for the upcoming year, it can be done so using predictive
analytics.

Insurance Operations
Today’s Insurance giants seem to be behind the times in their approach on how
they use predictive analytic methods. The current methods employ many data
analysts to run it. The method is too time consuming and it cannot handle the
massive amount of data flying through insurance companies data portals to be
effective systems. When you have data funneling through many sources, it
makes it improbable for Insurance Companies to analyze data in a quick and
efficient manner. These two factors are a must if the company wants to stay
abreast a highly changing market place.
Companies to be relevant must adopt predictive analytic models that will turn
business value-specific decisions and actions which will maximize operations
across the board. Solutions must continuously improve and refine needing
thousands of iterations. Analytic output must connect into business operations to
optimize revenues and be relevant. Modern Analytics provides the following
predictive analytic features to Insurance Companies who want improve business
operations:
Predictive analytics specifically for insurance
Predictive Modeling for Insurance
Fully operational predictive analytics
Big Data analytics
Prescriptive Analytics
Customized insight and knowledgeable answers

87
Insurance Companies must realize that proper predictive analysis will improve
overall business operations. Unfortunately, there is this mindset within the
insurance industry that current predictive models are not efficient. They worked
in former times but in today’s market they’ve become obsolete and will
eventually hurt the overall business operations revenue sources, resource
optimization and sales nurturing of the Insurance Companies.
They need to have an automated predictive model in place to keep up with the
influx of data constantly bombarding their systems now. The manpower labor is
too costly to continue to use manual predictive analytic methods. These
resources can be channeled into automated systems saving the companies a lot
of time and money. True with the advent of automated predictive models you
would have to hire additional technical support people who understand the
complexities of running these automated systems. But the revenue increase in
all the other areas will more than justify the manpower expense. Reluctance on
the part of analysts and statisticians to fully automate predictive processes is too
costly for Insurance Companies to continue operating in.
The predictive analytic model that Modern Analytics offers will be the very
model Insurance Companies need to increase sales, decrease fraud and increase
their policyholder’s market instead of losing them. Modern Analytics fully
automated predictive systems allows them to run thousands of models at one
time. Their methodologies run much faster than traditional analytic systems can.
Plus, they will assist insurance companies to increase overall productivity,
increase daily operations, effectively locate strategic business investments and
predict any changes in the current and future marketplace.
Their predictive analytic models will be game changers for any Insurance
Company that chooses to implement their services. How? By optimizing
business operations, improve internal processes and surpass competitors in
predictive analytic processes. Modern Analytics works very closely with their
clients, and they have helped a wide base of customers across many industries.
They gather and structure data with their cutting-edge algorithms and cutting
edge technology. They also rapidly supply top-notch data solutions uniquely
tailored for each client.

Shipping Industry
UPS has adopted a new program called (On-Road Integrated Optimization and
Navigation) Orion for short. UPS believes that Predictive Analytics needs to be
implemented all the way down to the front line workforce. UPS is deploying
this Predictive Analytics model to its front-line supervisors and drivers totaling
55,000 in all. The predictive model factors in the following variables:
Employee Work Rules

88
 Business Rules
Map Data
Customer Information

These are used to predict the optimal delivery routes for the UPS drivers. The
data sources will answer such questions like “
What is better to make a premium delivery 15 minutes earlier or is it better to
save a mile of driving, to reach peak performance objectives. UPS is taking
innovative steps to bring their predictive model to the execution. The front-line
supervisors will learn to interpret predictive analytics data to maximize their
delivery times on their routes, that is the key behind this move. Within UPS,
these data skills are no longer just used by analysts.
UPS originally distributed the ORION system in 2008. The problem was, there
were too many variables for the managers to manipulate. It confused them
because they didn’t understand how these variables applied to them. UPS
refined ORION and now only the essential variables have to be input by
managers. The managers teach the drivers how to interpret the ORION
predictive data. The data was put into terms the front-line could understand.
Understand they did one driver over a few months cut 30 miles off his route
interpreting the particular predictive data for his route.

Risk Management
Predictive analytics is employed to assess the risk associated with a business, in
a fashion similar to how it is used to detect fraud. Credit scoring is the crux of
ascertaining the risks associated with the business. Credit scores can determine
the likelihood of a customer to default in payments. How is this done so? The
credit score is nothing but a number, which is generated by a predictive model.
This model consists of all data associated with the creditworthiness of the
customer in question.

Employee Risk
There is another side of Risk Management that is typically overlooked. The risk
associated with a company’s employees. Companies can use predictive analysis
to gauge employees who may be a risk. There are two traditional methods in
place to detect employees who could become troublesome. They are the
“blanketed management” programs that focus more on non-risk employees and
the employees at risk slip into by without being noticed. The other method is the
“squeaky wheel” approach when company management focuses in on

89
employees that display active troublesome behaviors.
Another issue company’s face is if they have thousands of employees how can
they effectively monitor for employees at risk? They can’t use traditional
methods. So here is where predictive analytics comes into focus. Predictive
analytics will provide specific based risk management programs. They use
thousands of data points that are turned into tangible data. This data tracks
subtle but profound changes in an employee’s behavior that could lead to major
behavioral issues down the road. Often these changes otherwise go undetected
even with the staff of Risk Management employed. Previous trends are also
analyzed. These components are interpreted to predict future outcomes of an
employee’s unusual behavior. This information arms the managers and gives
them the ability to intervene in the employee’s issues before their issues impact
the business.
Dramatically undetected risk could cause employees to engage in tragic
behavior. So, the intervention of the manager early can sometimes avoid serious
negative behavior down the road. The managers can intervene with the right
person, at the right time and on the right subject. These interventions can stop
worker compensation claims and voluntary employee turnover.
Their blanket management methods are replaced with these streamlined task-
specific models, which will save time, money and most importantly reduce risk.
The value of these predictive models is they can historically analyze employees
past behaviors to identify present and future behavioral patterns.
I need to point out here that when employees are hired in the beginning, they
aren’t at risk. It’s events after their hiring that develop that put them at risk.
Problems in the employee’s lives can surface outside of the workplace. They
could be having medical or personal issues that carry over into their
performance at work. Again, as stated earlier predictive analytics will spot
subtle changes in an employee’s behavior from historical analyzing of their
records.
Let’s briefly talk about the transportation industry in relation to Risk
Management. This industry is highly vulnerable to risk with its employees. The
transportation uses predictive analytics to track unsafe driving patterns of its
employees. The data reveals that the employee may be driving unsafely which
could lead to ample violations, fines possibly costly losses due to accidents.
People may even be killed in an auto or other type of vehicle accident.
Specifically, one type of hazardous driving pattern could be the employee not
paying attention to his driving because he has a sick elderly parent. He is too
busy talking on the phone while driving, setting up doctor’s appointments and
other activities for this sick parent.
Based on historical data patterns the transportation manager can detect if the
employee is driving abnormally. They could be braking extra hard or spending

90
too much time in idle. The manager can communicate with the employee in
question to work out some type of solution to help the employee deal with their
issues to avoid a costly accident, such as a reduced workload or adjusted
schedule.
Predictive analytics can help a manager focus on why an incident happened,
instead of what happened. Managers tend to focus on what happened, instead of
why it happened. The manager’s focus is directed on the correcting the issue
and not necessarily the root cause. Changing focus on the incident can solve tie
employee’s personal issues much quicker.

91
Underwriting
Many businesses will have to account for their exposure to risk. This is because
of the different services offered by the organization. They will have to estimate
the cost associated for covering these risks. For instance, insurance companies
that provide insurance for automobiles need to determine the exact premium to
be charged for the purpose of covering every automobile and driver. In the case
of a financial company, they will need to analyze the potential and the repaying
capacity of a borrower, before they grant the loan. In the case of a health
insurance provider, predictive analytics can be employed to figure out the
following information – past medical bills of the person, pharmacy records etc.
This information can be used to predict the quantum of bills that the person may
submit in the years to come. The health insurance provider can arrive at a
suitable plan and premium.
Predictive analytics can help you underwrite these risks. With the help of
predictive analytics, it is possible to determine the risk behavior of a certain
customer as and when they make an application. This can help these companies
take a calculated decision.
Predictive analytics have also reduced the turnaround time for processing loans
and other claims, based on the credit scores. The reduced turnaround time has
made a big difference in the case of mortgages. These decisions are now taken
in a matter of few hours. This is different from the erstwhile procedure whereby
it used to take weeks for the financial institutions to offer a mortgage. Predictive
analytics also helps in the pricing of these financial products, which can reduce
the chances of default, for often the most common reason for default is the fact
that the interest rates are high or the premium is high.
Recently the insurance industry has turned to predictive analysis to help solve
the issues they face. They are using this technology to analyze credit scores to
foretell loss-ratio performance and the behavior of the insured. Now insurance
carriers are using predictive analytics to see into the future during underwriting.
They are creating models to see how future policies will perform. They face a
very serious issue where they really have no reliable data on homes they insure.
There is a big risk that the houses they insure could decrease in value in the
market. Or what if the house burns down? Is the house vulnerable to earthquake
damage? Is the house located in a potential flood zone? These are the types of
questions that predictive analysis will help them answer.
With unknown data about houses they insure, they are at risk of losing millions
of dollars in natural disaster damage or weather damage. There is also the
insurance/deficiency value and the potential for liability risk. Here is where
predictive analysis comes into play. Predictive analysis is cheaper, faster and
more efficient than traditional analysis methods. Time will tell if predictive
analysis will be a real game changer or not.

92
Predictive Analytics A Case Study In Underwriting-Freedom Specialty
Insurance
In the aftermath of the major economic slowdown of 2008, a risky underwriting
proposal was submitted to the executive management team of Scottsdale
Insurance Company. This case study is taken from the D&O insurance industry
(D&O is Directors and Officers Liability Insurance. This liability insurance is
paid to directors and officers or the organization itself. The money is
reimbursement for losses or a loan of defense fees in case an insured
experiences a loss because of a legal settlement.)
The Scottsdale Insurance Company accepted the proposal and Freedom
Specialty Insurance Company was born. Freedom developed an industry first
method, feeding external data into a predictive model to support risk intuition.
The theory was class action lawsuit data could predict D&O claims. A multi-
million-dollar unique underwriting platform was born. The dividends are
beginning to pay off as Freedom has grown to $300 million in annual direct
written premium. They have consistently kept their losses under control. In
2012, they were under the average industry loss rate of 49%.
Their model delivers a high level of honesty in their dealings. This has caused
great trust among personnel from the underwriter to the executive branch in the
parent company. The reinsured are confident as well. The point of this case
study is to show how Freedom built, maintained and refined their underwriting
model. In hopes that other companies can take this model and incorporate it into
their underwriting operations. Using lessons Freedom learned along the way.
The platform was a multi-team effort in its development. An actuarial firm built
and tested the predictive model. The external technology supplier built the user
interface and constructed the integration with corporate systems. Next,
technology from SAS was used for many parts such as many data repositories,
statistical analytics engines and reporting and visualization tools.
Their platform involves the following components:
Data Sources
Data Scrubbing
Predictive Model
Risk selection analysis
Interface with corporate systems

These are described as follow:

93
Data Sources-The system integrates with six external data sources and receives
data from corporate administrative applications. External sources, like class
action lawsuit and financial information, are commonly used in the D&O
industry. Other sources are specific to Freedom and drive their predictive
model. Data is central to the platform’s function, so Freedom spends a lot of
time maintaining quality data and vendor activities. The back-testing and
categorization processes illuminate vendor weaknesses and incongruities in
vendor data. Freedom goes to great lengths to work with vendors to classify
data and keep its quality high.
Freedom has realized they need to keep an eye on their eternal data as well.
They apply strict checks to improve policy and claims data. Freedom absolutely
maintains data freedom from vendor’s identification schemes. It takes extra
work to translate values but this insures Freedom can get rid of vendors quickly
if they have to.
Data Scrubbing-The data goes through many cleansing cycles when it’s
received. This guarantees maximum usefulness. One example of this is 20,000
individual class action lawsuits are reviewed monthly. Initially, they were
categorized by different parameters, but they are reviewed monthly to see if any
changes have taken place. In the beginning, this process took weeks to complete
when done manually. Now it’s done in hours using advanced data categorizing
tools.
Back-testing-This key process calculates the risk involved when an initial claim
is received. The system will funnel the claim back through the predictive model;
testing the selection criterion and changing tolerances as needed. The positive
feedback loop refines the system through many uses.
Predictive model- Data is condensed and fed through a model, which uses
multivariate analysis to determine the best range of pricing and limits.
Algorithms evaluate the submission across many predetermined thresholds.
Risk Selection Analysis- This produces a one-page analytical report of
suggestions for the underwriter. Comparisons with the same risks are displayed
along many risk factors- industry, size, financial structure and other parameters.
The underlying principle of the platform is that it’s driven by underwriter logic
helped by technology. It will never replace the human underwriter.
Interface with corporate systems- Upon a decision being made selected
information is sent over to corporate administration. The policy issuance
process is still largely done manually. There is room for complete automation in
the future. The policy is issued, and the statistical information is looped back
through the data source component. As claims file through loss, data is added to
the platform.
Like any D&O operation, underwriters possess deep technical insurance

94
knowledge. They have had to adjust and become comfortable with a system that
reduces mass amounts of information into several analytical pages, where the
underwriters traditionally spent time collecting, collating and analyzing data.
They now interface with policyholders and brokers in book and risk
management. Here again, predictive analytics has streamlined the data process
changing the traditional role of the underwriter.
Of course with this revised underwriting process in place, skilled technical
personnel were added. These people have legal and statistical knowledge that
enable them to be able to build predictive models in the financial field.
Incorporating this model has allowed Freedom to develop process expertise in
many areas. The data management tasks-data scrubbing, back-testing, and
categorization were learned from scratch. Initially, these processes were handled
manually, but they have been continuously automated since their inception. In
addition, the number of external sources is always expanding. Freedom is in the
process of evaluating the use of cyber security and intellectual property
lawsuits. The predictive model is constantly undergoing refinement. The tasks
involved in the care, feeding and maintenance of the predictive model used in
other industries has been developed in the D&O industry. Of special interest is
after the actuarial initially built the predictive model, it took many months for
Freedom to gain a full understanding of its intricate operations.
Processes were put in place to effectively manage all these external vendors
together. Several external groups came together to pull the predictive model
project together each group having to work in tandem with the other groups.
Again a list of those groups:
Actuarial Firm
IT Firm
Data vendors
Reinsurers
Internal IT

I mention, again, the groups involved in this project because it took a very
skilled and concerted effort, by Freedom, to bring these entities together to work
their skills in the same place at the same time.

