Let us assume that I’m working for some company A, and my manager has asked me to build a churn prediction engine given that the product to which the churn prediction is to be done is unknown. The engine must be a generalized one, that is it should predict the potential churns for any product data that is fed into it.

I have found a few basic parameters that are required for such a generalized engine. Recency, Frequency and Monetary value can be utilized for a generalized engine to determine the potential churns in any product. What are all the other generalized parameters like the above mentioned?

Say, you don’t know Machine Learning but how would you say if a person would churn or not with a parameter other than his subscription details, his activity or the complaints he made. What other parameters that would make sense here?

  • $\begingroup$ Have you considered contextual information about the customer, like Age, Gender, Location, Employment Status, Industry and so on? $\endgroup$
    – Dan Scally
    Commented Jan 14, 2020 at 10:37
  • $\begingroup$ @DanScally But how does that work as an early indicator of churn like how Recency, Frequency work? It is like you don't get to know the statistics of the people who already churned. And you're new to making one, so what would you consider as an early indicator other than this? $\endgroup$
    – Blackdeath
    Commented Jan 17, 2020 at 9:23
  • $\begingroup$ The likelihood that a customer will churn is probably going to be conditional on those things. Age and Gender are indicators of which services a person is likely to subscribe to, or remain subscribed to. Location, Employment Status, Industry and so on are all indicators of a person's disposable income, which is likely to have an affect on their Churn rate. Those aren't likely to be as informative as the RFM values, but I would expect them to be informative to a certain degree. You'd need some cross-validation to be sure. $\endgroup$
    – Dan Scally
    Commented Jan 17, 2020 at 9:57
  • $\begingroup$ Thanks for the information bud :) $\endgroup$
    – Blackdeath
    Commented Jan 18, 2020 at 17:11


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Browse other questions tagged or ask your own question.