A retention cohort is a group of users or accounts that share a starting event in the same time period, such as signing up or activating in a given week, and whose continued activity is tracked period by period afterward. In product-led growth, retention cohorts are the standard way to see whether the product keeps the people it attracts, independent of how fast new users arrive.
Because each cohort is followed from its own starting point, cohort analysis separates the quality of retention from the volume of acquisition. A rising total user count can hide worsening retention, and a retention cohort table makes that visible within weeks.
How retention cohorts are built
Building a retention cohort requires three decisions. The first is the cohort-defining event, usually signup, first activation or first payment. The second is the retention event, the action that counts as still being retained, such as logging in, completing a core action or paying. The third is the time grain, typically days for consumer apps, weeks for most SaaS products and months for billing analysis.
The result is a cohort table: each row is a cohort, each column is the number of periods since the start, and each cell holds the share of the original cohort that performed the retention event in that period. Plotting each row as a line gives the retention curve.
| Signup week | Cohort size | Week 1 | Week 4 | Week 8 |
|---|---|---|---|---|
| Week of March 2 | 1,200 | 48% | 31% | 26% |
| Week of March 9 | 1,350 | 50% | 33% | 28% |
| Week of March 16 | 1,100 | 55% | 38% | 33% |
| Week of March 23 | 1,400 | 56% | 39% | Not yet reached |
In this illustration later cohorts retain better at every age, which suggests that a change in onboarding or activation around mid-March improved the product experience. Reading down a column compares cohorts at the same age, which is the most informative comparison.
Types of retention measurement
The same cohort can be measured in several ways, and the choice changes the numbers considerably. Teams should state the definition explicitly whenever they share a retention figure, because different definitions are not comparable.
| Definition | What counts as retained | Typical use |
|---|---|---|
| N-day or N-week retention | Active exactly in period N | Consumer apps, habit-driven products |
| Unbounded retention | Active in period N or any later period | Infrequent-use products |
| Bracket retention | Active within a range of periods | Products with irregular usage |
| Logo retention | Account still subscribed | B2B SaaS revenue reporting |
| Net revenue retention | Revenue kept including expansion | Investor reporting and pricing analysis |
Behavioral retention, based on product usage, tends to lead revenue retention by weeks or months. A drop in weekly active use within a cohort usually appears well before the same accounts cancel, which gives teams time to react.
Reading the retention curve
A typical retention curve falls steeply in the first periods and then flattens. The flat part, often called the retention plateau, represents users who found lasting value. A curve that keeps declining toward zero indicates that the product has not yet reached product-market fit for that segment.
| Curve shape | Likely meaning | Typical response |
|---|---|---|
| Flattening at a stable level | A core group finds lasting value | Grow acquisition in similar segments |
| Steady decline to zero | No durable use case yet | Revisit value proposition and activation |
| Steep early drop, then flat | Onboarding loses many users | Shorten time to value |
| Smile curve that rises later | Users return through expansion or network effects | Invest in collaboration and reactivation |
| Sudden drop at one age | Trial end or billing event | Review trial length and conversion flow |
Benchmarks vary widely by category, price and audience, so comparing a product with its own past cohorts is more reliable than comparing it with published averages. The most useful signal is the direction of change between cohorts.
Retention cohorts in product-led growth
In a product-led growth model the product itself acquires, converts and expands customers, so retention is the foundation of every other metric. Activation rate, free-to-paid conversion and viral loops only compound if users stay. Retention cohorts are therefore used to evaluate onboarding changes, product tours, reverse trials and pricing experiments.
Segmenting cohorts makes them far more actionable. Common splits include acquisition channel, plan type, company size and whether users reached the aha moment. Comparing users who completed a key action in their first week with those who did not often reveals the behavior that best predicts long-term retention, which can then become the target of activation work or a candidate for the north star metric.
Pitfalls in cohort analysis
The most common mistake is choosing a retention event that is too weak, such as any login or email open, which inflates retention without reflecting real value. A second is mixing cohort sizes without weighting, which lets small cohorts distort averages.
- Survivorship bias: analyzing only accounts that are still active hides why others left.
- Seasonality: holiday or fiscal-year cohorts behave differently and should not be compared naively.
- Immature cohorts: recent cohorts have not reached later periods, so averages across ages can mislead.
- Changing definitions: altering the retention event mid-analysis breaks comparability.
- Ignoring reactivation: users who return after a gap may be missed by strict N-period definitions.
A good discipline is to document the cohort definition, retention event and time grain next to every chart, and to confirm results with qualitative research, since cohorts show what happened but not why.
Put simply
A retention cohort is a group of people who started using a product at the same time, followed over the weeks or months that come after. You check how many of them are still around after one week, one month and so on.
Think of a school class that starts together. Counting how many students are still enrolled each year tells you much more about the school than counting the total number of students, because the total also rises when many new students arrive. Cohorts show whether people stay, not just whether more people come in.
In practice
A team running a freemium design collaboration tool notices that weekly signups are growing, but revenue growth is flat. The growth analyst builds weekly retention cohorts using creating or editing a file as the retention event.
The table shows that only about 20% of each cohort is still active in week four, and that users who invited a teammate in their first three days retain at roughly twice the rate of solo users. The team redesigns onboarding to prompt an invitation during the first session and adds a short product tour.
Cohorts that sign up after the change show a noticeably higher week-four retention, and the team adopts teammate invitations in the first week as a leading indicator for its product-led growth plan.