Success of the Platform

This predictive analytic model was very successful for Freedom. It has opened
up new horizons for the Insurance Company. First, the analysis was so specific

95
it now opens up dialog with brokers and policyholders in risk management
discussions. Secondly, this model may be extended beyond liability lines to
property lines. Expanding this model will make it applicable to other lines of
liability such as surety. This predictive analytical model makes Freedom stand
out from its competitors. Finally, the methods used in the back-testing and
categorization elements can be used to forecast other data element. Freedom has
positioned itself to discover other potential information sources.

Predictive Analytics is Impacting Real Estate

How can Predictive Analytics impact Real Estate?
One example is a Real Estate corporation that used data devices to help a law
firm decide if it needed to move into a different office space.. They started out
by addressing employee retention factors. The law firm wanted to recruit and
maintain the best prospects. As this was the major concern for the client, the
Real Estate took this initiative instead of approaching it about a direct real estate
suggestion. The firm used various location-aware devices to track the
movements of the employees. The tracking data was implemented into data
mapping based on employee preferences and movements.
The interpretation of the data resulted in the law firm moving out of the high
rise office space into a cheaper office space. The data saved the law firm money
and positioned them into a new location. The improved location solved the
employee retention problem.

Predictive Analytics is changing the Way of the National Association of

Realtors (NAR)
Predictive analytics is helping NAR re-think the way they conduct business.
They are America’s largest trade association at 1million members. So their
results with predictive analytics could influence the Real Estate industry
dramatically. The NAR is looking how to add value to their client relationships.
They established a new analytics group who will:
Analyze member and customer data trends
The NAR will use the data results to add value to its realtors
Use disparate data models to build analytical models
The models will solve complex problems in the housing industry
Help Real Estate agents, state and local associations and others make
better data driven decisions

96
The data group will step forward in three phases. The first phase is the
experimentation phase where the data group will interpret the accumulative data
and identify trends and patterns. The last two phases will involve forming
partnerships with other data users; the partnerships will design and develop
products to help their agents. The golden result of applying predictive data
models will help the NAR to uncover behavioral patterns and build targeted
meetings with potential home buyers. This will result in increased operational
processes for NAR. Ultimately, the NAR will increase their financial bottom
line.
NAR’s traditional process of analyzing data is to gather holistic big picture
proprietary data. The data extracted show what had happened historically in the
group’s business dealings. The other part of analyzing data is gleaning
information from the customer relationship management system to determine
current events affecting the group.
The new and improved big data analyzing process will interpret focus on the big
picture and the finer details in one source. The new process will analyze both
proprietary and public data systems to discern trends that will predict the future
in real estate business practices.
On the other side of the coin Insurance companies will be changing the way
they conduct business as well. As predictive analytics helps doctors and patients
consult to create better and faster health care solutions. Patients will stay
healthier longer and will heal faster because the right treatment options are
being administered to them to meet their physical needs. Insurance companies
will have reduced premiums and claims to pay on. Yes, the insurance
companies will lose millions of dollars in premium costs because patients won’t
need extensive and unnecessary medical procedures. This will surely drive
down health insurance costs.
But insurance companies can increase revenues in new ways as the medical
profession becomes more focused and singular in its individual patient health
care plans. This will reflect in specialized but shorter hospital admissions.
Insurance companies may become more specialized in what they cover. The
result will be new avenues of revenue.
Medical device companies will be impacted as well. The familiar medical
devices they have been manufacturing to take care of traditional healthcare
issues will diminish. The obvious conclusion lost revenues for these companies.
Here too though, the device companies can adjust to the new playing field. They
can too shift gears and start manufacturing the specialized devices that will
undoubtedly be needed in the new healthcare environment. Thereby, bringing
increased revenues to their financial bottom line as well. Predictive analytics is
bringing widespread change to every facet of the healthcare system. There is yet
to come unseen changes as well as predictive analytics processes become more

97
refined and increase in accuracy.

Predictive Analysis is Changing Industries

It is obvious that Predictive Analysis is changing the game drastically in how
data is collected, interpreted and predicted. This section on “Why Predictive
Analytics” shows that Predictive Analytics is changing the face of so many
industries internationally. Many industries are on the cutting edge of the
advancements Predictive Analytics is making in various applications in specific
phases of those industries. Freedom Insurance who has developed a predictive
model that is spreading to other types of applications in the Insurance industry
has placed themselves as the leaders in their respective industries.
Predictive analysis is reported to be a very lucrative market reaching 5.2billion-
6.5 billion by 2018/2019. Mind you not millions but millions will be spent in
this market.
Take the healthcare industry for example predictive analytics is changing the
way physicians take care of their patients. Traditionally, the doctor would just
give the patient a medical treatment to follow. So the patient could recover and
return to his/her full health. Now physicians have become more like counselors
to advice patients on the best course of treatment to take. The patient is
intricately involved in his own care now because of predictive analytics. The
patient has access to his medical records through his personal physician.
Gone are the days where the patient was unaware of his medical history or
vulnerability to different diseases because of genetic factors and family history.
Now because of Predictive Analytics patients view their records and witness the
multiple predictive models that will steer them into the best care for their future
health. They now have a strong influence in which predictive treatment will best
bring about the desired result-good health. Here the role of doctors and their
patients are gravitating to definite role changes. Who would have thought 5-10
years ago Predictive Analytics would change healthcare so much? (Maybe
someone somewhere created a predictive model that anticipated the different
changes that are affecting the health industry now).
In all the research done for this book concerning Predictive Analytics, every
time the predictive analysis applications have been used, they have improved
the processes of companies. Every single case researched for predictive
analytics has shown that these newer models can help industries improve their
overall operational capabilities. Some industries have been radically changed by
Predictive Analytics:
Law enforcement (Crime Scene Reconstruction)
Insurance (Risk Management concerning employees and customers)

98
 Transportation (Predicting troublesome behavior and averting thousands of
citations, arrests, millions of dollars lost in property damage due to
accidents)
Healthcare. This is a big one because Doctors can now effectively diagnose
illnesses, conditions surgeries based on predictive models. Patient and
Physician’s roles are overlapping at integrating together. It’s more of a
democratic process where patient, and doctor jointly decide the best
treatment options. Many more lives will be saved because of updated and
accurate health care treatments based on healthcare predictive models.
Retail Industry. Predictive analytics are helping retailers to increase
customer retention rates. In some cases, by large percentages.
Real Estate. The NAR for example, is using Predictive Analytics to predict
what preferential property packages consumers will likely buy according to
historical home buying practices. Rental property investors are using
predictive analytics to predict where the most lucrative rental properties may
be found based on renter’s economic, geographical and family needs.

It’s not unrealistic to say that Predictive Analytics has caused industrial
revolutions in some industries. The potential market will exponentially increase
in the next few years as more and more success stories are detected about
innovative ways Predictive Analytics are being used by visionary companies.

99
100
Chapter 9: What is Data Science?

Data science is defined as “the deep knowledge discovery through data

inference and exploration”. Data scientists need to have experience at
algorithmic and mathematical techniques so that they can solve some of the
most complex of problems, using pools of raw information to find the insights
hidden beneath the surface. Data science is centered on analytical rigor, based
on evidence and on building decision-making capabilities.
Data science is important because it allows companies to operate more
intelligently and to strategize better. It is about adding significant value by
learning from data. A data scientist may be involved in a wide variety of
different projects, including:
Tactical optimization – improving business processes marketing
campaigns etc.
Predictive analytics – anticipating future demand, etc.
Nuanced learning – development of a deep understanding of consumer
behavior
Recommendation engines – recommendations for Netflix movies, or
Amazon products, etc.
Automated decision engines – self-driving cars, automated fraud
detection, etc.

While the object of all of these may be clear, the problems that arise require
extensive expertise to solve them. It may be that a number of models need to be
built, such as predictive, attribution, segmentation, etc. and this requires an
extensive knowledge of machine-learning algorithms and a very sharp technical
ability. These are not skills that you can pick up in a couple of days learning,
they can take years. Below we look at the skill set required to become a data
scientist:

101
Data Science Skill Set:
Data science is a multidisciplinary job and there are three main competencies
required:

Mathematics
The very core of determining meaning from data is the ability to be able to see
that data through in a quantative way. Data contains patterns, textures,
correlations and dimensions that are expressed numerically; determining any
meaning becomes a kind of a brainteaser requiring mathematical techniques to
solve it. Finding the solution to any number of business models will often
require that analytic models are built, models that are grounded in the theory of
hard math. It is just as important to understand how the models work as it is to
understand the process of building them.
One of the biggest misconceptions about data science is that it is all about
statistics. While these may be important, they are not the only math that has to
be understood by the scientist. There are two branches to statistics – classical
and Bayesian. Most people who talk about that are talking about the classical
ones but a data scientist needs to understand both types. As well as that, they
have to have a deep understanding of linear algebra and matrix mathematics. In
short, a data scientist has to have a very wide and very deep knowledge of math.

Hacking and Technology

Before we go any further, let me just clarify something – I am NOT talking
about breaking into computers and stealing sensitive data when I talk about
hacking. I am talking about the ingenuity and creativity required in using
learned technical skills to build models and then find the right, clever, solution
to a problem.
The ability to hack is vital because a data scientist needs to be able to leverage
technology to get the vast amounts of data sets and to work with algorithms that,
for the most part, are complex. Just being able to use Excel is not going to cut it
in the world of a data scientist. They need to be able to use tools like R, SAS
and SQL and for that they have to have the ability to code. With these tools, a
data scientist is able to piece together data and information that is not structured
and bring out the insights that would otherwise remain hidden.
Hackers are also algorithmic thinkers – they are able to break down a messy
problem and turn it into something that can be solved. This is a vital skill for a
data scientist especially as they work very closely with algorithmic frameworks

102
that already exist, as well as building their own, in order to solve an otherwise
complex problem.

Business Acumen
One of the most important things to recognize is that a data scientist is a strategy
consultant, before anything else. Data scientists are valuable resources in
companies because they and they alone are in the position to be able to add
significant value to the business. However, this means that they have to know
how to approach a business problem, how to dissect it and this is just as
important as knowing how to approach an algorithmic problem. Ultimately,
value doesn’t come from a number; it comes from the strategic thinking that is
based on that number. In addition, one of the core competences of a data
scientist is the ability to tell a story using data. This means they have to be able
to present a narrative that contains the problem, and the solution, derived from
insights gained from the data analysis.

103
What Is A Data Scientist?
One of the defining traits of a data scientist is their ability to thin deeply coupled
with an intense curiosity. Data science is about being nosy, about asking
questions, finding new things and learning. Ask any true data scientist what the
driving factor is in their job and they will not tell you it’s money. Instead, they
will tell you that it is all about being able to use creativity, to use ingenuity to
solve problems and to be able to indulge curiosity on a constant basis. Finding a
meaning in data is not just about getting the answer, it is about uncovering what
is hidden. Solving problems is not a task; it is a journey, an intellectually
stimulating one that takes them to the solution. Data scientists are passionate
about their work and they take great satisfaction in meeting a challenge head on.

104
How Analytics And Machine Learning Are Linked To Data Science
Analytics is now one of the most-used words in business talk and while it is
used quite loosely in some cases, it is meant as a way of describing critical
thinking of a quantative nature. Technically, analytics is defined as the “science
of analysis”, or, in easier terms, the process of making decisions based on
information gained from data.
The word “analyst” is somewhat ambiguous as it covers a range of roles –
operations analyst, market analyst, financial analyst, etc. Are analysts and data
scientists the same? Not quite but it I fair to say that any analysts are data
scientists at heart and in training. The following are a couple of examples of
how an analyst can grow to be a data scientist:
An analyst who is a master at Excel learns how to use R and SQL to get
into raw warehouse data
An analyst who has enough knowledge of stats to report on the results of
an A/B test goes ahead and learns the expertise needed to build predictive
models with cross validation and latent variably analysis.
The point I am trying to make is that to move from analyst to data scientist
requires a great deal of motivation. You have to want to learn a lot of new skills.
Many organizations have found a great deal of success in cultivating their own
data scientist by providing the necessary resources and training to their analysts.
Machine learning is a term that is always used when we talk about data science.
Put simply, machine learning is the art of training algorithms or systems to gain
insight from a set of data. The type of machine earning is wide ranging, from a
regression model to neural nets but it all centers on one thing – teaching the
computer to recognize patterns and recognize them well. Examples include:
Predictive models that are able to anticipate the behavior of a user
Clustering algorithms that can mine and find natural similarities between
customers
Classification models that can recognize spam and filter it out
Recommendation engines that can learn, at an individual level, about
preferences
Neural nets that learn what a pattern looks like

Data scientists and machine learning are tied closely together. The scientist will
use machine learning to build algorithms that are able to automate some
elements of problem solving, vital for complex projects that are data driven.

