Improving Decision Making and Structured decisions

Managing Knowledge - by contrast, are repetitive and

routine, and they involve a definite
procedure for handling them so that
Business Value of Enhanced they do not have to be treated each
Decision Making time as if they were new.
❖ semi structured decisions -
Accounts manager when only part of the
- A decision maker that allocate problem has a clear cut
support to most-valuable customers answer provided by an
accepted procedure.
Call center management
- A decision maker that predict call Examples:
center daily demand Senior Management (Unstructured) -
decide entrance or exit from markets,
Inventory manager approve capital budget, and decide
- A decision maker who decide parts long-term goals
inventory levels daily
Middle Management (Semistructured) -
Senior management design marketing plan, develop a
- A decision maker that identify departmental budgets, and design a new
competitive bids from major corporate website
suppliers
Operational Management (Structured) -
Manufacturing manager determine overtime eligibility, restock
- A decision maker who schedule inventory, offer credit to customers, and
production to fill orders determine special offers to customers

Production floor manager The Decision Process

- A decision maker who allocate labor 4 stages in decision making
to complete a job
Intelligence - consists of discovering,
Types of Decisions identifying, and understanding the
problems occurring in the
Unstructured decisions organization—why the problem exists,
- are those in which the decision where, and what effects it is having on the
maker must provide judgment, firm.
evaluation, and insight to solve the
problem. Each of these decisions is Design - involves identifying and exploring
novel, important, and not routine, various solutions to the problem.
and there is no well understood or
agreed procedure for making them. Choice - consists of choosing among
solution alternatives.
Implementation - involves making the analyzing data that come from the business
chosen alternative work and continuing to environment.
monitor how well the solution is working.
Business Analytics - is also a vendor
High-Velocity Automated Decision defined term; it focuses more on tools and
Making techniques for analyzing and
- Humans are eliminated from the understanding data.
decision chain because they are too
slow. The Business Intelligence Environment
- The intelligence, design, choice, and 6 elements of BIE
implementation parts of the decision
making process are captured by ● Data from the business
computer algorithms that precisely environment (Businesses must deal
define the steps to be followed to with both structured and
produce a decision. unstructured data from many
sources, including big data.)
Quality of Decisions and Decision ● Business intelligence
Making infrastructure (The underlying
- Accuracy is one important foundation of BI is a powerful
dimension of quality; in general, we database system that captures all
think decisions are better if they the relevant data to operate the
accurately reflect the real world data. business.)
- Speed is another dimension; we ● Business Analytics toolset (A set
tend to think that the decision of software tools is used to analyze
making process should be efficient, data and produce reports, respond
even speedy. to questions managers pose, and
track the progress of the business by
How do business intelligence and using key indicators of
business analytics support decision performance.)
making? ● Managerial users and methods
(Managers impose order on the
analysis of data by using a variety of
managerial methods that define
strategic business goals and specify
how progress will be measured.)
● Delivery platform—MIS, DSS,
ESS (The results from BI and
analytics are delivered to managers
and employees in a variety of ways,
Business intelligence - is a term depending on what they need to
hardware and software vendors and know to perform their job.)
information technology consultants use ● User interface (Business people
to describe the infrastructure for often learn quicker from a visual
warehousing, integrating, reporting, and representation of data than from a
 dry report with columns and rows of Operational Intelligence and
information.) Analytics

Business Intelligence and Analytics Operational Intelligence - Many decisions

Capabilities deal with how to run the business on a
5 analytic functionalities of BI daytoday basis. These are
largely operational decisions, and this
● Production reports (These are type of business activity monitoring is called
predefined reports based on industry operational intelligence.
specific requirements)
● Parameterized reports (Users Location Analytics and Geographic
enter several parameters to filter Information Systems
data and isolate impacts of
parameters.) Location Analytics - the ability to gain
● Dashboards/scorecards (These business insight from
are visual tools for presenting the location (geographic) component of
performance data users define.) data, including location data from mobile
● Ad hoc query/search/ report phones, output from sensors or scanning
creation (This allows users to create devices, and data from maps.
their own reports based on queries
and searches.) Geographic Information Systems (GIS) -
● Drill down (This is the ability to provide tools to help decision makers
move from a high level summary to visualize problems that benefit from
a more detailed view.) mapping. GIS software ties location data
● Forecasts, scenarios, models about the distribution of people or other
(These include capabilities for linear resources to points, lines, and areas on a
forecasting, what if scenario map.
analysis, and data analysis, using
standard statistical tools.) Support for Semistructured
Decisions
Predictive analytics use statistical
analysis, data mining techniques, historical Sensitivity Analysis - models ask what if
data, and assumptions about future questions repeatedly to predict a range of
conditions to predict future trends and outcomes when one or more variables are
behavior patterns. changed multiple times.

