How to Read a Cohort Retention Curve (Without a PhD in Statistics)

Most SaaS founders track retention by computing a single percentage: "we retain 60% of users month over month." This number feels useful. It is almost useless.

A flat 60% retention can hide two completely different stories. In one, you have a stable customer base that mostly stays, and your few losses are old contracts winding down. In the other, you bleed two-thirds of every new cohort within their first 30 days, and the survivors stick forever. Same headline number. Wildly different businesses.

The only way to tell them apart is cohort analysis. This post will show you how to read the chart, the three shapes that signal trouble, and what to do about each.

What a cohort actually is

A cohort is a group of users defined by a shared starting point. For SaaS, the starting point is usually signup month.

If 100 users signed up in March 2026, those 100 users are the "March 2026 cohort." From that moment, you track them as a group — separate from April signups, separate from May signups, separate from everyone who was already there.

Cohort retention answers a specific question: of the 100 users who joined in March, how many are still active each month after?

The data looks like this:

| Month | Active Users | Retention | |-------|--------------|-----------| | Month 0 (signup) | 100 | 100% | | Month 1 | 60 | 60% | | Month 2 | 45 | 45% | | Month 3 | 38 | 38% | | Month 4 | 35 | 35% | | Month 5 | 34 | 34% | | Month 6 | 34 | 34% |

Notice what happens between month 5 and 6: the number stops dropping. That is the flattening. It is the most important part of the chart.

Why aggregate retention lies

Imagine you compute retention this way:

"Out of all active users this month, how many were active last month?"

That number sounds like retention, but it averages every cohort together. New cohorts (which drop steeply in their first month) and old cohorts (which have already flattened) get blended into one figure. You can have terrible new-user retention and great long-tail retention — and never notice, because the average looks fine.

Cohort analysis separates the two. Each row is its own group. Each group tells its own story.

The healthy curve

A healthy cohort curve has three properties:

1. Sharp initial drop. Most signups will leave in the first 30 days. This is normal. Wrong-fit users, accidental signups, and tire-kickers self-select out.
2. A flattening tail. The drop slows, then plateaus. The plateau is your retention floor — the users who genuinely value your product.
3. A floor above zero. That floor is your true retention rate. For B2B SaaS, healthy floors land between 30% and 60% after six months. Consumer products run lower.

When you stack multiple cohorts on the same chart and they all flatten at similar heights, you have product-market fit. When they all flatten higher than previous cohorts, you have improving product-market fit. That is the chart you want to see.

For a refresher on the underlying retention math, start here. For broader unit economics, our LTV guide walks through the formulas.

Three failure shapes — and what to do about each

Shape 1: The cliff

You lose 70% or more of a cohort in the first 30 days, then continue dropping. The chart looks like a cliff with no ledge.

What it means: Users are signing up but never reaching value. Your onboarding does not work, your activation moment is unclear, or you are attracting the wrong people in the first place.

What to do:

• Track Time to First Value as a leading indicator. Cut it in half.
• Audit your onboarding flow end-to-end. Most users decide whether to come back within their first session. If they spend that session looking for a button, you have already lost them.
• Look at where they came from. If paid ads bring users who churn immediately, the channel is misaligned with the product. Adapty's playbook on aligning ad creatives with the actual app experience applies here too.
• Test your activation gate. The best SaaS products force a meaningful action within the first five minutes — connect a data source, invite a teammate, import existing data — anything that makes the product feel real.

Shape 2: The slow leak

The curve declines slowly and steadily and never flattens. Month-over-month retention looks fine — maybe 90% each month — but the cumulative effect compounds. After 12 months, you have 28% of the original cohort left, and that number keeps falling.

What it means: Your product delivers value for a while, but users eventually outgrow it, switch to a competitor, or solve the underlying problem and leave. There is no retention floor.

What to do:

• This is usually a product problem, not a marketing problem. The product needs a reason to stay relevant past the initial use case.
• Look at long-tenured users separately. What are they doing differently? Often they have adopted a feature or workflow the average user does not know about. Promote it.
• Consider expansion paths. If your product is "done" after three months, you cannot rely on retention alone — you need expansion revenue (seats, usage tiers, add-on modules) to compound growth without depending on a flat retention floor.
• Build switching costs intentionally: data accumulation, integrations, team collaboration. Stickiness is rarely accidental.

Shape 3: The bathtub

The curve drops, flattens, and then rises again. This sounds great. Often it is a measurement artifact.

What it means: You are counting reactivated users as part of the original cohort. They left, came back, and now your chart shows growth. In reality, you have lost most of the original cohort and replaced them with returning users.

What to do:

• Decide upfront whether reactivations belong in the cohort. There is no universal right answer. For pure retention analysis, they do not — exclude them and track reactivation as a separate metric. For total revenue impact, they do.
• Either way, define it clearly and apply it consistently across cohorts. The most expensive mistake is mixing definitions and not noticing.

Customer-count cohorts vs revenue cohorts

There are two ways to count cohort retention, and both matter.

Customer-count cohort: How many of the original users are still active? This is the chart we have been looking at. It treats every user equally.