105
106
Data Munging
Raw data is often very messy with no real structure and data munging is a term
that we use to describe the process of cleaning the data. This is so that it can be
analyzed and used in machine earning algorithms. Data munging requires very
clever skills in hacking and the ability to recognize patterns so that vast amounts
of raw information can be merged and then transformed. Dirty data hides the
truth that may be hidden beneath the surface and, if it isn’t cleaned, it can be
misleading. As such, a data scientist has to be good at data munging so that they
have accurate data to work with.

107
108
Chapter 10: Further Analysis of a Data Scientist’s Skills

One of the best people from whom to seek a worthwhile opinion regarding the
skills necessary for a data scientist is someone whose job is to hire data
scientists. The reason is that recruiters know precisely what skills they are
looking for in the potential employee or consultant. Burtch Works is one such
firm that deals with recruitment of senior personnel in the field of business and
industry. They have some work that was published in 2014, and it spells out the
skills a data scientist should have. Such information is credible, not just because
it comes from a renowned firm of experts, but also because the firm itself has
managed to walk the talk and climbed up the ladder of success to join the
Fortune 50.
Why then not borrow a leaf from Burtch Works and implement their
recommendations in regards to equipping potential data scientists with relevant
skills?

109
Further Skills Recommend For A Data Scientist
1) Solid educational background
From research, it is evident that people who manage to specialize as data
scientists have a solid educational background. In fact, a respectable 88% of
them have a Masters degree; a good 46% are PhD holders; while others are
generally well educated. Being a data scientist demands that the person is
capable of developing the required depth of knowledge, and one can only do
that with a good level of education.
Yet, even with a good level of education, a potential data scientist needs to have
an inclination towards fields that declare you are not afraid to do calculations;
fields that deal with analysis of numbers and formulas and so on. According to
Burtch Works, 32% of data scientists have a great background in Mathematics
as well as Statistics; 19% in Computer Science; and 16% in Engineering.

2) Competent in SAS plus/or R

SAS stands for Statistical Analysis Software, and it is only natural that a
potential data scientist should be comfortable using software that helps in data
analytics at an advanced level; data management; predictive analytics; and such.
It is even better if the person knows something about the R programming
language helpful in creating important functions. Generally speaking, any good
analytical tool available is good for a data scientist.

3) Skills in Python Coding

According to Burtch Works, employers who seek to employ data scientists want
someone who is has the technical skills in the use of Python, which is a popular
coding language. Often they are thrilled if the person is comfortable in the use
of Java; Perl; or even C/C++.

4) Knowledge of Hadoop
It is important that a data scientist be skilled in the use of the Hadoop platform.
These technical skills may not be mandatory, but it helps to be able to derive
statistical data with ease from such an open source library. Other technical skills
that are preferred include competence in the use of the Apache Hive, which is
some data warehouse software. The Hive helps in querying data using a
language almost similar to SQL, which goes by the name HiveQL.

110
And whoever is skilled in the use of Apache Pig on top of Hadoop stands a
good chance of securing a job as a data scientist. Apache Pig is another platform
that helps in data analysis. Also, as long as a person is interested in making a
career as a data scientist, it is advisable to get familiar with existing cloud tools
like Amazon S3, and any other that may be developed over time.

5) Skills in SQL Database

It is important that a potential data scientist be able to work with Structured
Query Language (SQL), which is a programming language used in data
management and stream processing of data. A data scientist needs to be skilled
at writing complex SQL queries as well as executing them. It is helpful for the
data analysts to know tools that will help with analyzing and compiling data.
Some of these tools are TOAD, DataSpy, Erwin, and many more effective
modeling and architecting data tools and software.

6) Ability to work with unstructured data

This is one area where a solid background in education helps a lot. As a data
scientist, not every aspect of the work involves calculations. And even if it does,
one needs to be able to determine which data is relevant in what equation for the
outcome to make business sense. A high level of competence in critical thinking
is, therefore, required, to be able to maximize the benefits of massive data from
the social media; video feeds; and other sources.

7) Having Intellectual Curiosity

You surely can’t afford to have someone for a data scientist, if they rely solely
on memory, like in remembering formulas and facts, and other stuff in the
books. A data scientist needs to be someone eager to learn more about things
that are happening in the world, which can impact the cost of doing business,
improve efficiency in doing business, or generally impact profitability.
Intellectual curiosity is one of the important soft skills Frank Lo,
DataJobs.com’s founder, wrote about back in 2014, as a guest blogger for
Burtch Works. It is a non-technical skill, but it makes a data scientist stand a cut
above others in the field of analytics

8) Competence in understanding a business situation

You can bank on someone with great business acumen to help your organization
better as a data scientist, than one who is least interested in matters of business.
Business acumen is another of those non-technical skills that are, nevertheless,

111
very helpful when a data scientist is analyzing data with a view to solving
business problems. A data scientist with business acumen is able to analyze data
and deduce what the lead problem for your business is, and what factors are
escalating it. This is very important because it then helps you in prioritizing
business decisions in a manner that allows you to protect the business, even as
you capture fresh business opportunities. On the overall, your organization can
leverage whatever data it has better than otherwise.

9) Skills in Effective Communication

Ever listened to learned people who addressed a group as if to impress them?
That shouldn’t happen with a data scientist. The marketing or sales manager, for
example, would care less if you used the IFFEROR or VLOOKUP formula; the
INDEX+MATCH formulas; or even went for the popular IF formula in your
data analysis. All they want is information that tells them the current market
situation as deduced from available data; and possible scenarios from that same
mass of data. Data analysis is meant to provide quantified insights into the
business situation, and unless those insights are communicated clearly and
effectively to the relevant decision makers by the data scientist, all resources
spent on data analysis will have been wasted.

112
Further Demystification Of Data Science
Have you heard someone describe data science as the sexiest job of this 21st
century? Well, those were the sentiments of expert contributors of the magazine,
Harvard Business Review, Thomas Davenport and his counterpart, D.J. Patil.
Although there are some skills that would make you better placed as a data
scientist, different employers often have different needs in mind when they
advertise for a data scientist. So, it is important that you don’t give
advertisements a casual look but a deep one; reading the details mentioned
therein. In short, when it comes to daily practice, the term data scientist is a
blanket term often covering four categories of engagements as follows:

1. Data Analyst
What we are saying here is that although you do not have the qualifications
mentioned earlier on of a data scientist, you would be doing yourself disservice
if you are great at Excel and have some knowledge of MySQL database and
dismiss a job advertisement just because its banner is data scientist.
At times people are hired to fill in the vacancy of a data scientist yet they end up
spending their days spooling reports and data from MySQL or doing work on
Excel pivot tables. Other times they are called upon to work with a team on
some Google Analytics account. Whereas you may not be doing the deep work
expected of a data scientist, the tasks just mentioned are not irrelevant or
useless. Rather, they accord you a platform to practice the basics of a data
scientist’s job if you are new in the field. In fact, if you were ever hired for such
a post, you can take the opportunity to explore additional use of data analytics
beyond what the company demands, hence expanding your skill set.

2. Data Engineer
When they advertise for a data scientist and somewhere in there you find the
mention of duties you are comfortable with as a data engineer, don’t let the job
slip by. If the job description includes things to do with statistics; machine
learning expert; and such other roles that you can do as software engineer, it is
fine to go for it. Alternatively, you may find the job description being one of
building data infrastructure for the organization. Companies often make such
advertisements when they have too much data around and they do not want to
discard it because they feel they are likely to need it sometime in the future.
Other times they see the data as being important, but since it is unstructured
they don’t know how to put it to good use.

113
 3. Statistician
Sometimes companies dealing with data based services need someone who can
handle consumer focused data analysis or even to deal with intense machine
learning activity. Most of such companies actually run a data analysis platform.
The tasks involved call for someone who is great at mathematics or statistics; or
even someone with a background in physics; presumably people who are
looking forward to advancing their academics along those lines. So, if you see a
data scientist’s job for a company of this kind that produces data driven
products, do not hesitate to apply if you are specializing in mathematics,
statistics or something relating to calculations and analysis of figures – you
could be just the person with the skills they require at that time.

4. General Analyst
You may find a big company seeking to employ a data scientist, but when you
look at the details you find that the tasks the person is to handle deal more with
data visualization or even machine learning. It is alright to apply for the job
because it means mostly the successful candidate is going to join an experienced
team of data scientists and just want someone to help out in the lighter chores.
What a great learning experience it would be for you especially if you haven’t
had much experience in a position of a data scientist! In short, if you are skilled
in some big data tools such as Hive or Pig, you should be comfortable applying
for this job irrespective of the heading, ‘Data Scientist’.

5. Data Architects & Modelers

With the increasing amount of data that companies are required to keep and
maintain, it is necessary to have people that can transform data from all the
various systems into databases that can make more structured sense of it. This
newly structured data is used to identify risk, fraud and alert triggers the data
might need to have. The Modelers and Architects role is to work with
development project teams to make sure any system changes get sent down to
the data repository and will be in a format that can be used for reports and
functions needed to use that data.

The Future Role of the Data Scientist

Experts are debating the future role of data scientists. Will they become extinct
by 2025? In the different regions of the data world, there is an emerging opinion
that data scientists could become unnecessary personnel anywhere from 5-10

114
years. Others say this could become the reality in 50 years.
There are those who say the need for a data scientist will always exist. The core
group who say data scientists will never become extinct base this on what is
called “expert level tasks.” The basic assumption is that there are some data
science tasks that are too complicated for robots or automation to be able to
perform. The need for human creativity and innovation in data expert level tasks
will always be needed. Robots won’t be able to think outside the box when data
interpretation calls for new data model methods to be applied or built to solve
unknown or hidden business issues arise.
The other side says that all data expert level tasks no matter what their level of
complexity will be automated within 5-10 years. Software tools will be able to
complete complex tasks that data scientists now perform. One example that was
given is that with software tools like Tableau the very difficult task of
visualization has been mastered with an application. Second generation data
science companies are making software tools that will improve company
workflow and automate data interpretation. Currently, this issue is far from
solved, so only time will tell if the Data Scientist’s job is in jeopardy or not.

115
116
Chapter 11: Big Data Impact Envisaged by 2020

Have you realized that it is not just advertising and marketing that organizations
are taking online these days? If you consciously think about what takes you
online on a daily basis, you will realize that a good part of it is business,
sometimes because you feel you can find a wider choice of items there, plenty
of information about the product or service that you want, and even a wide
range of pricing. The web also seems to provide much more entertainment than
you could physically reach in a short time, and most probably at a relatively
smaller price. And when it comes to connecting with others, individuals, private
and public institutions, and all manner of organizations are taking their
communication online, where they can reach a wider audience much faster and
more cheaply.

117
How Does Moving Online Impact Data?
Well, it means that the amount of data being generated on a daily basis is
growing exponentially, and we can no longer ignore big data. Secondly, even
before anyone can speak of the ability to analyze and organize the massive data,
tracking it is itself a mammoth challenge. Would you believe that internet users
are generating data in excess of 2½ quintillion bytes each day? This includes the
automated feedback like traffic monitors, weather related trackers, and all
manner of transactions. Is this a good thing?
Well, potentially, it is. But there is the question of what sources are reliable and
which ones are not; what data is relevant to your scenario and which one is not;
and so on. As you may already know, having massive data before you can also
be overwhelming. That is why big data has created room for unique business,
where you get people specially training to make good use of big data. That
potential power of big data is what specialized companies seek to help you
unravel and take advantage of, by handling for you big data that affects your
organization.
In fact, specialized technology has come up to make big data management
convenient. A good example of this is Apache™ Hadoop®, an advanced
database management technology that takes you beyond consolidation of
information to improved efficiency of your business and increased profits. As
long as organizations are open to the use of advanced technology that makes big
data analysis and management convenient, the world is headed for better times.
Instead of being overwhelmed by high data traffic and accumulated mass of
data, organizations are going to unleash the power of that data and if they are in
education they are going to bring down drastically the cost of education. If they
are in meteorology, they are going to have better comprehension of certain
complex phenomena like the weather. Those in business will be able to raise
their productivity by drastically cutting on all manner of redundancies; and the
job market is going to have better correlated data sets that will help to match job
seekers against potential employers as per the skills offered and needed
respectively. Some of these things are already happening and they can only
become better.
And the story of big data does not end with reduced cost of doing business and
acquiring education; increased efficiency and productivity; as well as increased
profits. Research that is already going on points to the potential of improving
the fight against crime; significant improvement of web security; and ability to
foretell early enough when there is likelihood of an economic or natural disaster
sometime in the future. In fact, as long as there is work going on regarding big

118
data and its intricacies, the world is likely to witness bigger and more radical
changes, much of it for the better.

