Cohort Analysis
Cohort Analysis is a technique that groups customers or users by a shared starting point and tracks how each group behaves over time to reveal patterns aggregate numbers hide.
Also known as: cohort study, cohort tracking, cohort segmentation
Cohort Analysis segments people by a shared starting event, such as the month they signed up, made a first purchase, or entered the funnel, then follows each group's behavior across subsequent periods. The technique reveals patterns that aggregate numbers hide by holding the starting point constant across the comparison, which exposes trends quarters earlier than aggregate reporting can catch them.
What Cohort Analysis Means
Cohort Analysis is a longitudinal segmentation technique. A cohort is a group of customers or users who share a defining starting event, most often the month they were acquired but sometimes the campaign that brought them in or the product version they signed up under. Every member is tracked from that common origin point forward. The analysis applies to retention, churn, cumulative revenue, expansion, and engagement frequency; any metric that evolves over a customer's lifetime can be tracked by cohort to reveal how behavior shifts as cohorts mature.
How Cohort Analysis Works
The technique tracks a metric for each cohort over time and compares cohorts side by side. A retention curve plots the percentage still active over time, showing whether retention is improving, deteriorating, or stable. A layered cohort chart shows each cohort's revenue contribution stacked on older ones, revealing whether topline growth comes from genuine cohort expansion or simply from new cohorts replacing dying old ones. Most subscription analytics tools generate these views natively; teams without specialized tools can compute them from billing or CRM exports using pivot tables and time-indexed grouping.
Common Pitfalls and Misconceptions
The key insight is that blended averages can disguise serious deterioration. A flat overall retention number may hide a steady decline in newer cohorts, masked by loyal older customers who have stayed long enough to anchor the average. The first pitfall is reading the blended average and missing the cohort trend underneath. The second is using too few cohorts: single-cohort analysis is anecdotal, and at least 12 monthly cohorts is the standard minimum to distinguish noise from trend. The third is cutting off at 12 months when the most important expansion or churn patterns develop in months 18 to 36.
Cohort Analysis in Practice
The practitioner application is Cohort Analysis as a feedback loop on acquisition decisions. Tag each cohort by the channel, campaign, or pricing tier that produced it, then compare retention curves across tags. The cohorts that retain best should get more budget; the cohorts that retain worst should be defunded even if they look cheap on a CAC basis. Most growth teams discover that 20 to 40 percent of their acquisition spend produces cohorts with sub-economic LTV, but they only see it once acquisition data is cut by cohort. The cleanest implementations refresh cohort views monthly so the data turns into decisions, not retrospective analysis.
Frequently asked questions
-
What is a cohort?
A cohort is a group of customers or users who share a defining starting event, most often the month they were acquired but sometimes the campaign that brought them in or the product version they signed up under. Every member of a cohort is tracked from that common origin point forward.
-
Why is cohort analysis better than overall averages?
Averages blend everyone together and can hide divergent trends. Cohort analysis isolates each group, so you can tell whether recent acquisition is improving or deteriorating, which an aggregate number conceals. The best cohort and the worst can both be visible at once, and the trend across cohorts becomes obvious.
-
What metrics work well in cohort analysis?
Retention, churn, cumulative revenue, expansion, and engagement frequency are all common. Any metric that evolves over a customer's lifetime can be tracked by cohort to reveal how behavior shifts. The richest cohort views combine a retention metric and a revenue metric so you see both who stays and how much they spend.
-
How does cohort analysis help measure marketing?
By comparing cohorts acquired through different channels or campaigns, you can see which sources produce customers who retain and expand best, not just which sources produce the most leads. This is what turns CAC from a snapshot into an LTV-adjusted measure and reveals which acquisition spend is net positive.
-
What is a retention curve in cohort analysis?
A retention curve plots the percentage of a cohort still active over time. Comparing curves across cohorts shows whether retention is improving, deteriorating, or stable, and where in the lifecycle customers tend to drop off. The shape of the curve often matters more than any single point on it.
-
How many cohorts do you need for the analysis to be meaningful?
At least 12 monthly cohorts is the standard minimum, which gives a year of comparable data and lets you see seasonal patterns. Fewer cohorts make it hard to distinguish noise from trend. For high-volume products, weekly cohorts can speed up signal detection, but monthly is the most common granularity.
-
What is a layered cohort chart?
A layered or stacked cohort chart shows each cohort's revenue or user contribution to the total over time, with newer cohorts stacked on older ones. It reveals whether topline growth comes from genuine cohort expansion or simply from new cohorts replacing dying old ones, which is a far less durable growth pattern.