The Fifth Elephant 2012

Finding the elephant in the data.

“I know what you are going to do next summer” – Predicting Repeat Purchase Behavior by using Bayesian Hierarchical model and Regression Techniques

Submitted by Biswajit Pal on Friday, 29 June 2012

videocam_off

Technical level

Intermediate

Section

Data Analytics

Session type

Lecture

Status

Confirmed

Vote on this proposal

Login to vote

Total votes:  +9

Objective

A major question in database marketing is that of identifying the customers who are most likely to make a repeat purchase in the near future. We will discuss a scalable repeat purchase scoring algorithm that assign a propensity score i.e. probability to transact in next ‘n’ period of time for each customer. This algorithm presently scores an entire database containing around 100MM customers in couple of hours and thus has the potential to be used in a Big Data scenario.

Description

In the context of targeted marketing to consumers, the ability to tell which customers are more likely than others to make a purchase with HP in the near future greatly enhances effectiveness of any marketing campaign. It helps to rank customers on their propensity to re-purchase, and leads to preferential treatment of the right customers. It also reduces the likelihood of bombarding customers, who are less likely to purchase, with marketing material (over email or postal mail), possibly alienating them from future interest in HP.

The propensity to make a repeat purchase depends on two parameters unique to each customer, the probability of churn and the frequency of transactions. The customer repeat purchase modeling framework we propose, based on a Regression based approximation to a Bayesian hierarchical model, answer these questions. Using the answer to these questions as inputs we predicted the likelihood of a customer making a transaction within a time span into the future (e.g. in the next six months).

Our algorithm can score massive databases for repeat purchase. Further, since it uses only transaction data, it is readily applicable to a wide array of customer segments across different business units

Speaker bio

Biswajit Pal currently holds the position of Project Manager in Customer Intelligence team which is a part of Global Analytics organization within HP. He has 6 years of experience in the areas of Predictive Analytics, Data mining, statistical modeling across verticals like Healthcare, retail and IT and holds a Post Graduate degree in Applied Statistics and Informatics from IIT Bombay.

Subhasish Mishra currently holds the position of Project Manager in strategy team which is a part of Global Analytics organization within HP. He has 5 years of experience in the areas of Predictive Analytics, Data mining, statistical modeling and holds a Post Graduate degree in Economics from Delhi School of Economics.

Manav Shroff is an analytics professional with 11 years of global experience across Technology, Financial Services, Banking and Insurance industries. He is currently leading a team of analytics practitioners supporting HP’s Corporate Customer Intelligence group, providing support in the areas of Customer Analytics (B2C and B2B), Web Analytics, and Market Research, Emerging Markets Strategy, Knowledge Management and Campaign Operations. Manav holds a MBA degree and also is a GE certified green belt and lean coach.

Comments

  • 1
    Arun Anantharaman (@birdonthewire) 6 years ago

    Interesting submission, Manav & team. Thanks!

    Could you also consider submitting a broader e-commerce analytics talk for the "Industry & Business" track - I think there's a potentially compelling story there that you can talk about as well.

Login with Twitter or Google to leave a comment