119
What’s The Market Like For Big Data?
As already explained, there are many innovators trying to create analytical and
data management tools that will turn the massive data generated into an
economic advantage. In 2012, the big data market was worth $5 billion. If you
remember what we said earlier that big data has been growing exponentially,
you’ll not be surprised to know that going by the situation today, market for big
data is projected to be $50 billion by the time we get to 2017.

Specific Ways Big Data Can Significantly Impact Life – Soon

1. Websites and applications functioning better
In this regard, not only is it expected that it will be much easier and convenient
to navigate websites and make sense of data availed there, but also that the sites
will be much safer than they are today.
This is particularly so because big data can be handy in identifying and tracking
fraudulent activity on a real time basis. Big data can introduce fresh and clearer
visibility on an organization’s website, and also make it possible to foretell
when attacks are imminent. Innovators have actually already begun designing
programs geared towards safeguarding data against destructive intrusion. For
example, there is the machine learning program known as MLSec that you can
find at MLSec.org. and it uses algorithms under supervision, to locate networks
harboring malicious programs. It must be very encouraging to web users to
learn that this machine learning program has been proven to be accurate to the
rate of 92% - 95% for every case tested.

120
How Bad Is The Security Situation Today?
Well, as per 2012 statistics:
Of all the websites hacked in the year, 63% of the owners did not realize
they had been invaded.
In fact, 90% of the web owners did not even seem to notice anything
strange at all going on within the site.
50% of the web owners learnt that their websites had been hacked from
their browser warnings or even warnings from the search engine they
were using.

2. Higher education becoming more accessible

Why is it so exciting that cost of education would fall? Well, accessing higher
education is generally a problem for the average person in most countries. In the
US, where it looks from the outside like the land of plenty, the cost of tuition
rises at a double rate compared to that of healthcare. And when you compare the
hike in cost of education to that of the country’s Consumer Price Index, it goes
four times as high.
Luckily, just as websites are becoming better designed and protected, something
is happening to make education more easily accessible to more and more
people. A good example is the emergence of online sites offering great courses.
The Khan Academy found at khanacademy.org; Big Data University found at
BigDataUniversity.com; Venture Labs found at venture-labs.org; Coursera
found at courser.org; are just some examples of institutions that are offering
higher education online at a much lower cost than conventional tertiary
institutions; and sometimes free of charge.
The good thing about many of the courses offered in this manner is that students
are tested on how well they have clinched the skills taught, particularly because
they the skills taught are applicable in the current high-tech environment. For
example, Big Data University teaches Hadoop plus some other big data related
technologies.

3. Relative ease in getting a job

Have you ever given thought to the number of job seekers there are worldwide?
It must be a staggering figure. In fact, on a monthly basis, the number of job
searches on the web alone has hit 1.5 billion. On the positive side, there are
websites that have already begun to match job seekers with potential employers
by amassing valuable information on various employers as well as information
on interested job seekers. One such website is indeed.com.

121
 4. Improved road safety
Are you aware that car accidents are the main cause of death in the category of
youth aged between 16yrs and 19yrs in America? And while it is dangerous to
drive while under the influence of either drugs or alcohol, ¾ of the deaths in this
category are not alcohol or drug related. What this means is that there are very
many accidents that occur because of merely poor judgment. It is envisaged that
as scientists continue to work on advancing the technology being used on
computers, big data is going to help them predict the behavior of drivers on the
road, as well as the move a vehicle is about to make at a certain point.
This route is geared towards reaching a point where cars on the road can
exchange data, so that drivers in different cars can see up to three cars that are
ahead of them; three that are immediately following them; and three on either
side of them at any one time. In fact, it is said that big data will enable the
drivers in data swapping vehicles to see the posture and focus of a driver in a
car near theirs. This may sound a little far-fetched, but think about it: Haven’t
Google’s self-drive cars taken the auto industry to a totally new level?

5. Ability to predict business future

What this means is that organizations will be able to use software like Hadoop
to analyze the data at their disposal in a way that will bring to the fore future
possibilities. The speed and accuracy with which data will processed will
enable organizations to take prompt action where there are business
opportunities emerging; damage control needed; and such other actions that
require accurate assessment. There are already big organizations using Hadoop
with great success, including eBay; Twitter; FaceBook; Disney and others. The
demand for Hadoop is rising rapidly. IDC, a renowned market research firm has
predicted that by 2016, the conservative worth of this software will be $813
million.
Another good example is Recorded Future, a technology company based on the
web. This one provides security intelligence mostly to data analysts, which they
use to keep their information safe. It puts businesses in a situation where they
can anticipate risks and also capitalize on business opportunities by unlocking
predictive signals using clever algorithms. There are other examples already
helping businesses prepare for eventualities, but suffice it to say, as
technological development continues, it will become all the more possible to
leverage data, hence avoiding surprises.

6. Ability to predict matters of weather

122
This ability to predict the weather brings the advantage of being in a position to
protect the environment better. Such protection in turn brings in huge savings to
the country and the world at large considering the massive expenses that are
brought about by weather related disasters. Just for example, weather and
climate related disasters cost the US losses in staggering figures; actually in
excess of $1 billion. Here, we are talking of disasters such as drought; wild
fires; incidences of flooding and storm; and such other unfortunate events.
Things are already looking bright on this front of data technology. There is the
Joint Polar Satellite System (JPSS), for instance, that is set to be launched in
2018. The JPSS, which will have the capability to utilize sensor technology, will
also use data to a hurricane’s path or a storm’s path, well before these disastrous
events occur. This will then give everyone concerned time to plan what to do to
safeguard life and property. As CNBC News has already noted, situations that
relied on guesswork some years back are slowly but surely becoming something
to predict with precision using predictive science.

7. Healthcare becoming all the more efficient

Big data is expected to bring improvement to the health sector, not just by
raising efficiency levels in service delivery, but also by customizing services to
suit respective consumers. McKinsey & Company, an advisor to the
management of many renowned businesses, says that between 50% and 70% of
business innovations depend to a good extent on the capacity to capture
customer’s data and not as much on external analytics. McKinsey & Company
relies heavily on qualitative as well as quantitative analysis, to be able to give
helpful advice to management.
It is said that 80% of data in the medical sector is unstructured. However, with
the trend the health sector in the US has taken in using big data creatively, great
improvement in service delivery is anticipated. It is actually envisaged that big
data is going to help the sector increase value through efficiency, adding value
in excess of $300 billion each year. By the same token, expenditure is
anticipated to reduce by a good 8%.
In treating patients effectively, caregivers benefit a lot from available patient
data. The reason is that the data helps the caregivers provide evidence based
advice. Currently, a medical center within Boston by the name of Beth Israel
Deaconess, is putting a Smartphone App in the market, meant to help medical
care givers access 200 million data points. These data points are expected to
avail data concerning around 2 million patients. There is also Rise Health that
utilizes the accessible mass of patient data, analyzing it from all dimensions,
and aligning it to the health providers’ goals with the aim of improving
healthcare through fresh insights. On the overall, big data brings speed to

123
innovation. A fitting example is the project on Human Genome that took a
whole 13yrs to complete, yet today it would only take a couple of hours to
accomplish it.

124
125
Chapter 12: Benefits of Data Science in the Field of
Finance

As far as data is concerned, there is no longer a shortage. There may even be

excess of it all around, if you consider the traffic going through social media;
real time market feeds; transaction details; and elsewhere. The volume of data
available for use in the finance sector is almost explosive. Its variety has also
expanded, and even the velocity at which the data becomes accessible has
sharply risen. That scenario can either take organizations to heights unknown
before, or leave them dumbfounded from the feeling of overwhelm caused by
the data influx.
Since organizations are in business to succeed, they have learnt that the best
way to utilize this flood of data is to engage data scientists. A data scientist is
that guru who takes the data, explores it from all possible angles, and makes
inferences that ultimately help him or her to make very informed discoveries.

A data scientist utilizing data:

Identifies and captures fresh data sources, analyses them, and then builds
predictive models. The data scientist also runs live simulations of various
market events. All this makes it possible to visualize the reality of
possible situations even before any measures have been implemented.
Data science, therefore, helps the organization to foresee trouble well in
advance and prepare accordingly, and foretell future opportunities as
different factors play out in the business arena and the environment in
general.
Utilizes software like Hadoop; NoSQL; and even Storm, to optimize data
sets of a non-traditional nature like geo-location and things like sentiment
data. After that the data scientist integrates the data sets with that which is
more traditional, like trade data.
Takes the precautionary move of ensuring there is ample raw data in
storage for future reference and analysis. In that regard, the data scientist
finds the relevant data in its raw form and selects the safest and most cost
effective way of storing it.
The expertise of data scientists in utilizing big data is being made even more
convenient by the emergence of other technology based storage facilities. There
is, for example, the cloud based data storage; analytical tools that are not only
sophisticated in the things they can accomplish, but also cost effective. Some
are tools that you can access online free of charge; presented as open source

126
tools. In short, there is a whole range of financial tools that are at the disposal of
data scientists, and they are being put to use to transform the way of doing
business.

A data scientist understands people’s sentiments

Is it strange that a data scientist should be able to analyze people’s sentiments
simply from data? Well, there are fitting tools for that. In doing sentiment
analysis, or what you can aptly call opinion mining, a data scientist makes use
of natural language processing; does text analysis; and makes use of
computational linguistics to consolidate the relevant material required for the
process. There are already firms using sentiment analysis to improve business.
Some good examples include MarketPsy Capital; MarketPsych Data; and Think
Big Analytics.

How firms work through sentiment analysis:

By building algorithms surrounding sentiment data from the market, for
example, by use of twitter feeds. These ones give ample data when
incidences with big impact occur, like a terrorist attack or even a big
storm.
By tracking trends, monitoring as new products get introduced into the
market; responding to issues affecting your brand; and in general,
improving the brand perception.
Analyzing voice recordings right from call centers, when those
recordings are unstructured; then recommending ways of reducing
customer churn, or in other words, recommending ways of raising
customer retention.
Considering that most of today’s business is on customer focus, the need for the
service of analyzing data in order to have a vivid picture what customers feel
about a brand cannot be overstated. In fact, data companies have emerged to
fulfill this need, and their role is that of intermediaries where they gather data,
identify sentiment indicators, and then sell that crucial information to retail
businesses.

Data Scientists in Risk Credit Management

The manner in which data scientists utilize the amount, the frequency and
variety of data available online, has enabled firms to offer online credit with
minimal risk. In some places, potential investors simply fail to access credit
because there is no way of giving them a credit rating. Yet a lender or financier

127
needs to know the extent of risk involved, any time lending is about to take
place. Luckily, with big data and expertise from data scientists, internet finance
companies have emerged, and they have found ways of approving loans and
managing risk. Alibaba Aliloan is a good example of online lending that is
enabled by use of big data.
Aliloan is not a conventional bank but an automated online system, which offers
flexible small size loans to online entrepreneurs. The recipients are often
creative and innovative persons who find it difficult to get credit from
traditional lenders like banks simply because they have no collateral.

Big Data Reducing Risk of Online Lending

Using Aliloan as an example of online lending:
i Alibaba monitors its e-commerce platforms as well as the ones it
uses for payments, with a view to understanding customer behavior and
financial strength. After analyzing the customer’s transaction records and
customer ratings, and also analyzing related shipping records as well as
other related information, Alibaba is able to determine the loan ceiling to
put for the customer; considering the level of risk learnt after the
comprehensive data analysis.
ii Alibaba also gets the online findings confirmed by 3rd party
verifiers, even as it seeks other external data sets to cross check the online
findings against. Such helpful external data sets include customs and
other tax records, electricity records and such other utility bills.
iii After granting the loan, Alibaba keeps tabs on the customer’s
activities, monitoring how the customer is utilizing the funds provided for
investment. The lender generally monitors the customer’s business
strategic development.

Other companies that are offering loan facilities by relying heavily on data
scientists’ expertise on big data are Kreditech and also Lenddo, both of which
offer small loans on an automated basis. These ones have come up with credit
scoring techniques that are very innovative yet very helpful in determining a
customer’s creditworthiness. There are also cases where much of the data used
to assess a customer’s position is from online social networks.

Real Time Analytics Improving the Finance Industry

Any decision maker in the finance industry will tell you it’s not enough to have
data somewhere within reach – it matters when you analyze it. As many people
dealing in critical thinking will confirm, it is not possible to make an informed
decision before you analyze the data before you. So, the longer you delay the

128
process of data analysis, the more you risk business opportunities passing you
by; and the higher the chance of other things going wrong in your business.
However, with the skills that data scientists have relating to big data, time lags
are no longer a handicap in the finance sector.

How real time analytics help:

1) Fighting fraud
It is possible today to detect attempts at fraud through data analytics relating to
people’s accounts. Institutions like banks, credit card companies and others,
have gotten into the trend of keeping on top of things as far as fundamental
account details are concerned, courtesy of big data. They want to ensure they
know if your employment details are up to date and your physical location too;
analyze and see the trend of your account balances; learn about your spending
patterns; analyze your credit history; and have such other important information
at their fingertips. Since data is analyzed on a real time basis, data scientists
ensure that there is a red flag triggered whenever there is suspicious activity
taking place; or even when an attempt is detected. What happens then is that the
account gets suspended so that the suspected fraudster cannot continue
operating, and also the owner receives an alert to that effect instantly.

