churn predictor model


I am working on identifying churn behavior. Let me give details about what I want to do. The customers whose subscription renewal + first charge was due in May, I want to see the usage pattern of those who successfully renewed vs those who cancelled. Further, among those who cancelled whether they were on trial or on the monthly plan and while they were on what what was their usage. Eventually I want to come up with some churn predictor model.

How can I go about doing this in Amplitude


Best answer by Saish Redkar 10 June 2021, 00:41

View original

2 replies

Userlevel 7
Badge +10

Hey @Abeeda 
Firstly you would want to create individual cohorts for users who have successfully renewed vs those who cancelled and also for those on trial vs monthly plans.

We did a similar analysis and from the experience, here are some charts which I found to be useful in understanding the difference in the usage pattern between the different cohorts -

  • Lifecycle Chart
  • Active % Event Usage 
  • Engagement Matrix
  • Average sessions per User
  • Retention Analysis


Hope this helps!


I would also love to hear from other Community members on this :)




Thanks a lot Saish, this gives me a starting point. Will explore this and will share my learnings as well along the way