8 Forecasting churn

 

This chapter covers

  • Predicting the probability of customer churn with logistic regression
  • Understanding the relative influence of different behaviors on churn
  • Checking the calibration of your forecasts
  • Using churn forecasts to estimate customer lifetime and lifetime value

At this point, you know all the steps necessary to analyze churn and to design great customer metrics. Those metrics will allow businesspeople to make targeted interventions that should reduce the churn on their product. And those things are the most important for most products, so that’s why the techniques beginning in this chapter (part 3 of the book) can be considered to be special or extra tactics: you can use them if you need to, but they are not always necessary. The most important thing in fighting churn is that the business should make data-driven decisions when segmenting customers and making targeted interventions.

8.1 Forecasting churn with a model

8.1.1 Probability forecasts with a model

8.1.2 Engagement and retention probability

8.1.3 Engagement and customer behavior

8.1.4 An offset matches observed churn rates to the S curve

8.1.5 The logistic regression probability calculation

8.2 Reviewing data preparation

8.3 Fitting a churn model

8.3.1 Results of logistic regression

8.3.2 Logistic regression code

8.3.3 Explaining logistic regression results

8.3.4 Logistic regression case study

8.3.5 Calibration and historical churn probabilities

8.4 Forecasting churn probabilities

8.4.1 Preparing the current customer dataset for forecasting

8.4.2 Preparing the current customer data for segmenting

8.4.3 Forecasting with a saved model