Big Data Analytics Predictive analytics are Pivot Table - Spreadsheets have a similar
starting to use big data from both private feature for multidimensional analysis.
and public sectors, including data from
social media, customer transactions, and
output from sensors and machines.
The Balanced Scorecard and What are the business benefits of using
Enterprise Performance Management intelligent techniques in decision making
and knowledge management?
Balanced scorecard method - is a
framework for operationalizing a firm’s Intelligent techniques (consist of expert
strategic plan by focusing on measurable systems, case based reasoning, genetic
outcomes of four dimensions of firm algorithms, neural networks, fuzzy logic,
performance: financial, business process, and intelligent agents.)
customer, and learning and growth. Artificial Intelligence (AI) - consists of
computer based systems (both hardware
4 Dimensions of Firm Performance: and software) that attempt to emulate
● Financial human behavior and thought patterns.
● Business Process
● Customer Expert Systems
● Learning and Growth - captures human expertise in a
limited domain of knowledge as a
Key Performance Indicators (KPIs) - set of rules in a software system that
measures proposed by senior management can be used by others in the
for understanding how well the firm is organization.
performing along any given dimension. - Expert systems model human
knowledge as a set of rules that
Business Performance Management collectively are called the
(BPM) - attempts to translate a firm’s knowledge base.
strategies (e.g., differentiation, low cost - The strategy used to search through
producer, market share growth, and scope the collection of rules and formulate
of operation) systematically into operational conclusions is called the inference
targets. engine.

Group Decision-Support Systems Case-Based Reasoning (CBR)

- are interactive computer based - Knowledge and past experiences
systems that facilitate the solution of human specialists are
of unstructured problems by a set of represented as cases and stored in
decision makers working together as a database for later retrieval when
a group in the same location or in the user encounters a new case with
different locations. similar parameters.

Collaboration Meeting Rooms Hybrid Fuzzy Logic Systems

(CMR) - allows groups of employees to - is a rule based technology that
meet using any device, via WebEx video represents such imprecision by
software, which does not require any creating rules that use
special network connections, special approximate or subjective values.
displays, or complex software. - cool is overlapped by cold or norm.

