Choose the return behavior carefully
Login activity can indicate access without meaningful value. Use a recurring action that fits the product's natural cycle, or analyze several measures separately. A monthly planning tool should not be judged by daily use as if it were a messaging product.
Distinguish user retention, account retention and revenue retention. They answer different questions and need different records. An account can remain subscribed while fewer people receive value.
Build the cohort view
- Choose an entry event such as first eligible signup or first paid purchase.
- Define the recurring action and interval.
- Exclude internal tests and duplicates consistently.
- Compare cohorts at equivalent ages.
- Segment meaningful changes in audience, product or onboarding.
Leave immature cells unobserved rather than treating them as zero retention. A cohort cannot have a third-month outcome before its third month occurs.
Interpretation table
| Pattern | Question |
|---|---|
| Recent cohorts retain less at the same age | Did acquisition quality or first-use experience change? |
| All cohorts drop at a similar step | Does a recurring value obstacle appear there? |
| A segment retains better | What customer situation or behavior differs? |
| Paid retention holds, usage falls | Is value weakening before cancellation becomes visible? |
Worked example: compare the right columns
Illustrative example: January's cohort has 60% of eligible accounts performing the chosen action in month two. February's cohort has 65% in month two. That is a comparable descriptive observation if definitions and conditions align. Comparing January month three with February month one would not answer the same retention question.
The difference does not by itself prove that an onboarding release caused improvement. Audience mix, seasonality and other changes may also matter.
Investigate composition
A stronger aggregate rate can result from acquiring more customers in a naturally better-retaining segment. Report segment mix alongside retention where it affects the decision. Otherwise the team may credit a product change for an acquisition shift.
Should cancelled accounts disappear from the cohort?
No. Removing unsuccessful customers from the original denominator can make retention look better than it is. Define exceptions carefully and consistently.
Can small cohorts support conclusions?
They can reveal questions, but rates may be unstable. Show counts and avoid presenting a few accounts as a reliable general pattern.
Put this into practice
Create one cohort table with a documented value event and inspect comparable lifecycle periods. Use the largest actionable difference to guide customer interviews or a focused experiment.
Primary-source reading for platform details: Google Analytics: reporting dimensions and metrics.
Related foundation: SaaS onboarding: help users reach their first useful outcome. How these guides are prepared.
Ayoub Mouhachtt
Growth & performance marketing. Explore the portfolio and working background.
Related portfolio work: eGrow. The worked examples in this guide are illustrative and are separate from the portfolio’s project evidence.