2) Improving credit ratings

Can anyone give a credit rating without ample data available? Needless to say,
the rating is only credible if it factors in current data and not just historical data.
That’s why big data is so important in this era when credit ratings play a crucial
role in determining the level of risk you carry, and the amount of credit you can
be allowed to enjoy by lending institutions. The fact that data analytics takes
place on a real time basis means that customers’ credit ratings are up to date and
they provide a reasonable picture of the customer’s financial capacity. In any
case, most of the categories of data necessary are already covered online,
including assets in the customer’s name; various business operations the
customer is engaged in; as well as relevant transaction history.

3) Providing reasonably accurate pricing

This pricing factor cuts across products and services. In financing, it may means
a customer can get a better rate of interest levied on money borrowed is the
current rating is better than before. For insurance, a policy holder can enjoy
benefits derived from data analysis, issuing timely warnings over accidents
ahead; traffic jams that may affect the driver; weather conditions; and such
other information that can help in reducing the rate of accidents. With a policy
holder having a clean driving record – or at least an improved one – it is
possible to win a discount on the price of insurance policy. It also means that

129
insurance companies will have less to pay out as compensation.
On the overall, cost of business goes down for everyone whenever data
analytics are used, mostly today that is mostly because of the benefits accruing
from real time analytics. In fact, major financial institutions like banks have
made PriceStats the order of the day. This online firm that began as a collector
of daily inflation rates for a few countries in South America now monitors
prices for around 22 countries (as at 2015), with a view to providing daily
inflation rates for those economies. This means you can easily follow the
fluctuating or steady trend of inflation rates and tailor your business actions
accordingly. PriceStats has a lot of useful data based information for the US too,
the largest world economy.

Big Data Also Great for the Customer

For those who may not be aware, many institutions, including banks and a good
number of other financial institutions do pay to acquire data from a myriad of
retailers as well as service providers. This underlines the importance of data,
particularly when you have the capacity to analyze it and use it where it matters
most. In fact, all data is important depending on the reasons someone wants it
for. You don’t want to store data just for the sake of it. Data held within your
system with nobody accessing it is an unnecessary distraction. Even when data
comes your way without you requisitioning for it, unless you have a good
processing plan in place, you can easily get overwhelmed.
That is why it is important to have a data strategy that is inter-departmental, so
that you can identify the category of data to shove to another department instead
of discarding it, and which portions of data to get rid of straightaway. And the
reason this chapter puts emphasis on the contribution of an analyst in data
handling and utilization is that not everyone can make good use of data from
scratch. But a data scientist has the necessary skills to handle big data from A to
Z.
That is why big data is helpful particularly when there is customer
segmentation.
In the case of institutions that pay to receive data, their aim is to use the data to
create a 360° visual image of their customer. As such, when they speak of KYC
(Know Your Customer), they are speaking from a point of credible information;
and that reduces the risk of doing business with those individuals. Can you see
predictive analytics coming into play right there? This aspect of using big data
to have an overall understanding of the customer has been emphasized by Sushil
Pramanick, a leading figure with IBM analytics. Pramanick also happens to be
the founder of The Big Data Institute (TBDI).

Improving business through customer segmentation

130
Once you can put together customers who have the same needs and probably the
same financial capacity; customers who have similar consumer tastes and are in
the same income bracket; customers who are in the same age bracket and are
from a similar cultural background; and other who match in various ways; it
becomes relatively easy to meet their needs.
a) You can conveniently design customized products and services with
them as a target group
b) You can adjust the manner of relating with them with a view to
retaining them as customers; avoiding customer churn
c) You can tailor your advertising and marketing to appeal to target
groups
d) You can re-engineer products or develop new ones with specific
groups in mind

131
132
Chapter 13: How Data Science Benefits Retail

As already mentioned elsewhere in the book, data in circulation is increasing in

volume each year at a rate that can only be described as exponential; and its
variety and velocity is also shooting. Retailers who are smart in business know
there is value in data. Yet they may not be certain how to trap and analyze data
at every possible interaction for use; where use in this case implies more
business profitability.
This is one reason the demand for data scientists has risen over recent years.
Experts like McKinsey have weighed in on the benefits of employing data
analytics in the retail business. In a report they released in 2011, McKinsey
projected that the retailers going for big data analytics are likely to have their
operating margins rise by a good 60%. That was a way of reassuring retailers
that data scientists have the capacity to turn big data upside down and inside out
– whether it is structured or not; internal or external – organizing it in a way that
makes business sense, and ultimately helping the retailer create gold out of the
mound of apparent clutter.

Retailer’s way of reducing cost and increasing revenue

Receiving recommendations that are customized to the needs of the
individual retailer
Taking advantage of sentiment analysis on data from social media or call
centers; or even the one displayed through product reviews. All this data
is important in giving a clear picture through customer feedback and
gives depth in market insights.
Improving customer experience through predictive analytics; that
improvement being both online and offline.
Improving how retailers lay out their promotional displays and such other
merchandizing resources, by utilizing heat sensors and even image
analysis to get a better understanding of customer behavior patterns.
Utilizing video data analysis to identify customers’ shopping trends. This
analysis is also helpful in identifying cross selling opportunities.
Making consistent daily profits because of the ability to respond to the
information derived from internal as well as external data. Would there
be any benefit, for instance, in taking advantage of reduced cost of raw
materials if the weather did not allow for deliveries of the final product?
So, data analysis puts the retailers in good stead because they can plan for
business with prior knowledge of the weather; economic forecast; traffic
reports; and even knowledge about whether it is low or High Holiday
season, among other details.

133
Growing revenues at a faster rate simply because it is possible to do a
detailed market analysis
Making use of product sensors to communicate on a real time basis,
important information concerning post purchase use.
As far as marketing goes, retailers are able to cut cost by offering
personalized offers via mobile devices. They are also able to deliver
location based offers, which in itself is much cheaper than generalized
marketing.
Retailers can communicate real time pricing; meaning it is possible to
issue prices based on metrics of second by second. In this manner,
retailers can also access data on their competitors; data regarding
consumer behavior; and other factors close to business.
Ability to identify appropriate marketing channels going by the indicators
from analytics on segment consumers. This helps the retailers to optimize
their Return on Investment (ROI).
Making use of web analytics as well as online behavioral analysis to
tailor marketing offers accordingly.
Retailers also benefit in the area of supply chain logistics. In this regard,
data science helps when data scientists use it to track inventory and do
general inventory management on a real time basis.
Retailers are also in a position to optimize on their routes of product
delivery through the telematics of GPS enabled big data. Of course with
well analyzed data, retailers are bound to select the cheapest, safest and
fastest route.
The area of supply chain logistics benefits also from both structured as
well as unstructured data, where retailers are able to foretell likely
customer demand well in advance.
Retailers also get the opportunity to negotiate with suppliers once they
read the scenario from analysis of available data is done.

134
135
136
Chapter 14: Data Science Improving Travel

The world of travel has always had an appetite for lots of data. And it has
always encountered lots of data too even when stakeholders are not deliberately
searching. The biggest challenge has mostly been how to store the data as well
as how to analyze it and optimize its use. The data that can be of benefit to the
travel sector is massive considering that all facets of life are involved, including
people’s cultures; air fares; security in different geographical areas; hotel
classes and pricing; the weather; and a lot more. For that reason, this sector
cannot afford to ignore data science, being the discipline that will help in the
handling of the large data sets involved in travel.

137
How The Travel Sector Uses Data Science
There are lots of benefits that come with the use of big data, which is possible
when you have a data scientist working with you. Some of the benefits that
come with the use of big data in the travel sector include:
Ability to track delivery of goods. This is possible whether it is a freight
shipment involved; a traveler on road; or a voyage. Major online firms
like Amazon and Etsy benefit a lot from such tracking, especially because
it is not just the seller who is able to track the shipment but the customer
too. This is likely to give customers more confidence in the company and
their mode of delivery, thus making them repeat customers – and that is,
obviously, good for business.
Analysis done at each data points. This is one way of increasing business
because then it means that different stakeholders will be able to share
information from well analyzed data. This ensures that nobody receives
data that is irrelevant or redundant.
Improving access to travel booking records; the use of mobile phones to
make inquiries, bookings as well as payments and other simplification of
transactions
Easy access to customer profiles; itineraries; positive as well as negative
feedback; and a lot of other data from internal sources like sensor data.
Easy access of external data such as reviews from social media; the
weather; traffic reports; and so on.
In short, the travel sector is a great beneficiary of big data, and it merges the
data from both its internal and external sources to find solutions to existing
problems. It also gets the same data analyzed in a way that gives helps the
stakeholders anticipate with relative precision the position of future events.
The sector is also able to cut drastically on cost of operation because they
can make better choices when they have massive data analyzed in a timely
manner as happens when data science is employed.
One reason there is a lot of interest in big data in the travel sector is that the
sector is lucrative and it is projected to become even more rewarding as far
as revenues are concerned. Global travel is projected to grow sharply so that
by 2022, its value will have reached 10% of the world’s Gross Domestic
Product (GDP). Of course, the main players involved realize the value of big
data and want to optimize its use so that they get the best business
intelligence to work with. And that makes them see big bucks ahead.

138
Big Data Enabling Personalization Of Travel Offers
Some years back, travel considerations were given as per general
classification of customers. Those in the high income bracket were sent
recommendations for certain facilities; those who regularly travel with
children were targeted for different offers; and so on. While that may have
increased revenues a little bit from the normal, its impact is still peanuts
compared to today’s use of big data.
The reason big data is very significant in the travel industry today, is that it
enables the Travel Company or agency to tailor offers as per individual
customers, by relying on a 360° view of the customer. So the potential of
offering ultra personalized packages or facilities to individuals is not far-
fetched where big data is at play.

Some data sets that help produce a 360° view

Data directed at reading behavior.
A good example of this is the regularity with which a person goes online; and
probably the websites the person is a regular visitor

The posts that one posts on social media

The analyst can establish if the person writes issues about travel on social
media; whether friends speak about travel with this person; or even if there are
any travel reviews contributed by friends.
Data from location tracking
Data on itineraries on various occasions
Historical data about a person’s shopping pattern
Data reflecting how the person uses mobile devices
Data regarding image processing

This list is by no means exhaustive. In actual fact, any data or information

relating to an individual’s preferences on travel would make the list. As you will
realize different data sets suit different people while others are common to all.
For example, the data sets targeting a person’s behavior pattern will mostly be
useful for a potential travel customer; or one who is new to the sector. However,
when you speak of historical patterns, then you are targeting old customers.

139
Big Data Enhancing Safety
If you consider the information transmitted between coordinators of travel from
the pilot to the control tower and vice-versa; driver and the fleet headquarters
and vice-versa; from traffic headquarters to all travelers; from electronic media
to travelers at large; and so on, you will easily acknowledge that big data is a
real life saver in the sector of travel. In fact, in present travel, the vehicles and
planes concerned are fitted with different types of sensors that detect, capture
and relay certain information on real time basis. The information relayed varies
greatly and may include airmanship; the behavior of the driver; mechanical state
of the vehicle or plane; the weather; and so on.
This is where a data scientist comes in to design complex algorithms that enable
the travel institution to foretell when a problem is imminent. Better still, the
work of the data scientist helps to prevent the problem before it crops up.
Here are some of the common problems that big data helps address:
If there is a part of the vehicle or plane that is not working well and is
detected, it is replaced before the damage on that part becomes too
serious for the part to be salvaged. Replacement or repair is also done to
prevent an accident.
In case the drive is missing one step or more when traveling, you need to
pull him or her out of active duty and take him or her through some
further training. This is necessary in order to maintain high standards of
performance.
For flights, big data helps to identify problems that can be tackled when
the plane is mid air and those that are possible to tackle in such a
position. For those that are found to be difficult to tackle when the plane
is in the air, the company puts a maintenance team on standby so that
they can dash to check the plane immediately on arrival.

From the facts provided, it is safe to say that big data plays a crucial role in
averting accidents during travel, whether by air; by sea; or even by road. In
fact, it is easy to appreciate that when you think about the transmission of
weather related data and information on real time basis.

140
Up-Selling And Cross-Selling
Sometimes big data is used in up-selling and other times in cross selling. Just to
re-cap, vendors are up-selling when they are trying to woo you into buying
something that is pricier than what you were seeking in the first place – great
marketing attempt. As for cross selling, vendors just try to get you to buy
something different from what you were searching for initially.
In case you want to travel by air, you are likely to receive many other offers
together with the one you inquired about in the first instance.
Here are examples of up-selling and cross selling:
You may find in your inbox an offer that is pretty personalized, and that
is a form of cross-selling
You may find yourself being booked for the Economy Plus, which means
you are being up-sold. With the advantages of additional leg room and an
opportunity to recline further on your seat, it is tempting to take up the
offer even if it means forking an extra amount of dollars.
You are likely to find discounted offers of hotels partnering with the
airline you are using for your travel.
You could also receive an offer for dining, courtesy of the steward, and
this is on a complimentary basis – so you get a coupon
You may find yourself looking at an ad, during an in-flight entertainment
session, trying to attract you into using a certain city tour on arrival or
during your local travel.
Companies particularly in the tourism sector are partnering and using big
data to cross sell each other. The usual suspects include airlines; hotels; tour
van companies; and companies offering taxi services.