Neural Networks
 - are used for solving complex,
poorly understood problems for Structured knowledge is explicit
which large amounts of data have knowledge that exists in formal
been collected. documents as well as in formal rules that
organizations derive by observing experts
Genetic Algorithms and their decision making behaviors.
- are useful for finding the optimal
solution for a specific problem by Enterprise content management (ECM)
examining a very large number of systems help organizations manage both
alternative solutions for that types of information. They have capabilities
problem. for knowledge capture, storage, retrieval,
- works by representing a solution as distribution, and preservation to help
a string of 0s and 1s. firms improve their business processes and
decisions.
Intelligent Agents
- are software programs that work in Learning management systems (LMS)
the background without direct provide tools for the management,
human intervention to carry out delivery, tracking, and assessment of
specific, repetitive, and predictable various types of employee learning and
tasks for an individual user, business training.
process, or software application.
Massive Open Online Courses (MOOCs)
- is an online course made available
What types of systems are used for via the web to very large numbers
enterprise-wide knowledge management of participants.
and knowledge work, and how do they
provide value for businesses? Knowledge work systems (KWS) - are
specialized systems for engineers,
Knowledge management - refers to the scientists, and other knowledge workers
set of business processes developed in that are designed to promote the creation
an organization to create, store, transfer, of knowledge and ensure that new
and apply knowledge. knowledge and technical expertise are
- There are two major types of properly integrated into the
knowledge management systems: business.
enterprise wide knowledge
management systems and Contemporary CAD systems are capable
knowledge work systems. of generating realistic looking three
dimensional graphic designs that can be
Enterprise-wide knowledge management rotated and viewed from all sides.
systems - are general purpose, firm wide
systems that collect, store, distribute, and Virtual reality systems - use interactive
apply digital content and knowledge. graphics software to create computer
- tacit knowledge and is rarely generated simulations that are so close to
written down.
reality that users almost believe they are - Velocity (the speed of data
participating in a real world situation. processing)
- Variety (the number of types of data)
Augmented reality (AR) is a related
technology for enhancing visualization. AR Reasons these are helpful to
provides a live direct or indirect view of a accountants:
physical real world environment whose Auditing - traditional process toward a
elements are augmented by virtual more automated one, which will allow audit
computer generated imagery. professionals to focus more on the logic and
rationale behind data queries and less on
Virtual Reality Modeling Language the gathering of the actual data.
(VRML) - is a set of specifications for Financial Reporting - substantially
interactive, three dimensional modeling on improve the quality of the estimates and
the World Wide Web that organizes multiple valuations.
media types, including animation, images, Taxes - Organizing tax data into a data
and audio, to put users in a simulated real warehouse to be able to consistently
world environment. model and query the data is an important
step toward developing the capability to
investment workstations - leverage the perform tax data analytics.
knowledge and time of its brokers, traders,
and portfolio managers. The Data Analytics Process Using the
IMPACT Cycle
Data Analytics in Accounting and
Business Identify the questions ( understanding a
business problem that needs addressing)
Master the data (requires one to know what
Data Analytics - the process of evaluating
data are available and whether those data
data with the purpose of drawing
might be able to help address the business
conclusions to address business
problem)
questions.
Perform the test plan (We take from all
- This provides a way to search
available data and see if we can identify a
through large structured and
relationship between the response or
unstructured data to discover
dependent variables and those items that
unknown patterns or relationships
affect the response (also called predictor,
explanatory, or independent variables))
Big Data - refers to data sets that are too
large and complex for businesses’
Classification - assign each unit (or
existing systems to handle utilizing their
individual) in a population into
traditional capabilities to capture, store,
a few categories.
manage, and analyze these datasets.
Regression - estimate or predict, for each
unit, the numerical value of some variable
3 Vs:
using some type of statistical model.
- Volume (the sheer size of the
Similarity Matching - identify similar indiv
dataset)
Clustering - divide individuals (like Data Preparation and Cleaning
customers) into groups (or clusters) in a
useful or meaningful way. Mastering the data - requires you to identify
Co-occurrence Grouping - discover and obtain the data needed for solving the
association associations between problem.
individuals based on transactions involving The ETL process is made up of the
them. following five steps:
Profiling - characterize typical behavior of Step 1 Determine the purpose and scope
individual, group or population of the data request (extract).
Link Prediction - predict relationship Step 2 Obtain the data (extract).
between two data items. Step 3 Validate the data for completeness
Data Reduction - reduce the amount of and integrity (transform).
information that needs to be considered to Step 4 Sanitize the data (transform).
focus on the most critical items. Step 5 Load the data in preparation for
data analysis (load)
Address and Refine results (Data analysis
is iterative. We slice and dice the data, find Unified modeling language diagram
correlations, ask ourselves further - support relational database
questions, ask colleagues what they think,
and revise and rerun the analysis) Relational Database
Communicate insights (insights are formed Data are complete
by decision makers and are communicated) Data aren't redundant
Track Outcomes ( Some outcomes will be Data follows business rules and internal
continuously tracked) controls
Data aids communication and integration of
business processes
An accountant can:
● Articulate business problem Primary keys - is typically made up of one
● Communicate with data scientist column, but it can occasionally be made up
● Draw appropriate conclusion and of a combination of columns. The purpose
make recommendation of the primary key is to ensure that each
● Present result in accessible manner row in the table is unique, so it is often
referred to as a “unique identifier.”
7 skills of analytic-minded accountants
➔ Develop an analytics mindset Foreign Keys - another type of attribute,
➔ Data scrubbing and data preparation and its function is to create the relationship
➔ Data quality between two tables.
➔ Descriptive data analysis
➔ Data analysis through data Composite Keys -two foreign keys from the
manipulation tables that it is linking combine to make up a
➔ Define and describe data problems unique identifier.
through statistical data analysis
➔ Data visualization and data reporting
Descriptive attributes - provides Class - manually assigned category applied
meaningful business information but do not to a record based on an event
contribute to database structure
2 types of approaches
Data dictionary - paramount in helping Unsupervised - no specific question and
database administrators maintain databases are simply exploring the data for potential
and analysts identify the data they need to patterns of interest
use. Supervised - historical data to predict
future outcome
Extraction, Transformation, and
Loading (ETL) of Data Unsupervised

