What Is Cohort Analysis?
Cohort analysis groups users by a shared starting point, usually the week or month they signed up, then tracks how each group behaves over time. It separates how new users behave from how older ones do, so you can see whether retention is genuinely improving rather than being masked by overall growth.
A single retention number hides the most important question in growth: are the users you acquire this month sticking around better than the ones you acquired last month? Cohort analysis is how you answer it.
Why a single retention number lies
A blended retention figure mixes everyone together: the users who joined this week and the ones who joined a year ago. When you are growing, fresh signups who have had no time to churn pile into that average and make it look healthy, even as each individual group decays. The reverse is just as dangerous: a real improvement in new signups gets buried under the mass of older users.
Cohort analysis fixes this by slicing users into groups that share a start date and following each group on its own. Instead of one number that blends three forces, you see one clean decay curve per cohort.
A retention cohort table
Rows are signup cohorts, columns are months since signup. The figures below are illustrative, chosen to show how the table reveals a trend an average would hide.
| Signup cohort | Month 0 | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| January | 100% | 48% | 39% | 35% |
| February | 100% | 52% | 44% | n/a |
| March | 100% | 55% | n/a | n/a |
| April | 100% | n/a | n/a | n/a |
Read down a column to compare cohorts at the same age. Month 1 retention climbing from 48 to 55 percent across the January, February, and March cohorts is the signal that an onboarding change is landing. A blended average would have smeared it away.
How to run a cohort analysis
1. Define the cohort
Most often this is signup date, but it can be first purchase, acquisition channel, or plan tier. The definition decides which question the analysis answers.
2. Pick the behavior to track
Choose what counts as retained: active in the product, returning to a core action, or still paying. Revenue cohorts and active-usage cohorts can tell different stories.
3. Match the time grain to your usage cycle
Daily for a habit product, weekly or monthly for a considered purchase. The wrong grain either hides the decay or buries it in noise.
4. Read both directions
Down a column compares cohorts at the same age, which exposes trends. Across a row follows one cohort aging, which shows the shape of decay.
Retention curves vs cohort tables
Plot one cohort's retention over time and you get a retention curve. The shape is what matters: a curve that keeps sliding toward zero means users leave and never settle, while a curve that flattens into a plateau means a core group keeps coming back. That flattening is the visual most teams hunt for, because a curve that levels off is a sign the product has found a durable base of users. A cohort table is simply many of these curves stacked so you can compare them, which is how you tell whether the plateau is rising cohort over cohort.
Cohorts and experiments
Cohort analysis and controlled tests answer different questions. After you ship a change, comparing cohorts from before and after is a quick read on whether retention moved. The catch is that it is observational: seasonality, a shift in acquisition channels, or a pricing change can all move cohorts at the same time, so a before-and-after jump is a signal, not proof. Use cohorts to monitor the trend and to generate hypotheses, then use a randomized A/B test when you need to attribute a result to a specific cause.
Frequently asked questions
What is the difference between cohort analysis and a funnel?
A funnel tracks how users move through a sequence of steps, such as signup to activation to purchase. Cohort analysis tracks groups of users defined by a shared start date over time. Funnels show where users drop within one journey; cohorts show whether behavior is improving across groups acquired at different times.
What is a retention cohort?
A retention cohort is a group of users who share a starting period, usually the week or month they signed up, measured by how many remain active over time. Following each cohort separately reveals the decay curve that a blended average hides.
How do you read a cohort table?
Read down a column to compare different cohorts at the same age, which surfaces trends such as rising or falling early retention. Read across a row to follow one cohort as it ages, which shows the shape of its decay. The two directions answer different questions.
What is the difference between cohort analysis and an A/B test?
Cohort analysis is observational: it watches groups defined by when they joined, so outside factors like seasonality can confound a before-and-after comparison. An A/B test is a randomized controlled experiment that isolates a single change as the cause. Use cohorts to monitor and form hypotheses, and an A/B test to attribute a result.
Related terms
Go deeper
- North Star Metric: The Ultimate Guide
- Subscription Experiments Guide
- Growth Loops Guide
- Growth Experimentation: The 2026 Ultimate Guide
About GrowthLab
GrowthLab is an experiment management tool where AI drafts the hypotheses, ICE and ROTI prioritize them, and every learning compounds into the next batch.
