It's not just about survival: using survival analysis to study customer behaviour

Ardo Illaste

Prior knowledge:
No previous knowledge expected


Survival analysis is relevant beyond its widespread use in clinical studies. Analyzing time to event data with survival analysis methods provides unique insights into product usage and customer behaviour. This talk will briefly cover some of the applications of survival analysis in a SaaS setting such as customer churn and activation as well as how to get started with this kind of analysis easily.


  • Looking beyond medical sciences and why it's more appropriate to call survival analysis time to event analysis.
  • Censor this! What is censoring and why naive methods don't work well for analyzing continuously accumulating behaviour data.
  • Useful tools: lifelines and PyMC3 for Python and survival for R
  • Examples using lifelines:
    • How to properly determine customer churn and product activation
    • Survival regression: estimating effects of covariates on customer behaviour
    • Testing to see if visual differences are statistically significant