# Cohort Analysis > Group users by when they signed up, then watch how many of each group stick around month after month. - Type: Calculator: Tracking signup groups over time - Tags: Metrics, Retention - Growth levers: Retention (primary), also Activation - ~1354 words --- **Cohort retention curve.** Cohort analysis groups users by a shared start point, usually the month they signed up, and tracks what percentage of each group is still active over the following weeks or months. Instead of one blended retention number that mixes everyone together, you get a separate curve per cohort, so you can see whether the people who joined in March stick around better than the ones who joined in January. That is the difference between knowing your retention is 40% and knowing which signups produced it. The tool above does not crunch a spreadsheet of raw events. It projects one cohort's retention curve from three numbers you already have a feel for: how many people joined, how many survive the first month, and how slowly the survivors leak out after that. It is the shape every cohort table converges to, which makes it useful for planning before you have a year of real data, and for sanity-checking the curves you already track. ## How the retention curve is calculated Real cohorts almost always drop hard in month 1 and then flatten. Most of the people who were never going to stick leave fast; the ones who find the value moment settle into a habit and leak out slowly. The tool models exactly that two-stage shape: a steep month 1 fall, then a gentle decline that compounds at a fixed monthly rate. > **Formula:** Month 0 = 100%. Month 1 = your month 1 retention. Every month after that = previous month x terminal monthly retention. So month n (for n of 2 or more) = month1 x terminal^(n - 1). Multiply any month's percentage by the cohort size to get the users still active that month. Worked example, using the tool's default inputs: a cohort of 1,000 users, 60% retained at month 1, and a terminal monthly retention of 95%. Month 1 holds 60% (600 users). Month 2 is 60% x 0.95 = 57%, month 3 is 57% x 0.95, and so on. By **month 6 the curve sits at 46%** (about 464 users) and by **month 12 at 34%** (about 341 users). That is the same math the widget runs, so leave the sliders on their defaults and you will read 46% at month 6 and 34% at month 12. The terminal rate is the lever that decides everything past month 1. At 95% monthly retention, a cohort keeps roughly 0.95 to the 11th of its month 1 base over the next year, which is the slow bleed you see. Push terminal retention to 98% and the curve nearly plateaus; drop it to 90% and the same starting cohort halves again well before month 12. Small changes there compound, which is why one extra point of monthly retention is worth more than it looks. ## How to read the curve and use the tool Watch the shape, not just the endpoint. A retention curve that keeps falling at a steady slope means you have no habit: every month you lose the same fraction, and the cohort eventually goes to zero. A curve that flattens means you found a core of users for whom the product became a default. The month the slope goes flat is your real retention story, because everything to the right of it is the base you actually keep. - **Month 1 retention is an activation problem.** The steep early drop is mostly people who never reached the value moment. If this number is low, fix onboarding before anything else. Pair it with your [activation rate](https://www.productgrowth.blog/calculators/activation-rate) to find where new users fall off. - **Terminal retention is a habit problem.** The flattening slope is your retained core leaking out. Raising it means giving long-term users a reason to keep coming back, which is the same lever behind your [retention rate](https://www.productgrowth.blog/calculators/retention-rate). - **Compare cohorts, do not average them.** Run the tool once per signup month or per channel and line the curves up. If a recent cohort holds higher at month 3 than an older one did, an onboarding or product change is working. A blended number hides that entirely. Two things to keep honest. First, pick what counts as active and stick to it: logged in, performed a key action, or paid. A login-based curve flatters you next to a paid-active one. Second, the projection assumes a constant terminal rate, which real cohorts only roughly follow. Use it to plan and to compare scenarios, then validate against your actual cohort table once you have the months of data to fill it. ## Why cohort analysis matters for growth Cohort analysis is how you tell whether the product is getting stickier or you are just buying more traffic. A single retention number can hold flat while your best cohorts quietly decay, because fresh signups paper over the loss. Split by cohort and the trend shows up: improving curves mean the changes you ship are landing, flat-then-falling curves mean you are filling a leaky bucket. It is also the cleanest way to read the impact of an onboarding rework, a pricing change, or a new activation flow, since you can watch the cohort that experienced the change against the ones that did not. ## Related calculators - [Retention rate](https://www.productgrowth.blog/calculators/retention-rate) for the single-period version of what each cohort curve measures. - [Churn rate](https://www.productgrowth.blog/calculators/churn-rate) for the mirror image: the share of a cohort you lose each period instead of keep. - [Activation rate](https://www.productgrowth.blog/calculators/activation-rate) to attack the steep month 1 drop at its source. - [Engagement rate](https://www.productgrowth.blog/calculators/engagement-rate) to check whether the users who stay are actually using the product. #### What is a good cohort analysis? A good cohort analysis is one where the retention curves flatten instead of falling to zero, and where newer cohorts hold higher than older ones at the same age. The flattening is the signal that matters: it means a core of users turned the product into a habit, so the cohort settles at a stable floor rather than bleeding out month after month. For most B2B SaaS, a curve that plateaus somewhere in the 30 to 50% range by month 6 to 12 is healthy; consumer products plateau lower. The other half of good is direction. If each new signup cohort retains better than the last, your onboarding and product changes are working, which is the whole reason to run cohorts instead of one blended number. #### How is cohort retention calculated? You group users by a shared start point (usually signup month), then for each later period count how many of that original group are still active and divide by the cohort's starting size. Month 0 is 100% by definition. The calculator above projects the curve from three inputs: month 1 retention sets the first big drop, and a terminal monthly retention rate compounds each month after that, so month n equals month 1 retention times the terminal rate raised to the power of (n minus 1). With 60% month 1 retention and a 95% terminal rate, a cohort lands at about 46% by month 6 and 34% by month 12. #### What is the difference between cohort analysis and retention rate? Retention rate is a single number for a single period: the share of customers you kept this month or this year. Cohort analysis breaks that number apart by start group and tracks each group over many periods, so you see a curve instead of a point. The retention rate tells you how you did; the cohort curve tells you which signups produced it and whether the trend is getting better or worse. They answer the same question at different resolutions. #### How many months should a cohort analysis cover? Long enough for the curve to flatten, which for most SaaS is 6 to 12 months. The early months are dominated by the month 1 activation drop; the real retention story is the slope after it settles, and you cannot see that until you have a few flat months to read. The tool above projects 12 months for this reason. If your curve has not flattened by month 12, that is itself the finding: you have no retained core yet, and the fix is upstream in activation and habit, not further down the funnel. --- All posts: https://www.productgrowth.blog/archive ยท Site: https://www.productgrowth.blog