# Lessons Learned

The previous chapter introduced survival analysis and the calculation of hazard and survival probabilities using SQL and Excel. This chapter extends these ideas, showing ways to calculate survival in other situations and to measure the effects of covariates on survival.

The chapter starts by showing how to understand the effects on survival of variables known at the beginning of the customer relationship. The effects might change over time, even though the variables remain constant during each customer’s lifetime. Hazard ratios capture the effects for categorical variables by taking the ratio of the hazards. For numeric variables, the right measure is the average of a numeric variable at different points in the survival curve for active and stopped customers.

One of the biggest challenges in using survival analysis is calculating unbiased hazard probabilities. This is particularly challenging when the data is left truncated—that is, when customers who stopped...