Extraction - Requesting data is often an Clustering - natural groupings within the

iterative practice, but the more prepared you group
are when requesting data in the first place, Profiling - typical behavior in a data
the more time you will save for yourself and Data reduction - filter and reduce the
the database administrators in the long run. information
Co-occurrence grouping - discover
Transformation - Anytime data are moved association
from one location to another, it is possible
that some of the data could have been lost Supervised
during the extraction. Classification - predict whether a new
vendor belong to one class
Loading - It is so simple, in fact, that if your Regression - to predict specific values
goal is to do your analysis in Excel and you Causal modeling - an event influences
have already transformed and cleaned your another
data in Excel, you are finished. Similarity Matching - group data by
attributes
Flat File - A means of storing data in one Link prediction - predicts whether a
place, such as in an Excel spreadsheet, as connection will for between two nodes
opposed to storing the data in multiple
tables, such as in a relational database. Profiling approach - process of examining
source data and understanding structure,
content, and interrelationship between data
Modeling and Evaluation: Going
- uses structured data
from Defining Business Problems
and Data Understanding to Steps of Data Profiling
Analyzing Data and Answering 1. Identify the objects or activity you
Questions want to profile.
2. Determine the types of profiling you
Data Specific Terms: want to perform
Target - an expected attribute or value that 3. Set boundaries or thresholds for the
we want to evaluate activity.
 4. Interpret the results and monitor the A filter might require the use of a
activity and/or generate a list of computer-assisted technique, called fuzzy
exceptions. match, to match addresses that do not
5. Follow up on exceptions. perfectly match 100 percent.

Structured Data- refers to organized XBRL (eXtensible Business Reporting

information in formats like databases or Language) is used to facilitate the
spreadsheets exchange of financial reporting information
between the company and the Securities
Importance of Data Profiling and Exchange Commission (SEC).
- Data quality assurance
- Improve decision making Data Regression
- Informed data cleaning Goal: understand relationships between
- Efficiency different variables; to predict future
outcomes based on those relationships
Benford's Law - describes the relative Purpose: Support informed strategic
frequency distribution for leading digits of choices; Focus on high-impact factors for
numbers in datasets. better outcomes
- is an observation about the
frequency of leading digits in many Regression Analysis Process
real-life sets of numerical data. - Identify variables that might predict an
outcome
- Determine the functional form of the
Data Reduction relationship
Goal: obtain a smaller representation of - Identify the parameters of the model
data
Purpose: Complex data analysis may take Goal of Classification
a very long time to run on the complete data to predict whether an individual we know
set. Therefore, data reduction attempts to very little about will belong to one class or
reduce information to focus on the most another.
critical, interesting, or abnormal items
- The data reduction approach is done Here are the general steps:
primarily using structured data—that 1. Identify the classes you wish to predict.
is, data that are stored in a database 2. Manually classify an existing set of
or spreadsheet and are readily records.
searchable. 3. Select a set of classification models.
4. Divide your data into training and testing
sets.
Data Reduction Steps 5. Generate your model.
1. Identify the attribute you would like 6. Interpret the results and select the
to reduce or focus on “best” model.
2. Filter the results
3. Interpret the results Classification Terminology
4. Follow up on the results
Training Data - are existing data that
manually evaluated and assigned data Data Clustering in Auditing
- Anomaly detection in payments to
Test Data - are existing data used to insurance beneficiaries, suppliers, etc.
evaluate the model
Visualization: Using Visualizations
Decision Trees - are used to divide data and Summaries to Share Results
into smaller groups
with Stakeholders
communicating results is more art than
Decision Boundaries - mark the split
science.
between one class and another

