You’ve learned a lot about understanding churn with metrics derived from events and subscriptions. You’ve seen that simple behavioral measurements can be powerful for segmenting customers who may be at risk for churn and who have different levels of engagement. But you’ve also seen some of the limitations of simple behavioral metrics.
Many simple metrics are correlated, and correlations arise because customers who have a lot of product-related events tend to have a lot of other events as well. Correlations make it harder to tell which types of behaviors are most important. The problem is deeper than a lack of refinement. In this chapter, you’ll learn that correlation between metrics can make you misread the influence of a behavior. A behavior that’s negative (in the sense that it takes utility and enjoyment away from customers) can appear to enhance engagement when it’s correlated with other behaviors that provide utility and enjoyment.