141
142
Chapter 15: Big Data and Law Enforcement

Big data is making critical inroads in criminal investigations. Today some law
enforcement agencies aren’t using data tools to their fullest potential. They will
work independently of one another and maintain individual standalone systems.
This is impractical and not a cost effective way of running these agencies. Law
enforcement is expected to do more with less. Manpower has been cut back but
crime increases in some areas.
It is to the benefit of these agencies to pool their shrinking resources into a data
networking system where law enforcement agencies network data across the
board to help solve crimes and protect the public more efficiently.
Big Data is the answer to finding effective solutions to stopping crime in a time
when agencies are being asked to perform more tasks with less help. This is
where big data is the go to for these agencies. They are being forced to
streamline operations and still produce the desired results. Data has to be
interpreted holistically to be able to streamline their operations Data platforms
make these objectives a reality.
The data platform can be placed in the Human Resources system or even in the
plate identification system. It doesn’t matter where the platform is integrated it
just has to be inserted into one of the data systems. So there will be a foundation
from which criminal activity can be dissected, analyzed and translated into
effective crime fighting techniques. Challenges facing Law Enforcement:
Law Enforcement is being straddled continually with shrinking financial
resources
Limited money equates to less manpower
Even with less manpower Law Enforcement is expected to maintain the
same level of services
Law enforcement must find innovative ways to fight crime through data
systems
Big Data can meet all of the above requirements if use to its fullest
potential

Law enforcement is the foundation upon which our society is kept safe, where
people can feel secure and safe walking down the street in broad daylight.
Where a young woman can walk safely in remote areas at night and not feel
threatened.

143
Lately, the public view of law enforcement has turned negative. Police brutality
is perceived to be on the rise. Crime seems to be escalating around the nation.
Where is law enforcement when you need them most? We must realize law
enforcement is in a catch 22 situation here in our 21st century. How can they be
expected to do more with fewer resources? Are demands being put on them
improbable for them to realize?
They have fewer personnel to cover the same amount of ground as before when
law enforcement agencies had full rosters to meet all the demands of the
criminal activity.

Data Analytics is the solution to law enforcement’s current

dilemma of shrinking resources
Once the previously mentioned data platform is put into place, then data
analytics can be applied at a very detailed level. Police departments can now
solve crimes faster than ever before. They can stop criminal activity with a
greater success rate. They can even now use data analytics to set in potential
criminal behavioral tendencies and stop these lawbreakers from committing the
crime before they ever get the chance. Here are some technologies Microsoft
uses to help in this quest:
Microsoft Azure Data Lake
Microsoft Cortana Analytics
Windows speech recognition

There is one police department working with Microsoft using their data analytic
tools. Some of these tools are Microsoft Azure, which allows crime analysts to
store criminal data in any size, shape or speed. The beauty of this technology is
that it can be used for any analytics or processing across platforms and
languages. No matter how much data they have, it can be stored all in one data
central system. Do you catch the significance of this?
Law enforcement agencies can network with each other worldwide. This would
be helpful in tracking terrorist activities worldwide. Using predictive analytics
data can be extracted and used to project where terrorists will strike next. They
can decipher the most likely form of attack the group or individuals will use.
What will be the extent of the operation be based on historical data stored from
previous attacks? They can stop the attacks before they materialize. Microsoft
Cortana Analytics: This analytic application allows crime analysts to predict
crime scenarios using machine-learning algorithms at lightening quick rates of
speed. The analysts can also use this tool to support and enhance decision-

144
making processes by recommending the best course of action to take in the
event of a real-time crime or predictive crime taking place. It also helps them to
automate and simplify critical decisions, which include many criminal variables
that constantly change in real time, giving them a greater capability to make the
right decisions quicker to stop crimes before they happen and while they are
happening.
Windows voice recognition: assists police officers in dictating critical
commands pertaining to a driver’s outstanding warrant history, DUI violations
and ticket violation patterns. Knowing this information quickly in real time
allows the police officer to position himself in a place to stop potential crimes or
traffic violations from occurring. Essentially the police officer could reduce the
threat or actually diffuse a life-threatening event before it occurs.

Microsoft A Central Player In Police Department Crime

Prevention
Microsoft is using their data applications in assisting police departments to
prevent crime before it happens. There is one police department they are
working with, to measure public safety in real time using varied sources. This
particular police department uses a plethora of data streams to analyze live 911
calls, live feeds from cameras and police reports to fight terrorism and other
types of violent crimes. Crimes at all levels can be measured, which allows for
quicker reaction time and the ability to develop strategies to fight these crimes
in real time to proactively stop them before they happen.

Advanced Analytics Used For Deciphering Across The

Board Criminal Activities
Police departments are also using advanced analytics to automatically interpret
crime patterns such as burglaries, homicides, and domestic violence incidents.
This data tool helps police departments quickly sort through large amounts of
data quickly. Police can now identify a burglar’s crime patterns. When will the
burglary be committed and what will be the point of entry into the house or
building? What area will the burglar next try to steal from? What objects will he
steal? Armed with this information police officers can be positioned to stop the
burglaries from happening.
Another facet is to preventing burglaries from happening in certain geographical
areas and in certain timeframes. Let’s say a certain area of the city is vulnerable
to burglaries in certain times of the year. For example, a row of apartment
complexes will be hit in the summer months between 8: 00a.m-12:00p.m.

145
Knowing this, police departments can dispatch police officers to the designated
areas during the crucial time period. Therefore, they can stop a series of
robberies before they start.

Police Departments Can Respond To Major Terrorist

Crimes Successfully After They Occur
Remember the Boston Marathon bombing? Police used a few data sources to
respond to the crime promptly and make arrests quickly. The Boston police used
Social Media outlets to gather huge amounts of citizen data. It was a huge task
to filter through the data, but it sure paid off. The police were able to identify
and track down the 2 suspects.
Cell phone videos and security camera footage were painstakingly pored
through but the police’s task paid off. The police were able to identify the two
suspects in the video. They tracked them down from the images caught on these
two data sources.

146
147
Chapter 16: What about Prescriptive Analytics?

148
What is Prescriptive Analytics?
The latest technical data buzz word-“Prescriptive Analytics” is the third branch
of the big three data branches. It is predicted this analytic model will be the last
of the three. Is it the next groundbreaking data model in technology? This
analytic system is a summation of the previous two-Descriptive Analytics and
Predictive Analytics. Definition of the three Analytic models:
Descriptive Analytics-uses Data Aggregation and Data mining techniques,
to give enlightenment to the past. Answering this question “What has
happened?”
Predictive Analytics-uses statistical models and forecasting techniques to
giving insight into the future. Answers this question “What will happen?
Prescriptive Analytics- uses optimization and simulation algorithms to
options for the future “What should we do?”

The prescriptive analytics model is the newest branch in the data world. Some
are already arguing it is the only way to go in the analytical world in the coming
future. It doesn’t seem like Prescriptive Analytics hasn’t caught fire in the
business world as of now. It will become more widespread in different
industries. Companies will realize the benefits of having a model that will
suggest answers to future questions which is the most advanced component of
this relatively new technology.
It takes predictive analytics one-step further. It goes beyond predicting what
will happen in the future. It gives options for the future then it suggests the best
option for that course of action. It has some “artificial intelligence” involved in
its processes. It analyzes the optimization and simulation algorithms then
“thinks” of all the options based on those systems of algorithms. It will even
suggest the best option to take out of all the options analyzed. Prescriptive
Analytics are comparatively speaking difficult to administer than the other two
analytical models. It also uses machine thinking and computational modeling.
These various techniques are used against several many data sets historical data,
transactional data, real-time data feeds and big data. It also incorporates
algorithms that are released on data with minimal parameters to tell them what
to do. The algorithms are programmed to conform according to the changes in
established parameters instead of being controlled externally by humans. The
algorithms are allowed to optimize automatically. Over time, their ability to
predict future event improves.

149
What Are The Benefits Of Prescriptive Analytics?
Many companies aren’t currently using Prescriptive Analytics in their daily
operations. Some of the huge corporations are using them in their daily
operations. These big companies are effectively using prescriptive analytics to
optimize production and successfully schedule and guide inventory, in their
respective supply chains. To be sure they are delivering the right products at the
right times and to the right places. This also optimizes the customer experience.
You know it is advised to use prescriptive analytics to tell customers what to do.
This can be a very effective management tool for end-users.

150
The Future Of Prescriptive Analytics
What does the future hold for prescriptive analytics in the international business
world? In 2014, there were approximately 3% of the business world was using
prescriptive analytics. Scott Zoldi, Chief Analytics Officer, at FICO, made the
following predictions concerning prescriptive analytics for 2016:
Cornerstone of prescriptive analytics “streaming analytics” will become
dominant in 2016
Streaming Analytics is applying transaction-level logic to real time
events (They have to occur in a prescribed window-last 5 seconds, last
10,000 observations, etc.)
Prescriptive analytics will be a must have cyber security application
(This process will analyze suspicious behavior in real time)
Prescriptive analytics will go mainstream in lifestyle activities in 2016
(From home appliances to automated shopping it is poised for explosive
growth)

Analytical experts are predicting the next major breakthrough for prescriptive
analytics will be it will go mainstream spreading across all industries. In 2015, it
was predicted that prescriptive analytics would increase in consultancy
partnering with academics. As of 2016, there still was no off-the-shelf general
purchase for prescriptive analytics, which seems to be a long way off.
It seems to be a hit and miss preposition for prescriptive analytics within
departments of the same company. One department will be using prescriptive
analytics while the other department won’t be.

151
Google Using Prescriptive Analytics For Their “Self-
Driving Car”
The search engine giant is using prescriptive analytics in their five-year-old self-
driving cars. The cars drive automatically without a human guide. They make
lane changes, turn right or left, slow down for pedestrians and stop at lights.
Generally, they drive like any other car but without the human driver.
The car uses prescriptive analytics to make every decision it will need to make
during the driving experience. It makes decisions based on future events. For
instance, the car approaches a traffic light intersection it has to decide whether
to turn right or left. It will take immediate action on future possibilities
concerning which way to turn. The factors it considers before it makes that
decision are many. It considers what is coming towards it in terms of traffic,
pedestrians, etc. before it determines which way to turn. It even considers the
effect of that decision before it makes the decision to turn right or left at the
intersection. Google being such a high profile multi-billion-dollar company will
have a major impact on prescriptive analytics moving forward across the
business world.

152
The Oil and Gas Industry is using Prescriptive Analytics
The oil and gas industry is currently using prescriptive analysis to interpret
structured and unstructured data. It also uses this analytical model to maximize
fracking (is the process of injecting liquids at high pressure to subterranean
rock, boreholes, etc. to force open existing fissures to extract out oil and gas)
Here are applications the oil and gas industry is using prescriptive analytics in:
Maximize scheduling, production and tune the supply chain process, so
the right products are shipped to the right person, at the right time and to
the right location.
Maximize customer experience
Locate functioning and non-functioning oil wells
Optimize the equipment materials needed to pump oil out of the ground
The above tasks are to deliver the right products at the right time to the
right customers at the perfect time

Most of the world’s data (about 80%) are unstructured as videos, texts, images,
and sounds. The oil and gas industry used to look at images and numbers but in
separate silos. The advent of prescriptive analytics changed this. Now, the
industry has the ability to analyze hybrid data- the combination of structured
and unstructured gives the industry a much clearer picture and more complete
scenario of future opportunities and problems. This, in turn, gives them the best
actions to determine more favorable outcomes. For example, to improve the
hydraulic fracking process the following datasets must be analyzed
simultaneously:
Images from well logs, mud logs, and seismic reports
Sounds from fracking from fiber optic sensors
Texts from drillers and frack pumpers’ notes
Numbers from production and artificial lift data

Along these lines, hybrid data is essential in analyzing because of the multi-
billion-dollar investment and drilling decisions made by oil companies. They
must know where to drill, where to frack and of course how to frack. Any one
of these steps is skipped or done incorrectly it could have disastrous results for
these energy companies. The cost could be astronomical. It would be
unimaginable and a waste of time, energy and manpower if they fracked or

153
drilled in the wrong way or place.
There are more to the components involved in completing the prescriptive
analytics. Scientific and computational disciplines must be combined to
interpret different types of data. For instance, to algorithmically interpret images
like (log wells) machine thinking must be connected with pattern recognition,
computer vision, and image processing. Mixing these various disciplines
enables energy companies to visualize a more holistic system of
recommendations: where to drill and frack, while minimizing issues that might
come along the way. By forming detailed traces-using data from production,
subsurface, completion, and other sources–energy companies are able to predict
functioning and non-functioning wells in any field.
This process is anchored by prescriptive analytics technology’s ability to
digitize and understand well logs to produce dispositional maps of the
subsurface, knowing where to drill oil companies save untold millions in
resources. This process helps them to pinpoint where non-functional wells are
located, obviously saving them optimum drilling time. It needs to be said that
the environmental impact is minimized on the particular landscape. Prescriptive
analytics should be used in other areas of oil and gas production. In both
traditional and non-traditional wells, simply by using data from pumps,
production, completion and subsurface characteristics, oil companies are able to
predict failure of submersible pumps and reduce production loss. ( Again saving
untold millions of dollars.)
The Apache Corporation is using prescriptive analytics to predict potential
failures in pumps that extract oil from the subsurface, thereby minimizing the
associated costly production loss due to the failure of these pumps. Another
potential application of prescriptive analytics is too hypothetically predicting
corrosion development on existing cracked oil pipelines. The prescriptive model
could describe preemptive and preventive actions, by analyzing video data from
cameras. It could also analyze additional data from robotic devices known as
“smart pigs” that are positioned within these damaged pipelines.
Using prescriptive analytics in their daily operations will help oil companies
make wiser decisions. This will result in fewer manpower accidents, reduce
production loss, maximize resources and increase financial profits. Another
valuable benefit will be to reduce the environmental impact in the particular
landscapes and surrounding communities. Oil Companies will put themselves in
a favorable light with environmentalists. By showing concern and taking steps
to minimize or reduce environmental hazards around their operating facilities.
Hopefully, the oil companies will see fit to pass on the cost-savings to the
consumers in the form of reduced gas prices, from their reduced costs using the
prescriptive analytics model. Further investigation into the oil and gas industry
reveals they use prescriptive analytics in deeper ways. They use it to locate the

154
oil fields with the richest concentrations of oil and gas. It helps track down oil
pipes with leaks in it. When caught early these leaks can be repaired, therefore,
a major environmental hazard from an oil spill. Prescriptive analysis also helps
refine the fracking process to avoid endangering the environment and to
improve the output of oil and gas.
Conversely, the oil and gas industry will have to increase expenses to hire the
proper experts to run the prescriptive analytics models. The training involved in
getting these technicians up to speed with the company operations will further
inflate company costs. But streamlining all the processes outlined earlier in this
chapter will outweigh the increased costs to increase manpower. To further
justify the hiring of analysts, statisticians and of course data scientists is a
necessary expense. The prescriptive analytics model is more complicated to
implement than the descriptive and predictive analytical models. But the
potential to change industries is a given as the prescriptive analytics system has
already drastically changed the oil and gas industries.

