I have simple dataset here. Supposed I want to find out which customers who bought a certain item are more likely to come back after 10 months.

I have 2 sets of data

  1. The repeat purchase % of users who bought various items
  2. The # of users who bought various items

In the example dataset below, in total 150,000 users bought items, and 20% of them returned after 10 months to buy something.

In the first table, it represents how many users purchased that item X times. For example, D3 means that 60,000 purchased Item A 3 times.

In the second table, it represents how likely the user are to come back after 10 months if they purchased that item X times. For example, D12 means that if a user purchased Item A 3 times, the likelihood they will come back is 34%

My question is - given that there is a natural correlation between people who buy more things and people who come back, how do we find the outliers, and the inflexion points?

I am particularly puzzled by how to use these 2 tables together. Multiplying the related cells together somewhat works, but i wonder if there is a more significant way to do that.

Example Data In Google Docs

enter image description here

  • 1
    $\begingroup$ The data you are using may be oversimplified. If one customer bought item A, then item B, then item A, is he counted as a returning customet twice? If can access the raw data (customer id, date, item purchased) you will be able to analyze the data more easily. $\endgroup$ – KishKash Apr 7 '15 at 21:58
  • 1
    $\begingroup$ hmmm.... this is an example problem set. i have access to the raw data, but i am trying to see if i can arrive at a simplified solution before i go over to the complex one $\endgroup$ – ming yeow Apr 8 '15 at 13:52

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.