Skip to content
Learn · Glossary

Cohort analysis

Cohort analysis is the practice of grouping users by a shared starting point — usually signup week or month — and tracking each group's behavior over time. Because every cohort is measured from its own day zero, it separates real changes in retention from mix effects caused by growth or seasonality.

Updated 8 Jul 20263 min readBy fromHello
Key takeaways
  • A cohort shares a start date; measuring each group from its own day zero keeps retention comparisons honest.
  • Blended retention hides mix effects — a fast-growing product can look healthier than it is because new users dominate the average.
  • A retention curve that flattens above zero means a core of users stays for good — a signal growth practitioners often read as product-market fit.

How does cohort analysis work?

You group users by when they first showed up — the January signups, the February signups — then measure the same behavior for each group at the same relative age: day 1, day 7, day 30. The result is usually a grid where each row is a cohort and each column is an age. Reading down a column answers the question a blended average cannot: do newer cohorts retain better than older ones, or is the product standing still while acquisition grows?

How do you read a retention curve?

Plot one cohort's retention rate against time since signup and you get a curve that starts at 100% and falls. The shape matters more than any single point. A curve that keeps sliding toward zero means every user eventually leaves — growth is refilling a leaking bucket, and your churn rate will catch up with acquisition. A curve that flattens means a core of users has settled in. Growth practitioners commonly read a flattening curve as a product-market-fit signal; it is a heuristic rather than a formal test, but a widely trusted one.

Cohort analysis and its adjacent terms.

Why it matters for a two-person team

Blended metrics flatter growing products: when this month's signups outnumber last year's, the average is dominated by users too new to have left. Cohorts strip that distortion out with one grouping — and the query is simple enough to run in SQL against your own events. Pair the day-30 column with your activation rate to check whether onboarding changes actually stick. What to do when a cohort sags — win-back journeys, lifecycle messaging — is covered in the guide to retention and win-back flows.

What are the common mistakes in cohort analysis?

Three recur. Cohorts too small to be signal — in a 20-user cohort, one person moves retention by five points, so read the trend across several cohorts rather than any single one. Comparing cohorts at different ages — a two-month-old cohort always looks better than a twelve-month-old one at month twelve, because it has no data there yet. And defining 'active' loosely — if opening an email counts as retention, the curve flatters you. Tie 'active' to the action that delivers your product's core value, not to a login.

FAQ

Common questions

  • What is the difference between cohort analysis and segmentation?

    A cohort is fixed by a shared starting point: users who signed up in March stay in the March cohort. A segment is defined by attributes or behavior, and membership changes as users match or stop matching the rules. Cohorts measure how behavior evolves over time; segments decide who gets a message.

  • How many users do you need for cohort analysis?

    There is no hard threshold, but percentages on small groups are noisy: one user in a cohort of 20 shifts retention by five points. Weekly cohorts in the hundreds are readable on their own; below that, switch to monthly cohorts or read the trend across several cohorts.

  • Should cohorts be daily, weekly, or monthly?

    Match the interval to your product's natural usage cycle: daily cohorts for products used every day, weekly for most SaaS, monthly for low-frequency products. Too fine an interval fragments users into tiny, noisy groups; too coarse an interval hides the effect of recent changes.

  • Is a flattening retention curve proof of product-market fit?

    It is a widely used signal, not proof. A curve that flattens above zero shows a durable core of users, which growth practitioners often associate with product-market fit. But a curve can flatten at a level too low to sustain the business, so read it next to cohort size, acquisition cost, and revenue per user.

See the platform the team runs.

Related guides
Early access

Put your growth teamon autopilot.

Early access opens gradually, so the team tunes to real use cases. Small teams with big ambitions go first.

500+ already on the list

Not ready to share an email? It's open source. Run it yourself today. View on GitHub

No spam. One email when your spot opens. Unsubscribe at any time.