hashtagEPC.

com
 Part A

What are the different career roles related to

Data Science

hashtagEPC.com
Let’s consider a Simple example:

Assume that you are part of an EPC company.

Your company has been bidding to win contracts for FEED scope from
clients. But for the past few years, you are losing the bids to your
competitors.
So, let’s have a typical business case conversation on this case.

hashtagEPC.com
Management: We have been losing our You:
FEED contracts for the past few years. Ok. I will arrange for the
bidding data for past
We got feedback from clients that our years.
bids are not competitive.
I will analyze the projects
As per our clients, we were higher on the where we lost the bids
total hours estimated. and attempt to find
reasons for it.

hashtagEPC.com
Management: How much time do you You: I will need to get all
need? the bidding summaries.

I don’t think all the data is

saved in a common
location.

The file formats may also

be different.

Will take 2 months to

complete the study.

hashtagEPC.com
Management: Will wait to see what you You: See you in two
find out. All the best! months!

hashtagEPC.com
You start your analysis with the below steps:

• Create a folder in LAN drive.

• Add all the files into the folder.

• Review all the data.

• Prepare a table in excel and decide on what attributes (or parameters)
you think will impact the FEED hours.

• Add all the data into the excel table.

• Clean the data (remove the data set which does not have all data
information).

hashtagEPC.com
• Review outliers.
• Use Excel features like pivot, graph to find a correlation between
various data sets.
• Prepare a visualization of key inferences like hours/equipment count,
$/hr, % margin by which you lost the bid, count of how many times
your key competitor won, etc.

hashtagEPC.com
• Obliviously, the analysis till now gives a status of what your company is
doing today. But the business case is about what you should be doing in
the future – that is - make a prediction.

• So, you will try to find out what are some other factors impacting the
FEED hours (i.e find out relations between variables. You may use some
statistical tools to find out these co-relations like sensitivity analysis on
mean, medians, ranges or do regression analysis.

hashtagEPC.com
Finally

• At the end of your analysis, you may replace the simple formula to
estimate FEED hours by a more complex algorithm.

Example, if the existing formula is:

FEED Hours = (Historical hours/equipment count) X Equipment
Count

The new formula based on your analyze may look like:

FEED Hours = {(Historical hours/equipment count) X Equipment
Count} X Project Complexity Factor X Client Specification Factor

hashtagEPC.com
It was found that your company was losing bids because they
were not considering lower hours for:
• Projects which were less complex
• Projects for which work productivity will be higher as the
client specification is known to your company

Based on your recommendation, your

company utilizes the above strategy in the
next bid.

hashtagEPC.com
And the result is that you win the
biggest FEED contract for the
company!!

hashtagEPC.com
 Ok, so nice story.

Now, if you reflect on the case scenario, you will figure out that, at a high level
there are two sides of the story:
• Business side
• Data side

hashtagEPC.com
 BUSINESS SIDE
When you are representing the business side, you should have a good
understanding of the technicality of the business you are representing.

In this case, being from EPC business you should have a thorough understanding
of what is FEED, how FEED hours are estimated, what are different types of
equipment, how FEED work is executed, etc.

Usually, the business side will define the problem, expect the solution to it, and
then make a decision to implement (or not implement) the recommendations.

hashtagEPC.com
 DATA SIDE
When you are representing the data side, you need to have a good
understanding of how you will store data, which file types are best for analysis,
tools/software’s/methods for analyzing the data, statistical methods.

Let us understand the data side in more detail. To do this lets look back into the
steps you took to analyze the FEED hours estimate.

Can we group the steps into some categories?

