What Is a Cohort in Product Analytics? Definition + Example
A cohort in product analytics is a group of users who share a common starting characteristic within the same time window — most often signing up in the same week or month. Grouping users this way lets you watch how each group behaves over time, instead of blending everyone into a single average that hides what is actually happening.
Quick Answer: A cohort is a set of users bucketed by a shared starting point — usually their signup date — so you can compare each group's behavior over its own timeline rather than averaging all users together.
How a cohort works: the shared-start-event rule
A cohort is defined by one rule: every member shares the same starting event, anchored to the same time window. That anchor is what separates a cohort from a random pile of users. Pick the event, pick the window, and everyone who qualifies is "in."
The most common anchor is acquisition — the week or month a user first signed up. All users who joined in March form the "March cohort," and you follow that fixed group forward through April, May, and beyond. Nobody joins the March cohort later; the membership is locked the moment the window closes.
Why the lock matters. Because the group never changes, any change you observe over time comes from the users' behavior, not from new people being added. That is the whole point. A blended metric like "total active users this month" mixes veterans and brand-new signups, so a flood of newcomers can make a leaky product look healthy. Cohorts strip that illusion out by holding the population still.
Eric Ries makes this the backbone of measurement in The Lean Startup, arguing that founders should favor cohort-based accounting over vanity totals — track how each successive group of users performs rather than watching cumulative numbers that only ever go up.
Time-based cohorts vs. behavior-based cohorts
Cohorts come in two main flavors: acquisition (time-based) and behavioral (action-based). Both follow the shared-start rule; they differ in which starting event they anchor to.
Here is how the two compare across the dimensions that matter when you are deciding which to build.
| Dimension | Acquisition cohort | Behavioral cohort |
|---|---|---|
| Grouping rule | Joined in the same time window | Performed the same action |
| Typical anchor | Signup week or month | First purchase, invited a teammate, hit a key feature |
| Best for | Retention curves, lifecycle trends | Isolating what drives value |
| Question it answers | "Do newer signups stick better than older ones?" | "Do users who do X retain better than those who don't?" |
| Time window | Fixed calendar period | Fixed event, timed from when it occurred |
Takeaway: reach for an acquisition cohort when you want to see whether the product is getting stickier over time, and a behavioral cohort when you want to test whether a specific action predicts long-term retention. Behavioral cohorts are how you find your product's "aha moment" — the early action that separates users who stay from users who leave. If you are still framing which behavior even matters, our guide to startup idea validation covers how to pick the signal before you instrument it.
Worked example: a signup-week cohort grid
A cohort grid stacks each signup group as a row and reads retention across the columns. The example below is illustrative only — the numbers are made up to show the shape, not measured from any real product.
Imagine you group users by the week they signed up, then measure the percentage still active in each following week:
| Signup week (hypothetical) | Week 0 | Week 1 | Week 2 | Week 3 |
|---|---|---|---|---|
| Cohort A | 100% | 62% | 48% | 41% |
| Cohort B | 100% | 55% | 39% | 30% |
| Cohort C | 100% | 68% | 55% | 50% |
How to read it. Every cohort starts at 100% in Week 0 — that is just "everyone who joined." Moving rightward shows decay as users drop off. Reading down a column compares cohorts at the same age: at Week 3, the hypothetical Cohort C holds 50% while Cohort B holds 30%.
Takeaway: because these illustrative numbers compare groups of the same age rather than the same calendar date, a rising trend down a column would suggest the product is improving for newer users — a signal a blended average would completely bury. This row-by-time layout is the raw material for cohort analysis, and each row, plotted as a line, becomes a retention curve.
Cohort vs. segment vs. blended average
A cohort is time-anchored, a segment is attribute-anchored, and a blended average collapses both. They are easy to confuse because all three group users — but they answer different questions.
| Concept | Grouped by | Tracks over time? | Main use |
|---|---|---|---|
| Cohort | Shared start event + time window | Yes — that is the point | How behavior changes as a group ages |
| Segment | A shared attribute (plan, region, source) | Not inherently | Slicing users by who they are right now |
| Blended average | Nothing — all users pooled | No — single snapshot | A headline number, easily misleading |
Takeaway: use a cohort when the question involves change over time, and a segment when the question is which kind of user. The two combine well — you can cohort within a segment, such as tracking retention for paid users who signed up in Q1. For the attribute side of that pairing, see what a user segment is and how it differs from a time-based group.
Why cohorts matter for retention and product-market fit
Cohorts matter because they answer the one question a blended metric cannot: is the product getting better or worse for the people who use it? New-user growth can mask the churn of everyone who came before. Hold the population still with a cohort, and the truth surfaces.
In Lean Analytics, Alistair Croll and Benjamin Yoskovitz treat cohort analysis as a core tool precisely because it exposes whether product changes actually move retention. If Cohort C retains better than Cohort A at the same age, something you shipped between them is working.
This is also the clearest read on product-market fit. A flattening retention curve — where a cohort stops losing users and stabilizes — is one of the strongest signals that a real, durable group of people needs what you built. Cohorts turn "are people sticking around?" from a gut feeling into a number you can watch improve. Tools built for early teams lean on this cohort-first framing so founders measure retention instead of vanity totals.
Key Takeaways
- A cohort is users grouped by a shared start event in the same time window — most often the week or month they signed up.
- Two main types exist: acquisition (time-based) cohorts and behavioral (action-based) cohorts, differing only in which starting event they anchor to.
- Cohorts beat blended averages because they hold the population fixed, so new users cannot mask the churn of older ones.
- A cohort grid reads groups by age, not calendar date — comparing down a column reveals whether the product is improving for newer users.
- A cohort differs from a segment: cohorts are anchored to time and tracked forward, segments are anchored to an attribute in the present.
- Cohorts underpin cohort analysis and retention curves — each cohort row plotted over time becomes a retention curve.
- Flattening cohort retention is a leading product-market-fit signal, which is why The Lean Startup and Lean Analytics both build measurement around cohort-based accounting.
Frequently Asked Questions
What is a cohort of users in simple terms?
A cohort of users is a group who all started at the same time — usually the week or month they signed up. Instead of averaging every user together, you follow that fixed group forward to see how many stay active over time, which reveals behavior an overall average would hide.
What is the difference between a cohort and a segment?
A cohort groups users by a shared starting event in a time window and tracks them forward over time. A segment groups users by an attribute they have right now — plan tier, region, or acquisition source — without an inherent time dimension. Cohorts answer "how does behavior change as a group ages"; segments answer "which type of user."
Why are cohorts better than a blended average?
A blended average pools all users into one number, so a surge of new signups can hide the churn of earlier users. A cohort holds the group fixed, so any change over time reflects real behavior rather than population turnover. That is why cohort-based accounting is a core idea in The Lean Startup.
What is the difference between a cohort and cohort analysis?
A cohort is the group of users itself; cohort analysis is the practice of comparing multiple cohorts to spot trends. You build cohorts first, then analyze them — for example, checking whether newer signup groups retain better than older ones, which turns a static grid of numbers into an actionable read on whether the product is improving.