Cohort analysis for PMs: How I’d read retention the useful way
Our reviewers evaluate writing services independently. Learn how we stay transparent, our methodology, and tell us about anything we missed.
If you only look at blended retention, you can feel great while new cohorts quietly die. Cohort analysis is how I separate "users from March behave like this" from "the product overall looks fine."
Cohort, simply
A cohort is a group sharing a starting characteristic—signup week, acquisition campaign, plan type, first feature used.
A cohort analysis watches behavior over time for each group, so you can compare apples to apples.
Graduation-year analogy: class of 2010 vs 2011. Same school, different starting line.
The retention table mental model
Classic B2C retention grid:
- Rows = cohort (e.g., users who signed up in week W)
- Columns = periods since start (week 0, 1, 2…)
- Cells = % still active (by your definition of active)
Read across a row to see how a cohort decays. Read down a column to compare different cohorts at the same age.
That second read is where product changes show up: did April signups retain better at week 4 than January signups?
How I'd define cohorts usefully
- Acquisition time — default for retention health
- Acquisition channel — paid vs organic vs referral
- Activation quality — completed onboarding vs not
- Persona / segment — when the product serves distinct jobs
Don't cohort on everything. Start with signup-week retention, then slice the questions you actually have.
Mistakes I try not to make
- Changing the definition of "active" mid-comparison
- Celebrating early spikes from a one-off campaign cohort
- Ignoring seasonality
- Treating B2B like B2C (seats, contracts, and multi-user accounts change the story—B2B cohorts deserve their own care)
- Tiny cohorts with huge variance presented as truth
What I'd do with the insight
- Spot which recent cohorts underperform
- Hypothesize why (onboarding change? acquisition quality? product regression?)
- Validate with qualitative + funnel diagnostics
- Ship a fix aimed at new cohort health
- Recheck the same aged column later
My take
Cohorts turn retention from a vibe into a comparison. If you can't explain how this month's new users retain versus last quarter's, you're flying on blended averages.
Want tighter analytics instincts as a PM? The Product Manager Certification helps—and the newsletter keeps the field notes coming.
