Cohort analysis and retention

A cohort is a group of users who share a temporal characteristic: they signed up the same week, made their first purchase the same month, or installed the app in the same period. Grouping by cohort lets you compare how users from different eras behave under similar conditions.

Without cohorts, you mix new users with veterans and metrics lose meaning. A retention rate of "30%" can be excellent if measured at month 6 or disastrous if measured at week 1. Cohorts provide that essential time dimension.

The starting point is choosing the anchor event that defines the cohort: sign-up, first purchase, first session. Then you define the time interval (daily, weekly, monthly) and the return metric: what action must the user repeat to be considered "retained"?

In GA4, you can use cohort explorations for basic analysis. For deeper analysis, export to BigQuery and build cohort tables with SQL. Tools like Amplitude or Mixpanel offer advanced cohort analysis without requiring SQL.

A retention curve shows what percentage of a cohort remains active over time. The shape reveals product health: a sharp initial drop followed by a plateau indicates that users who survive the activation phase stick around. A continuous decline signals a systemic value problem.

The ideal curve flattens quickly at a high percentage. Compare your curves against industry benchmarks: a B2B SaaS can expect 90–95% monthly retention, while a consumer app might consider 20–30% at 30 days as good.

Churn rate is the percentage of users who stop using your product in a given period. It sounds simple, but the calculation is full of pitfalls: do you count from the start or end of the period? Do you exclude new sign-ups? Do you use user churn or revenue churn?

For SaaS, distinguish between gross churn (users lost / total) and net churn (which subtracts expansions). A negative net churn — existing customers generate more revenue than is lost to cancellations — is the hallmark of a thriving business.

Cohort analysis becomes valuable when it generates actions. If the January cohort retains better than the February one, investigate what changed: new onboarding? pricing change? a different acquisition campaign that brought less qualified users?

Segment cohorts by acquisition channel, pricing plan or behaviour in the first few days. Users who complete onboarding within the first 24 hours typically retain 2–3 times better. Identifying these patterns lets you optimise activation and reduce early churn.

GA4 offers basic cohort explorations within its explorations section. For more sophisticated analysis, Amplitude and Mixpanel let you define complex cohorts with multiple conditions and visualise retention with granular segmentation.

If your data lives in BigQuery or your own warehouse, building cohort tables with SQL gives you maximum flexibility. You can cross-reference product data with CRM, payment or support data to understand retention from multiple angles.