Revenue cohort: How much of the original cohort's MRR is still being paid? This treats every user differently — a $500/month customer counts ten times more than a $50/month customer.

The two charts often tell different stories. You might retain 50% of customers but 80% of revenue, because your biggest customers stay. Or you might retain 90% of customers but 60% of revenue, because the small accounts that don't churn don't account for much.

For unit economics decisions, prioritize the revenue cohort. For product decisions, prioritize the customer-count cohort.

A revenue cohort that grows above 100% over time is the dream: it means expansion (upsells, seat additions, usage growth) is outpacing churn within the cohort. This is the foundation of Net Revenue Retention above 100%, and it is how the best SaaS companies grow without depending on new logos.

How to actually build the chart

You do not need fancy tools to start. Here is the minimum viable workflow:

1. Export a list of every user with their signup date. Most product databases can do this in one query.
2. Group by signup month. That gives you your cohorts.
3. For each cohort and each subsequent month, count how many users were active. Define "active" specifically — logged in, performed a key action, paid, whatever fits your product.
4. Express each cell as a percentage of the original cohort size. That is your retention rate for that cohort at that month.

The output is a triangle: cohort months down the left side, calendar months across the top. Each cell is the retention percentage. Add color shading (lighter for low retention, darker for high) and you have a heatmap. The heatmap is the standard format because it lets you spot patterns in seconds.

If you are doing this in a spreadsheet, you can build it in about two hours. If you are doing it in your application database, an hour with the right SQL. Use Saasly's free calculator for the underlying LTV, churn, and NRR math that feed your cohort interpretation.

When to start tracking cohorts

The temptation is to wait until you have "real" data. Do not wait.

Three rules of thumb:

1. Start tracking once you have three monthly cohorts of at least 30 users each. Before that, the swings are too noisy to read.
2. Stop trusting the chart until you have six cohorts. Patterns in three or four cohorts can be coincidence. Six cohorts is enough to see a trend.
3. Make decisions on twelve cohorts. A year of monthly data is when cohort retention becomes a serious planning input.

Do not let the wait-for-perfect-data instinct stop you from setting up the system. The system is the cheap part. Twelve cohorts of data will show up on its own.

What cohort retention does not tell you

Cohort retention is powerful but partial. Three things it cannot answer:

• Why users leave. Cohort charts show you where the leak is, not what is causing it. You need exit interviews, support transcripts, or qualitative research for the why.
• Which users are most at risk. Aggregate cohorts blend everyone together. Per-segment cohorts (by plan, channel, persona) are how you find specific failure modes — see the Growth Machine framework for how to slice user states productively.
• The financial impact of a retention change. Cohort retention is a percentage. To translate it into dollars, you need LTV math and a clear view of ARPU per cohort.

Use cohort retention to find the problem. Use other tools to diagnose and price it.

What to do this week

If you have never built a cohort chart for your product:

1. Export signup dates and activity timestamps from your database. One spreadsheet, two columns each.
2. Build the cohort triangle in a spreadsheet. Group by signup month. Count active users by month-since-signup.
3. Plot two or three cohorts on the same line chart. Look for the flattening point.
4. Compare last month's cohort to a cohort from six months ago. If the new cohort is flattening higher, your product is improving. If lower, something has regressed.

If you have already built one and it has been gathering dust:

1. Pull up your most recent six cohorts.
2. Identify which one of the three failure shapes you have, if any.
3. Pick the one corresponding action from the list above. Run it for 30 days. Re-check the new cohort.

Quick reference

| Term | What it means | |------|---------------| | Cohort | A group of users defined by shared signup window | | Customer-count cohort | Tracks how many original users remain active | | Revenue cohort | Tracks how much of original cohort MRR is still paid | | Flattening | The point where retention stops dropping — your floor | | Retention floor | The percentage of a cohort that stays long-term | | Healthy floor (B2B) | 30-60% after 6 months for most B2B SaaS |

Track yours alongside the rest

Cohort analysis is one piece of a healthy metrics practice. The others — LTV, CAC, churn, NRR — work together. Use Saasly's free calculator for the underlying unit-economics math, and the benchmark library to see where your numbers sit against B2B / B2C / Mobile / Fintech / E-commerce SaaS peers at your stage. No signup needed for either.

---

Frequently asked questions

What is a cohort in SaaS?

A cohort is a group of users who signed up in the same time window — typically a week or a month. Cohort retention tracks how that specific group behaves over time, separate from new signups. It answers questions like "are the users who signed up in March still here in May?"

How do I calculate cohort retention?

Pick a cohort (e.g., everyone who signed up in March). For each month after, count how many of those original users are still active. Divide by the original cohort size. Month 1 retention for a March cohort of 100 users with 70 still active in April is 70%.

What is a healthy cohort retention curve?

A healthy curve drops sharply in the first month (this is expected — many signups never come back) then flattens. The flattening is the important part: it means you have a stable core of users who keep returning. A curve that never flattens means you have no retention floor.

When should I start tracking cohorts?

As soon as you have three or more monthly cohorts of at least 30 users each. Earlier than that, the numbers swing wildly with each individual user. By month four or five, the patterns become readable. Do not wait for perfect data — start tracking and let the picture sharpen over time.