Pruning - removes branches from a

Quadrants 1 and 3 versus Quadrants
decision tree to avoid overfitting the model 2 and 4: Qualitative versus
Quantitative
Linear Classifiers
- are useful for ranking items rather than Qualitative data - are categorical data.
simply predicting class probability - Nominal data is the simplest form of
data.
Support Vector machine - Ordinal data can also be counted
- is a discriminating classifier that is defined and categorized like nominal data
by a separating hyperplane that works first but can go a step further—the
to find the widest margin (or biggest pipe) categories can also be ranked.
and then works to find the middle line - The proportion is calculated by
counting the number of items in a
Evaluating Classifiers particular category, then dividing that
Overfitting - Rarely will datasets be so number by the total number of
clean that you have a clear decision observations. (ordinal data but with
boundary. the use of quantitative data)
- ALWAYS BE WARY OF CLASSIFIERS
THAT ARE TOO ACCURATE. Quantitative data - are more complex than
- HAVE A GOOD AMOUNT OF qualitative data because not only can they
ACCURACY BEING TOO PERFECT be counted and grouped just like qualitative
data, but the differences between each data
Sweet Spot point are meaningful—when you subtract 4
- where the model is most accurate without from 5, the difference is a numerical
being so complex to thus allow classification measure that can be compared to
of both the training as well as the test data subtracting 3 from 5.
- The normal distribution is a
Data Clustering phenomenon that many naturally
- The clustering data approach works to occurring datasets in our world
identify groups of similar data elements follow, such as SAT scores and
and the underlying drivers of those heights and weights of newborn
groups. babies.
 - Standard normal distribution - A
special case of the normal
distribution used for standardizing
data. The standard normal
distribution has 0 for its mean (and
thus, for its mode and median, as
well), and 1 for its standard
deviation.
- Standardization - The method used
for comparing two datasets that
follow the normal distribution.
- Ratio data have a meaningful 0 and Charts Appropriate for Qualitative Data
interval data do not. The most
sophisticated type of data on the When it comes to visually representing
scale of nominal, ordinal, interval, qualitative data, the charts most frequently
and ratio. considered for depicting qualitative data are:
- Interval data do not have a • Bar charts (a bar chart is more easily
meaningful 0; in other words, in interpreted than a pie chart because our
interval data, 0 does not mean “the eyes are more skilled at comparing the
absence of” but is simply another height of columns (or the lengths of
number. horizontal bars, depending on the
- Discrete data are data that are orientation of your chart) than they are at
represented by whole numbers. comparing sizes of pie, especially if the
- Continuous data are data that can proportions are relatively similar.)
take on any value within a range. • Pie charts (most famous (some would say
infamous) data visualization for qualitative
Declarative versus Exploratory data. It shows the parts of the whole; in
other words, it represents the proportion of
Declarative visualizations are the product each category as it corresponds to the
of wanting to “declare” or present your whole dataset.)
findings to an audience. • Stacked bar chart
• Tree maps and heat maps (These are
Exploratory visualizations are made when similar types of visualizations, and they both
the lines between steps P (perform test use size and color to show proportional size
plan), A (address and refine results), and C of values.)
(communicate results) are not as clearly • Symbol maps ( are geographic maps, so
divided as they are in a declarative they should be used when expressing
visualization project. qualitative data proportions across
geographic areas such as states or
countries.)
• Word clouds ()are formed by counting the
frequency of each word mentioned in a
dataset; the higher the frequency
(proportion) of a given word, the larger and
bolder the font will be for that word in the Tableau earns high praise for being
word cloud.) intuitive and easy to use, which makes it
ideal for exploratory data analysis.
Charts Appropriate for Quantitative
Data FURTHER REFINING YOUR CHART TO
COMMUNICATE BETTER
• Line charts (Show similar information to
what a bar chart shows, but line charts are ● Data Scale and Increments
good for showing data changes or trend ● Color
lines over time.)
• Box and whisker plots (Useful for when COMMUNICATION: MORE THAN
quartiles, median, and outliers are required VISUALS— USING WORDS TO PROVIDE
for analysis and insights) INSIGHTS
• Scatter plots (Useful for identifying the
correlation between two variables or for ● Content and Organization
identifying a trend line or line of best fit.) (IMPACT cycle)
• Filled geographic maps (As opposed to ● Audience and Tone (Carefully
symbol maps, a filled geographic map is considering your audience is critical
used to illustrate data ranges for to ensuring your communication is
quantitative data across different effective.)
geographic areas such as states or ● Revising (Just as you addressed
countries) and refined your results in the fourth
step of the IMPACT model, you
should refine your writing.)

Tools to Help When Picking a Visual

Tableau - is a newer product and has

placed the majority of its focus on data
visualization.

Microsoft - has a much more robust

platform for data analysis.