Guides · Cohort Retention
Guide

Cohort retention analysis: the clearest view of whether customers stick

Blended churn can look stable while your business quietly changes underneath it. Cohort analysis is the fix — it holds time constant so you can finally see whether the customers you win today stick better than the ones you won last year.

Get your number in seconds

Skip the math — the free cohort retention calculator runs this from your own figures.

Open the free calculator →

What a cohort is

A cohort is a group of customers who share a starting point — usually the month or quarter they first signed up or purchased. Cohort retention analysis tracks what share of each group remains active as time passes: how the January cohort looks after one, three, and six months, and so on. Because every customer in a cohort started together, their retention curves are directly comparable across groups.

That shared starting point is what gives the analysis its power. It lets you line up the six-month retention of every cohort side by side and ask a precise question: are the customers we're acquiring now retaining better or worse than the ones we acquired before? A single blended figure can't answer that, because it mixes together cohorts of every age.

Why a blended churn number misleads

A company-wide churn or retention rate blends customers who joined last week with those who joined two years ago, and it moves for reasons that have nothing to do with product quality. A surge of new sign-ups can temporarily flatter retention, because brand-new customers haven't had time to churn yet; a change in customer mix can shift it either way. The headline number can hold steady while the underlying retention of each cohort is quietly deteriorating.

Cohorts remove that distortion by isolating each group. If your product genuinely got stickier, newer cohorts will show higher retention at the same age than older ones — a signal a blended number would bury. Conversely, if newer cohorts retain worse, cohort analysis catches it early, before the aggregate number reacts and long before it becomes obvious in revenue.

Reading a retention curve

Plotted over time, a cohort's retention almost always declines — the question is how fast and where it settles. A curve that drops steeply in the first month or two and then flattens tells a very different story from one that erodes steadily forever. The flattening point matters most: a retention curve that stabilises means you've found a group of customers for whom the product genuinely sticks, and that plateau is what compounds into durable revenue.

Comparing curves across cohorts shows whether that plateau is rising. If each new cohort flattens at a higher level, your improvements to onboarding, product, or targeting are working. Revenue retention adds another layer — a cohort can lose customers while the survivors spend more, so tracking both customer and revenue retention gives the complete picture the tool is built around.

Putting cohort analysis to work

The practical payoff is a feedback loop. Make a change — a better onboarding flow, a pricing tweak, a sharper target customer — and the cohorts that start after it will reveal, over the following months, whether it improved retention. Because cohorts hold age constant, you can attribute the difference to the change rather than to shifting mix or seasonality, which makes cohort analysis the standard way product and growth teams judge what's working.

The free calculator gives you customer and revenue retention for a single cohort from its starting size, active count, and revenue per customer, and the full tracker holds multiple cohorts so you can compare curves over time. It's assumptions-driven and not financial advice, but it turns retention from a fuzzy aggregate into a clear, comparable signal you can act on.

Questions

What's the difference between customer and revenue retention in a cohort?

Customer retention counts how many members of the cohort are still active; revenue retention measures how much of the cohort's revenue remains, including expansion from the survivors. They can diverge sharply: a cohort might lose a third of its customers while the remaining two-thirds upgrade enough that revenue retention stays flat or even grows. Tracking both matters — customer retention tells you whether people stick, revenue retention tells you whether the money sticks, and the gap between them tells you how much expansion is offsetting customer loss.

How many customers do I need for cohort analysis to be useful?

Cohort analysis works best when each cohort is large enough that its retention rate isn't dominated by the behaviour of a handful of customers. With very small cohorts, one or two churns swing the percentage wildly and the curves get noisy. If your cohorts are small, grouping by quarter instead of month, or combining periods, gives steadier signal. Even with modest numbers the directional comparison across cohorts is informative — you just read the trend with appropriate caution rather than treating small differences as meaningful.