Reading retention curves and cohorts in a product case
Analytics case · 7 min read
How to define a retention event, read the shape of a curve, and compare signup cohorts so you can explain why retention moved.
Define it before you measure it
Retention is the share of a cohort that comes back and does a meaningful action within a time window. Pick the action (not just any event) and the window that matches how often the product is naturally used: daily for messaging, weekly for planning tools.
Read the shape
A curve that flattens means you have a core of retained users. A curve that keeps falling to zero means no habit has formed. A curve that rises later (a smile) usually means resurrected users, which is worth calling out.
Compare cohorts, not averages
A blended retention number mixes old and new users. Split by signup week and by acquisition channel; a drop often traces to one new channel bringing lower-intent users rather than a product problem.
Common mistakes
- Using any event as the return action, which inflates retention.
- Comparing a young cohort's week 1 to an old cohort's week 8.
- Blaming the product before checking a mix shift in acquisition.
Further reading
Try it now, free
Smiling retention curve
No account needed. You get graded feedback right away.
Answer the question