Next Best Product - Best Approach?




I am currently working on a side-project and stuck up in the possible approach/technique to solve this problem.

I currently have a mock data-set for Insurance Domain which consists of fields like;

  • ID
  • Name
  • Age
  • Gender
  • Martial Status
  • Current Product
  • Premium Paid
  • Claims Made
  • Annual Salary
  • Province
  • Country

I would like to suggest the Next Best Product for the Client (Cross-Sell/Up-Sell), for this approach I decided to first perform Customer Segmentation followed by using Apriori Algorithm and find out the most suitable product for the current customer.

For Customer Segmentation I am using k-Prototype Algorithm.

I would like to know whether my approach is right or is there any better technique to determine which the next best product for Cross-Sell/Up-Sell.

I am using R for solving this problem, and any inputs and suggestion will be helpful.




I have built few NBP solution for my clients. I would like to share my findings, hopefully it helps -

Basically the approach depends on which kind of product you are talking about. Lets take two extreme examples -

Example 1: Bank X wants to X-sell Personal Loan, Credit Card, 2 wheeler Loan, Over Draft, High Ticket Personal loan to its Mass Market base. Contact policy is: Every customer can be called only once in every 2 months.
Best feasible solution would be -
Build response or Look-a-like model for every product, translate the probability to a score and pick the product with maximum score.

Example 2: Music Company has 1 Mn customer base and want to promote only 2-3 albums every week to each customer.
This one, kind of matches with your approach. First do a segmentation, Objective: To see if there is any homogeneous consumer group (it is not necessary that you will get a very useful output from this) and then do a CF to identify 2 most lovable albums for each customer.
You can avoid CF and do Apriori only if you are working for a retail client - Big Bazar, Unilever where every customer buys many products. You cant justify Apriori for Banking, Telecom (like VAS X-sell, who cares these days :slight_smile: ). I should mention that I tried Apriori only in SPSS, not quite sure how different it is in R.

Hope I did not confuse or bore you.


Thanks chiranjit7.