hashtagEPC.com
 The steps can be grouped into 3 Categories or Roles
Steps Summary of the Steps Role
Create a folder in LAN drive
Add all the files (or data sources) into the folder
Review all the data Extract, Transform and Load
Data Engineer
Prepare a table in excel and decide on what attributes will impact the FEED cost. Data
Add all the data into the excel table
Clean the data
Use Excel features to find a correlation between various data sets
Interpret data, Identify
Prepare a visualization of key inferences like hours/equipment count, $/hr, %
patterns and trends, Prepare Data Analyst
margin by which we lost the bid, count of how many times your key competitor
visualizations
win, etc
Make a prediction – what should be the estimated FEED hours
Find out what are some other factors impacting the FEED hours (i.e. find out
Predictive Analysis,
relations between variables. You may use some statistical tools to find out these
co-relations like sensitivity analysis on mean, medians, ranges or do regression Identify not just existing Data Scientists
analysis trends but factors impacting
At the end of your analysis, you may replace the simple formula to estimate FEED the trend
hours by a more complex algorithm
hashtagEPC.com
 To keep it simple
Data Engineer : Develop, construct, test and maintain data architectures
Data Analyst : Extract data from data warehouse, Interpret data,
Prepare visualizations
Data Scientist : Predictive analysis, complex data sets, not just find the
trend but other factors impacting the trend

Sounds very simple!

In fact, you carried out all the three different roles all by yourself !!
So what’s the fuss about Data Science then?

hashtagEPC.com
 It’s all about complexity and the amount of data

From Simple Case in our example To Complex cases in Reality:

The FEED case we discussed was a very Imagine the analysis to be done for:
simple case where we had:
- Hundred or thousands of files
- A small amount of data - Continues inflow of data
- Simple analytics - Multiple factors affecting the result
- Excel was sufficient tool - No simple/direct inferences
- Pretty clear business case or problem - Multiple cases to analyze
- Just one case to analyze

As data type becomes more complex, amount of data becomes humongous,

predicting co-relations becomes more challenging,
each of the three roles needs specialized tools and specialized skillsets.

hashtagEPC.com
 For Complex Data
Our Data Engineer has to manage a huge amount of data. Depending on the type of industry they are
supporting, they may be getting data regularly (sometimes as frequently as thousands per minute, think of
Twitter, YouTube). So from LAN folders and excel tables, they have to be an expert in developing large data
warehouses and writing queries on data so that others can use it. These need a core computer science (or
software skills).

Data Analysts need to collect data from a warehouse with the help of programming queries, utilize detailed
statistical analysis & data interpretation, rely on enterprise-level tools for analytics and visualization.

Data Scientists don’t just have a simple case to analyze and recommend. They have to go into a sea of data,
find out the co-relations without even knowing what co-relations may exist. And this is where they need to
use tools like Python, Jupyter and technology like Machine Learning, Deep Learning.

hashtagEPC.com
 Part B

Switching to a Career in Data Science

hashtagEPC.com
 Before you make a decision
The analytics work which you are doing presently, (like we explained the FEED analysis) does not
mean that you are ready to start a career in Data Science. For a career in Data Science, you will
need to develop expertise in the core field which includes:
- Understanding Data Architecture
- Programming
- Tools/Software such as SQL, Python, Apache Hadoop, etc
You may also need to MOVE OUT from your current role and become a Data Scientist.
That is, don’t expect that you will be able to continue to be a Project Controls Engineer or a
Project Execution Engineer and wherever there is any case to analyze, you become a Data
Scientist.
For small cases or tasks probably you will be able to manage both your project functions and data
role, but in complex cases, it may not work like this.
Data Science needs a full-time effort.
hashtagEPC.com
 Two Paths towards Data Science
Path A
- Find out how your current company is implementing or utilizing data science.
- If there is already a team implementing data science, try to get into the business side of the analytics
process.
- As you interact with the data team from the Business side, you will get to understand the nuances of
data science.
- Join a training program on Data Science. Don’t assume that you can become a data scientist without
having professional-level training.
- When you get an opportunity, you preferably start as a Data Analyst and not directly as Data Scientist.

But:

- What if your current company is not utilizing data science? You may need to find a company that is doing
it, move into it, and get into the data science process. This may not be an easy task and will take time.
hashtagEPC.com
 Two Paths towards Data Science
Path B
- Go for Master Degree in Data Science
- Use your industry contacts and your business knowledge to do a real world project on data science.

But:
- You will be leaving your existing job. Unless you find a part time course.
- If you have pretty high years of experience (say 8+ years), after completing the course it may be
challenging to find a satisfactory entry level job.

hashtagEPC.com
Read the full blog on our website to understand further about:
- Common Tools/Software used in Data Science
- EPC Project Roles that can easily transition into Data Science career
- Example of Data Science application in EPC project which can be considered for academic projects

hashtagepc.com/featured-blogs