Cohort Analysis

Cohort analysis is a method of grouping customers (or users) by a shared starting point and then tracking how each group behaves over time. A cohort is simply a group that shares a defining experience in the same period — most often the month they first signed up or made their first purchase, though it can be any common starting event. Having formed the groups, you follow each one forward across subsequent periods, watching a behavior such as how many of them are still active, still buying, or still subscribed. The result is usually shown as a table or chart where each row is a cohort and each column is a period since they started, so you can read across a cohort's life and down across cohorts to compare them. Cohort analysis turns a flat customer base into a set of groups you can watch age.

The reason cohort analysis matters is that aggregate, whole-base numbers hide what is actually happening to customers over time. A business can show rising total active users while every individual group of new users is churning faster than before — new sign-ups simply mask the decay underneath. Cohort analysis pierces that illusion by following each group's own trajectory, so you can see whether customers who joined recently retain better or worse than earlier ones. That makes it the natural tool for studying retention and the long-run impact of changes: improve onboarding in May, and you can watch whether the May cohort and those after it retain better than the cohorts before. It answers questions about trend and trajectory that a single snapshot of the whole base simply cannot.

Cohort analysis and RFM analysis both segment customers, but along entirely different axes, and using one where the other belongs leads to the wrong conclusion. Cohort analysis groups customers by time — a shared starting point — and follows each group's behavior forward to reveal trends and retention. Its lens is longitudinal: how do groups evolve as they age, and are newer groups doing better or worse than older ones. The questions it answers are about trajectory over time. It does not, by itself, tell you which individual customers are most valuable right now; it tells you how groups of customers behave across their lifetimes and whether the patterns are improving.

RFM analysis, by contrast, scores each customer on the recency, frequency, and monetary value of their purchases to rank them by current value and responsiveness. Its lens is the present state of the base, sorted by behavior, producing segments like best, loyal, at-risk, and lost. Where cohort analysis is a longitudinal view of how groups change over time, RFM is a snapshot that ranks customers by current behavior. They are complements, not rivals: RFM tells you who your most valuable customers are today and how to treat each segment, while cohort analysis tells you how customer groups retain and evolve and whether your efforts are moving those trends. Use RFM to act on the present and cohort analysis to understand the trajectory.

Using cohort analysis well begins with choosing the right cohort definition and the right behavior to track. Group customers by a meaningful starting point — usually the period they joined or first purchased — and follow a behavior that matters, most commonly retention or repeat purchase. Read the table both ways: across a row to see how a single cohort decays or holds over its life, and down a column to compare cohorts at the same age and spot whether newer groups are doing better or worse. Tie the analysis to changes you make, so you can see whether a new onboarding flow or pricing change actually improved the retention of the cohorts exposed to it. The whole value is in comparing trajectories, not admiring a single number.

The failures come from leaning on aggregate metrics that hide cohort-level decay, or from misreading the table. Watching only total active users can mask the fact that each new cohort is churning faster, because fresh sign-ups paper over the loss — exactly the illusion cohort analysis exists to break, so ignoring it is the cardinal error. Other traps include picking a cohort definition that does not fit the question, tracking a vanity behavior instead of a meaningful one, and reading cohorts without controlling for their different ages. And as with RFM, using cohort analysis alone leaves gaps — it shows how groups evolve but not which individuals to prioritize today. The discipline is to define cohorts thoughtfully, track retention over time, compare groups at equal age, connect the trends to your decisions, and pair the longitudinal view with present-state tools like RFM.

Cohort analysis — grouping customers by a shared starting point and tracking each group over time — reveals retention trends that whole-base totals hide, complementing the present-state ranking of RFM analysis.

Etymology: source.

Search interest for this term over the last five years:

View interest-over-time on Google Trends →

Areas of marketing where cohort analysis is a core concern: