Part 3. Special weapons and tactics

 

I call the techniques in this part “special” weapons and tactics because not all companies need to use them. In my opinion, however, all companies fighting churn need a great set of customer metrics. To someone who trained as a data scientist, this may be a surprise because the subjects in this part include what most people think of as the heart of data science: prediction! But I explained back in chapter 1 that churn is different: predicting churn has only a few use cases, whereas there are many more use cases for great customer metrics. Nevertheless, prediction can be an important weapon in your arsenal, with a few wrinkles unique to churn.

If you have never worked on any predictive analytics before, you might find that chapters 8 and 9 have a steep learning curve. That said, these chapters do cover all the basics, and I think anyone who learned the techniques in parts 1 and 2 can master the part 3 techniques as well. But if you have no experience in predictive analytics, you may need to put in a little extra time and use some of the recommended online resources.

Chapter 8 teaches you how to forecast churn probability with logistic regression. With this technique, you can see the combined influence of all factors that affect churn and rank them in importance. Regression also gives you a forecast that you can use to calculate customer lifetime value.