1
Data Curry Hello Readers, By now you must be aware of the power of data science and how you can derive amazing business insights using big data analytics. Today I will share another story on how you can take correct business decisions by backing your intuitions with the power of data science. Test Even Your Surest Assumption With Data These stories are written by Dr. Dakshinamurthy V Kolluru, Chief Advisor – Data Science, Usha Martin Education: President, International School of Engineering (http://www.insofe.edu.in) The best place in the world to learn Applied Engineering. In any project I handle, I always request the business users to share enough information about the business like the names of variables in the database, a 50,000 feet view of the data collected, etc. and leave me alone for a week or two. I explore the data on my own and find amazing truths. Interestingly, several timeswhat I find turn out to be intuition crushers.I am not talking about results derived out of complex models, but insights gained by just looking at the entire data. Recently, I was working on data collected from an e- commerce company. They get a lot of first time sales in December compared to any other month of the year. The distribution of first time sales per month in a year is shown below. Number of customers who made their first purchase by month 1 2 3 4 5 6 7 8 9 10 11 12 Month 6000 4000 2000 Count More or less, this trend is followed each year. Now, everyone in the company somehow believed that customers acquired in Christmas are better(may be because they are more in number when compared to the other months). However, I wanted to validate this and checked the revenue earned fromthe customers in the two years since their first purchase.The following is a mind blowing revelation from the data analyzed: Median Revenue per Customer since their First Purchase As shown in the above graph, actually customers who made their first purchase in December generate a lot less overall revenue. This is a complete contrary to the company's intuition. Now the company has the insight that if it wants to cross sell or upsell its products, it is better off focusing on customers who bought in months other than December, January, and June. So, always ensure that your intuition is supported and validated by data. www.datacurry.com

Test Even Your Surest Assumption With Data - DataCurry Recipe

Embed Size (px)

DESCRIPTION

Another story on how you can take correct business decisions by backing your intuitions with the power of data science.

Citation preview

Page 1: Test Even Your Surest Assumption With Data  - DataCurry Recipe

Data CurryHello Readers,

By now you must be aware of the power of data science and how you can derive amazing business

insights using big data analytics. Today I will share another story on how you can take correct

business decisions by backing your intuitions with the power of data science.

Each story is a very simple one where I know for sure that decision makers gained

powerful insights

You can most probably read these in 5 minutes

You and your team can implement them with equal ease in your own business

I will also write them for the business users and do away with any

engineering/technology stuff

Test Even Your Surest Assumption With DataTest Even Your Surest Assumption With Data

I was taught to always spend quality time with data; you know,

asking questions and exploring answers, prior to starting

complex modelling on the same. I was told, rightly, that this

exercise would give me the needed comfort and intuition with the

data.

In the initial days, I used to explore hundreds of records.

However, in the past few years, thanks to the unbelievable

computing power available, I am able to play with hundreds of

thousands of records even for exploration.

Now, what I find with customer after customer is that the simple

sums and averages I find from this big data itself are the intuition-

crushers that they are looking for.

Here is a customer whose customers predominantly come from

two geographies. When I looked at the entire customer base, I

found that approximately 73% of the customers are from one

geography and 27% from the other.

Now, I verified the distribution amongst high value customers

(top 10%) and found that the ratio changed drastically to 94-6 and

finally in the extremely high value customers (top 5%), it is 98-2.

The graphs provided below tells the story I am talking

about.

So, what could they do with this insight? They could

focus more on selling more to high value customers in

Region1 if they believed it has as much potential as

Region2 or spend more on Region2 as that seems to be

the low-hanging fruit. They may even want to see what

aspects of the product make it dear to Region2.

But, that is not the point. What I presented here is not

complex analytics. It is a simple graph that I plotted

probably in the first half hour but on really big data.

So, before you go for really complex million dollar

models, ensure that you milked all the powerful insights

that you can get from simple sums, averages, etc.

27%

73%

Entire Customer Base

0

20000

40000

60000

80000

Region 1 Region 2

6%

94%

Top 10% customers

0

5000

10000

Region 1 Region 2

98%

Top 5% customers

0

1000

2000

3000

4000

Region 1 Region 2

2%

5000

These stories are written by Dr. Dakshinamurthy V Kolluru, Chief Advisor – Data Science,

Usha Martin Education: President, International School of Engineering (http://www.insofe.edu.in)

The best place in the world to learn Applied Engineering.

In any project I handle, I always request the business

users to share enough information about the

business like the names of variables in the database,

a 50,000 feet view of the data collected, etc. and leave

me alone for a week or two. I explore the data on my

own and find amazing truths. Interestingly, several

timeswhat I find turn out to be intuition crushers.I am

not talking about results derived out of complex

models, but insights gained by just looking at the

entire data.

Recently, I was working on data collected from an e-

commerce company. They get a lot of first time sales

in December compared to any other month of the

year. The distribution of first time sales per month in

a year is shown below.

Number of customers who made their first purchase by month

1 2 3 4 5 6 7 8 9 10 11 12

Month

6000

4000

2000

Co

un

t

More or less, this trend is followed each year.

Now, everyone in the company somehow believed that customers acquired in Christmas are better(may be because they are

more in number when compared to the other months). However, I wanted to validate this and checked the revenue earned

fromthe customers in the two years since their first purchase.The following is a mind blowing revelation from the data analyzed:

Median Revenue per Customer since their First Purchase

As shown in the above graph, actually customers who

made their first purchase in December generate a lot less

overall revenue. This is a complete contrary to the

company's intuition.

Now the company has the insight that if it wants to cross

sell or upsell its products, it is better off focusing on

customers who bought in months other than December,

January, and June.

So, always ensure that your intuition is supported and

validated by data.

www.datacurry.com