155
Prescriptive Analytics Is Making Significant Inroads Into
The Travel Industry
The Travel Industry has come on board and is using Prescriptive Analytics in
their daily activities as well as in their overall operational functions. Prescriptive
analytics calls for many large data sets. Because of this factor the travel industry
sees great potential in this latest round of data analytics. Online traveling
websites like airline ticketing sites, car rental sites or hotel websites have seen
the tremendous benefits of predictive analytics in their local business functions.
They are implementing prescriptive analytics to filter through many multiple
and complex phases of travel factors, purchase and customer factors such as
demographics and sociographics, demand levels and other data sources to
maximize pricing and sales. Predictive analytics may turn the travel industry
upside down as it did the gas and oil industry. It could be said that predictive
analytics is like reinventing the wheel in data analytics. It is more complicated
in its daily inter-facing than its two earlier predecessors have been.
Other applications the travel industry is finding applicable for prescriptive
analytics are as follows: Segmenting through potential customers predicated on
multiple data sets on how to spend marketing dollars. (By doing this the travel
industry can optimize every dollar to attract a wider customer base. They will be
able to predict where customers preferred traveling locations are based on past
traveling preferences. This will result in marketing and advertising strategies
that will appeal to the particular target age groups.)
An example of this: The Intercontinental Hotel Group currently utilizes 650
variables on how to spend their marketing dollars to increase their customer
base.

156
Other Industries Using Prescriptive Analytics
The Healthcare Industry is waking up to the cutting edge technology that
prescriptive analytics can provide for their intricate overall operational
activities. Using so many variables offers doctor, nurses and physician assistants
can choose the optimal options for the best treatment programs for their
patients. Also, the prescriptive analytics model can suggest which treatment
would be the best fit for the illness or physical condition of the patient. This will
streamline the diagnostic process for the medical profession. Imagine the hours
saved because the many options for any medical condition will be populated
through the many data sets used in the prescriptive analytics model.
One very successful example of a medical company saving a large amount of
money is noted as follows: The Aurora Health Care Centre was able to improve
healthcare and reduce re-admission rates by a healthy 10% resulting in a
significant savings of $ 6 million dollars this is the type of potential prescriptive
analytics has in impacting industries worldwide.
The Pharmaceutical industry will not be left out of the loop either. Prescriptive
analytics can help the industry across the board by helping it to reduce drug
development and minimize the time it takes to get the medicines to the market.
This would definitely reduce drug research expenditures and greatly reduce
manpower hours and resources. Also, drug simulations will quicken the time it
takes to improve the drugs and patients will be easier to find to trial the
medications on.
We can see that prescriptive analytics is the wave of the future in the data
analytics world. The model does require massive amounts of data sets, though.
Currently, only the largest size corporations have the amount of data sets to run
the prescriptive analytics system feasibly. If the data sets are not available in
great quantities, the system absolutely will not work. This is a drawback for
smaller companies who don’t have the large quantities of data sets to effectively
run the prescriptive analytics in their operations. This could be one reason why
this model has not taken off to this point. But the model is starting to make
inroads into the larger corporate settings. As they realize the great potential this
analytical model has. It could take 5-10 years for prescriptive analytics to
become a household system in businesses at the International level.
The measurable impact that prescriptive analytics has had in the gas and oil and
travel industries just cannot be ignored. Truly, Prescriptive Analytics could be
the dominant data system of the near future. It has made noteworthy inroads
thus far.

157
Data Analysis And Big Data Glossary
The following words and terms are related to data analysis, big data and data
science. While not all of them have been used in this book, they are words and
terms that you will come across in your studies. The whole list of terms is quite
exhaustive and would take forever to list so I have chosen to give you
definitions for the more common terms

A
ACID Test - this is a test that is applied to data to test for consistency,
durability, isolation and atomicity – in other words, to make sure it all works as
it should
Aggregation – the collection of data from a number of different databases for
analysis or data processing
Ad hoc Reporting – reports that are generated for one-off use
Ad Targeting – an attempt made by business to use a specific message to reach
a specific audience. This usually takes the form of a relevant ad on the internet
or by direct contact.
Algorithm – mathematical formula that s inserted into software to analyze a
certain set of data
Analytics – the art of using those algorithms and certain statistics to determine
the meaning of the data
Analytics Platform – either software or a combination of software and
hardware that provides you with the computer power and tools that you need to
perform different queries
Anomaly Detection – The process by which unexpected or rare events are
identified in dataset, these events will not conform to any other events in the
same dataset
Anonymization - the process of severing the links between people and their
records in a database. This is done to prevent the source of the records being
discovered
API – acronym for Application Program Interface. These are programming
standards and a set of instructions that enable you to access or build software
applications that are web-based
Application - a piece of software designed to perform a certain job or a whole
suite of jobs

158
Artificial Intelligence – the apparent ability of a computer to apply previously
gained experience to a situation in the way that a human being would
AIDC – Automatic Identification and Capture – any method by which data is
identified and collected on items and then stored in a computer system. An
example of this would be a scanner that collects data through an RFID chip
about an item that is being shipped

B
Behavioral Analytics – the process of using data collected about people’s
behavior to understand that intent and then predict their future actions
Big Data – there are many definitions for the term but all of them are pretty
similar. The first definition came from Doug Laney in 2001, then working for
META Group as an analyst. The definition came in the form of a report called
“3-D Data Management: Controlling Data Volume, Velocity and Variety”.
Volume is referring to the size of the datasets and a McKinsey report, called
“Big Data: The Next Frontier for Innovation, Competition and Productivity”
goes further on this subject saying “Big data refers to datasets whose size is
beyond the ability of typical database software tools to capture, store, manage,
and analyze”.
Velocity is referring to the speed the data is acquired at and used. Companies
are not only collecting the data at a faster speed, they also want to determine the
meaning of it at a faster rate as well possibly in real-time.
Variety is referring to the different data types available for collection and
analysis as well as the structured data that would be found in a normal database.
There are four categories of information that make up big data:
Machine generated – including RFID data, data that comes from
monitoring devices and geolocation data from a mobile device
Computer log – including clickstreams from some websites
Textual social media – from Twitter, Facebook, LinkedIn, etc.
Multimedia social media – and any other information gleaned from place
like YouTube, Flickr and other sites that are similar
Biometrics – the process of using technology to use physical traits to identify a
person, i.e. fingerprints, iris scans, etc.
Brand Monitoring - the process of monitoring the reputation of your brand
online, normally by the use of software
Brontobyte – a unit representing a vast amount of bytes. Although it is not yet
officially recognized as a unit, a brontobyte has been proposed as a unit measure

159
for data that goes further than the yottabyte sale
Business Intelligence – BI - term that is used to describe the identification of
data, and encompasses extraction and analysis as well

C
CDR – Call Detail Record Analysis - a CDR contains data collected by
telecommunications companies and includes the time and the length of a phone
call. The data is used in a variety of different analytical operations
Cassandra – One of the most popular of the columnar databases normally used
in big data applications. Cassandra is an open source database that is managed
by The Apache Software Foundation
Cell Phone Data - a mobile device is capable of generating a vast amount of
data and most of it can be used with analytical applications
Classification Analysis – data analysis used for assigning data to specific
groups or classes
Clickstream Analysis - the process of analyzing web activity by collecting
information about what a person clicks on a page
Clojure – this is a dynamic program language that is based on LISP and uses
JVM (Java Virtual Machine); suited for use in parallel data processing
Cloud - a wide term covering any service or web based application hosted
remotely
Clustering Analysis – the process of analyzing data and using it to identify any
differences or similarities in data sets; this is so that any that are similar can be
clustered together
Columnar Database/Column-Oriented Database – a database that stores data
in columns, not rows. Row databases could contain information like name,
address, phone number, age, etc., all stored on one row for one person. In
column databases, all of the names would be stored in one column, addresses in
another, ages and telephone numbers in their own columns and so on. The
biggest advantage to this type of database is that had disk access is faster
Comparative Analysis – analysis that compares at least two sets of data or
processes to see if there are any patterns in large sets of data
Competitive Monitoring - using software to automate the process of
monitoring the web activity of competitors
CEP – Complex Event Monitoring - this is the process of monitoring the
events in the systems of any organization, analyzing them and then acting when
necessary

160
Complex Structured Data - structured data made up of at least two inter-
related parts which is therefore difficult for structured tools and query language
to process
CLASS – Comprehensive Large Array-Data Stewardship System – digital
library of historical data gained from NOAA (US National Oceanic and
Atmospheric Association
Computer-Generated Data - data that is generated through a computer instead
of a human being, for example, a log-file
Concurrency – the ability to execute several different processes at once
Confabulation – the process of making a decision based on intuition look as if
it were based on data instead
CMS – Content Management System – software that allows for the
publication and the management of web-based content
Correlation Analysis - A process of determining the statistical relationship
between two or more variables, usually to try to identify if there are any
predictive factors present
Correlation – refers to multiple classes of dependent statistical relationships.
Examples would be the correlation of parents and their children or the demand
for a specific product and its price
Cross-Channel Analytics - analysis that can show lifetime value, attribute sales
or how an average order
Crowdsourcing – the process of asking the public to help complete a project or
find a solution to a problem
CRM – Customer Relationship Management - software used to help manage
customer service and sales

D
Dashboard – graphical report of either static or real-time data on a mobile
device or a desktop. The data is usually high level in order to give managers
access to quick reports of performance
Data – a qualitative or quantative value. Examples of the more common data
includes results from market research, sales figures, readings taken from
monitoring equipment, projections for market growth, user actions on websites
customer lists and demographic information
Data Access – the process of retrieving or viewing data that has been stored
DATA Act – Digital Accountability and Transparency Act 2014 - A

161
relatively new US law that is intended to make it easier to gain access to federal
government expenditure information by requiring the White House Office of
Management and Budget and the Treasury to standardize data on federal
spending and to publish it
Data Aggregation - the collection of data from a number of sources for the
purpose of analysis or reporting
Data Analytics – the use of software to determine meaning of or information
from data. The result may be a status indication, a report or an automatic action,
based on what information is received
Data Analyst – the person who is responsible for preparing, modeling and
cleaning data so that actionable information can be gained from it
Data Architecture and Design - This is the structure of enterprise data. The
actual design or structure will vary, as it is dependent on the result that is
required. There are three stages to data architecture:
The conceptual representation of the entities in the business
The representation, logically, of the relationship between each of the
entities
The construction of the system that supports the functionality
Data Center - physical place that is home to data storage devices and servers
that may belong to one or more organizations
Data Cleansing – the review and revision of data to eliminate any duplicate
information, to correct spelling errors to add in any missing data and to provide
consistency across the board.
Data Collection – a process that captures any data types
Data Custodian – The person who is responsible for the structure of the
database and the technical environment, including data storage
Data Democratization – the concept of ensuring that data is available directly
to all workers in an organization, instead of them having to wait for the data to
be delivered to them by another party, usually the IT department, with the
business
Data-Directed Decision Making – The use of data to support the need to make
crucial decisions
Data Exhaust - the data that is created by a person as a byproduct of another
activity, for example, call logs or web search histories
Data Feed - the means by which data streams are received, for example,
Twitter, or an RSS feed

162
Data Governance - the rule or processes that ensure data integrity and that
management best practices are followed and met
Data Integration - The process by which data from various sources is
combined and presented in one view
Data Integrity – a measure of trust that an organization places in the
completeness, accuracy, validity and timeliness of data
Data Management – The Data Management Association says that data
management should include this practice to ensure the full lifecycle of data is
managed:
Data governance
Data design, analysis and architecture
Database management
Data quality management
Data security management
Master data management and reference
Business intelligence management
Data warehousing
Content, document and record management
Metadata management
Contact data management
DAMA – Data Management Association – non-profit international
organization for business and technical professionals that is “dedicated to
advancing the concepts and practices of information and data management”
Data Marketplace – an online location where people can purchase and sell data
Data Mart - access layer of a data warehouse that provides users with data
Data Migration – the process by which data is moved between different
formats, computer systems or storage places
Data Mining – the process of determining knowledge or patters from large sets
of data
Data Model/Modeling - data models are used to define the data structure
needed for communication between technical and functional people to show
which data is needed for the business. Also used for communication of
development plans for data storage and access among specific team members,
usually those in application development

163
Data Point – individual item on a chart or graph
Data Profiling – the collection of information and statistics about data
Data Quality - the measurement of data to determine if it can be used in
planning, decision making and operations
Data Replication – the process by which information is shared to ensure
consistency between sources that are redundant
Data Repository - the location where persistently stored data is kept
Data Science – this is a relatively recent term that has several definitions. It is
accepted as a discipline that incorporates computer programming, data
visualization statistics, data mining database engineering and machine learning
in order to solve complex problems
Data Scientist - a person who is qualified to practice data science
Data Security – the practice of ensuring that data is safe from unauthorized
access or from destruction
Data Set – a collection of data stored in tabular form
Data Source – a source from which data is provided, such as a data stream or
database
Data Steward - the person who is responsible for the data that is stored in a
data field
Data Structure – a certain method for the storage and organization of data
Data Visualization – Visual abstraction of data that is used for determining the
meaning of the data or for communication the information in a more effective
manner
Data Virtualization – process by which different sources of data are abstracted
through one access layer
Data Warehouse – a place where date is stored for analysis and reporting
purpose
Database - digital collection of data and the structure that the data is organized
around.
Database Administrator – DBA – a person who is usually certified for being
responsible for the support and maintenance of the integrity of content and
structure of a database
DaaS – Database as a Service - a database that is hosted in the cloud and is
sold on a metered basis. Examples of this include Amazon Relational Database
Service and Heroku Postgres

164
DBMA – Database management System - software used for collecting and
providing structured access to data
De-Identification – the process of removing data that links specific information
to a specific person
Demographic Data - data that relates to the characteristics of the human
population, i.e. in a specific area, age, sex, etc.
Deep Thunder – weather prediction service from IBM that provides specific
organizations, such as utility companies, with weather data, allowing them to
use the information to optimize the distribution of energy
Distributed Cache – data cache that spreads over a number of systems but
works as a single system; usually used as a way of improving performance
Distributed File System - a file system that is mounted on several servers at the
same time to enable data and file sharing
Distributed Object - software module that has been designed to work with
other distributed objects that are stored on different computers
Distributed Processing - execution of a specific process over several
computers that are connected to the same computer network
Document Management - the process of tracking electronic documents and
paper images that have been scanned in, and then storing them
Drill - this is an open source system, distributed for carrying our interactive
analysis on very large datasets.

E
Elastic search – open source search engine that is built on Apache Lucene
HER – Electronic Health Records - a digital health record that should be
accessible and usable across a number of different health care settings
ERP – Enterprise Resource Planning - software system that enables a
business to manage resources, business functions and information
Event Analytics – shows the steps taken to lead to a specific action
Exabyte - 1 billion gigabytes or one million terabytes of information
Exploratory Data Analysis - data analysis approach that focuses on identifying
general data patterns
External Data - data that lives outside a system
ETL – Extract, Transform and Load – process used in data warehousing to
prepare data for analysis or reporting

165
F
Failover – the process of automatically switching to a different node or
computer when one fails
FISMA – Federal Information Security Management Act - US federal law
that says all federal agencies have to meet specific information security standard
across all systems
Fog Computing - computing architecture that enables users to have better
access to data and data services by moving cloud services, like analytics,
storage and communication closer to them though a device network that is
distributed geographically

G
Gamification – the process of using gaming techniques in applications that are
not games. This is used to motivate employees and encourage specific behaviors
from customers. Data analytics often is applied to this in order to personalize
rewards and encourage specific behaviors to get the best result
Graph Database – a NoSQL database that makes use of graph structures for
semantic queries with edges, nodes and properties that store, query and map
data relationships
Grid Computing - performance of computing functions making use of
resources from systems that are multiple distributed The process involves large
files and is mostly used for multiple applications The systems that make up grid
network do not need to be designed similarly, neither do they have to be in the
same location geographically

H
Hadoop - open-source software library that is administered by Apache Software
Foundation. Hadoop is defined as “a framework that allows for the distributed
processing of large data sets across clusters of computers using a simple
programming model”.
HDF – Hadoop Distributed File System - fault-tolerant distributed file system
that is designed to work on low-cost commodity hardware. It is written for the
Hadoop framework and written in Java
HANA - a hardware/software in-memory platform that comes from SAP. It is
designed for real-time analytic and high volume transactions
HBase – distributed NoSQL database in columnar format

166
HPC – High Performance Computing – Also known as supercomputers.
Normally these are custom-built using state of the art technology, as a way of
maximizing computing performance, throughput, storage capacity and data
transfer speeds.
Hive - a data and query warehouse engine similar to SQL

I
Impala - open source SQL query engine distributed for Hadoop
In-Database Analytics – the process of integrating data analytics into the data
warehouse
Information Management - the collection, management and distribution of all
different types of information, include paper, digital, structured and
unstructured.
In-Memory Database – a database system that relies solely on memory for
storing data
IMDG – In-Memory Data Grid – data storage in memory across a number of
servers for faster access, analytic and bigger scalability.
Internet of Thing – IoT - this is the network of physical things that are full of
software, electronics, connectivity and sensors to enable better value and service
through the exchange of information with the operator, manufacturer and/or
another connected device/s. Each of these things is identified through its own
unique computing system but is also able to interoperate within the internet
infrastructure that already exists.

K
Kafka - open source message system from LinkedIn that is used to monitor
events on the web

L
Latency - a delay in the response from or delivery of data to or from one point
to another one
Legacy System - an application, computer system of a technology that is now
obsolete but is still used because it adequately serves a purpose
Linked Data - this is described by Time Berners Lee, the inventor of the World
Wide Web as “cherry-picking common attributes or languages to identify
connections or relationships between disparate sources of data”.

167
Load Balancing - the process of the distribution of a workload across a
network or cluster as a way of optimizing performance
Location Analytics - this enables enterprise business systems and data
warehouses to use mapping and analytics that is map-driven. It enables the use
of geospatial information and a way to associate it with datasets.
Location Data - data used to describe a particular geographic location
Log File - files created automatically by applications, networks or computers to
record what happens during an operation. An example of this would be the time
that a specific file is accessed
Long Data - this term was created by Samuel Arbesman, a network scientist
and mathematician, referring to “datasets that have a massive historical sweep”

M
Machine-Generated Data - data that is created automatically from a process,
application or any other source that is not human
Machine Learning - a process that uses algorithms to enable data analysis by a
computer. The purpose of this is to learn what need to be done when a specific
event or pattern occurs
Map/Reduce - General term that is referring to the processing of splitting a
problem down into bits, each of which is then distributed among several
computers that are on the same cluster or networks or a map, a grid of
geographically separated or disparate systems. The results are collected in from
each computer and combined into a single report.
Mashup – a process by which different datasets from within one application are
combined to enhance the output. An example of this would be a combination of
demographic data with a real estate listing
MPP – Massively Parallel Processing - the process of breaking up a program
into bits and executing each part separately on its own memory, operating
system and processer
MDM – Master Data Management - master data is any data that is non-
transactional and is critical to business operation, i.e. supplier or customer data,
employee data, product information. MDM is the process of master data
management to ensure availability, quality and consistency
Metadata – data that is used to escribe other data, i.e. date of creation, size of a
data file
MongoDB - NoSQL database that is open source under the management of
10gen

168
MPP Database - database that is optimized to work in an MPP processing
environment
Multidimensional Database - a database that is used to store data in cubes or
multidimensional arrays instead of the usual columns and rows used in
relational databases. This allows the data to be analyzed from a number of
angles for analytical processing and complex queries on OLAP applications
Multi-Threading - the process of breaking an operation in a single computer
system up into several threads so it can be executed faster

N
Natural Language Processing - the ability for a computer system or program
to understand human language Application of this include the ability for humans
to interact with a computer by way of automated translation, speech and by
determine the meaning of data that is unstructured, such as speech or text data.
NoSQL – database management system that avoids using the relational model.
NoSQL can handle large volume of data that don’t follow a fixed plan. It is best
suited to use with large volumes of data that do not need the relational model

O
OLAP – Online Analytical Processing - the process of using three operations
to analyze multidimensional data:
Consolidation – aggregation of available
Drill-down – allow users to see details that are underlying the main data
Slice and Dice – allows users to choose subsets and see them from
different perspective
OLTP -Online Transactional Processing - process by which users are given
access to vast amounts of transactional data in a way that they are able to
determine a meaning from it
ODCA – Open Data center Alliance - a consortium of IT organizations from
around the globe who have a single goal – to hasten the speed at which cloud
computing is migrated
ODS – Operational Data Store – location to store data from various sources so
that a higher number of operations can be performed on the data before it is sent
for reporting to the data warehouse

169
Parallel Data Analysis – the process of breaking an analytical problem up into
several smaller parts. Algorithms are run on each part at the same time. Parallel
data analysis can happen in a single system or in multiple systems
Parallel Method Invocation – PMI - a process that allows programming code
to call many different functions in parallel
Parallel Processing - the ability to execute several tasks at once
Parallel Query - the execution of a query over several system threads to speed
up performance
Pattern Recognition – labeling or clarification of a pattern that is identified in
the machine learning process
Performance management – Process of monitoring the performance of a
business or system against goals that are predefined to identify any specific
areas that need looking at
Petabyte - 1024 terabytes or one million gigabytes
Pig - data flow execution and language framework used for parallel
computation
Predictive Analytics – the use of statistical functions on at least one dataset to
predict future events or trends
Predictive Modeling - the process by which a model is developed to predict an
outcome or trend
Prescriptive Analytics- the process by which a model is created to “think” of
all the possible options for the future. It will suggest the best option to take.
Q
Query Analysis – the process by which a search query is analyzed to optimize
it to bring in the best results

R
R - Open source software environment that is used for statistical computing
RFID – Radio Frequency Identification – Technology that makes use of
wireless communication to send information about a specific object from one
point to another point
Real Time - descriptor for data streams, events and processes that have actions
performed on them as soon as they occur
Recommendation Engine - an algorithm that is used to analyze purchase by a
customer and their actions on a specific e-commerce site; the data is used to

170
recommend other product, including complementary ones
Records Management - the process of management of a business’s records
though their whole lifecycle. From the data of creation to the date of disposal
Reference Data - this is data that describes a particular object and its
properties. This object can be a physical one or t may be virtual
Report – information gained from a dataset query and presented in a
predetermined format
Risk Analysis - the use of statistical methods on at least one dataset to
determine the risk value of a decision, project or action
Root-Cause Analysis – the process by which the main cause of a problem or
event is determined

S
Scalability – ability of a process or a system to maintain acceptable levels in
performance as scope and/or workload increases
Schema – Defining structure of data organization in a database system
Search - the process that uses a search tool to find specific content or data
Search Data - the aggregation of data about specified search times over a
specified time period
Semi Structured Data - data that has not been structured with the use of a
formal data model but provides alternative means of describing the hierarchies
and data
Server – virtual or physical computer that serves software application requests
and sends them over the network
SSD – Solid State Drive – sometimes called a Solid State and is a device that
persistently stores data by sing memory ICs
SaaS – Software as a Service – application software used by a web browser or
thin client over the web.
Storage - any means that can be used to persistently store data
Structured Data – data that is organized with the use of a predetermined
structure
SQL – Structured Query Language – programming language that is
specifically designed to manage data and retrieve it from a relational database

171
T
Terabyte – 1000 gigabytes
Text Analytics - application of linguistic, statistical and machine learning
techniques on sources that are text-based, to try to derive insight or meaning
Transactional Data – Data with an unpredictable nature, i.e. accounts
receivables data, accounts payable, data or data that relates to product shipments

U
Unstructured Data - data that does not have any identifiable structure, i.e.
emails or text messages

V
Variable Pricing - the practice of changing a price on the fly, as a response to
supply and demand. Consumption and supply have to be monitored in real-time

172
173
Conclusion

I am sure that by now, you would have realized the importance of having a
sound system in place to manage your data. In order to effectively manage that
data, you might need to expand your organization to include people that are
skilled in analyzing and interpreting it effectively. With effective data
management, you will find it easier to analyze data.
With the increasing competition, predictive analytics is also gaining more and
more importance, as the days go by. I have shown you several case studies of
large organizations that are using data to further expand and improve their
operations. I hope that the information mentioned in this book gave you an
insight into the field of predictive analytics.
I hope you enjoyed this book and apply the techniques mentioned in this book
in your business.
Lastly, I’d like to ask you a favor. If you found this book helpful or enjoyed this
book, then I’d really appreciate you leaving a review and your feedback on
Amazon by clicking the link below.

Click Here To Leave a Review for This Book on Amazon

Thank you again.

Daniel Covington

174