Когортный анализ
Cohort Analysis
Cohort analysis is a method of studying the behavior of user groups that share a common trait, such as registration date or first deposit, over a single observation period.
What is Когортный анализ (Cohort Analysis)?
Cohort analysis splits the audience into groups (cohorts) by a shared trait and compares how each group behaves over time. A cohort is most often formed by the date of a first action: registration, first deposit, or first purchase. Then it measures what share of the group returns, deposits, or completes the target action on day 1, day 7, day 30, and so on.
## How it works
A table is built where rows are cohorts (for example, the registration week) and columns are periods of lifetime (day 0, day 1, day 7). Each cell holds a metric, such as retention, ARPPU, or payout share. Reading a row shows how a single group degrades or grows. Reading a column lets you compare different cohorts at the same stage of their lifecycle and tell whether the product or the traffic has improved.
## Where it is used
In gambling and betting, cohorts help assess player retention after a bonus campaign and the effect of changing the wagering requirement. In crypto and fintech, they track wallet activity and the share of repeat transactions. In dating, they show user churn after the first week. In nutra, they reveal repeat orders and cancellations across traffic batches from different creatives.
## What it affects
Cohort analysis shows whether acquired users pay off in the long run. It is used to calculate LTV, estimate CAC and payback period, and decide which traffic sources to scale and which to shut down.
## Common mistakes
Comparing cohorts with different observation windows without adjusting for age: a newer group has not yet shown its long-term behavior. Mixing users from different traffic sources in one cohort, which blurs the conclusions. Looking only at averages without the distribution: a few large players can distort the picture. Failing to record product or offer changes that altered a cohort's behavior.
FAQ
How does cohort analysis differ from standard analytics?
Standard analytics looks at overall metrics for a period, which mix old and new users together. Cohort analysis separates them by the date of first action, so you can see the real behavior of each group over time.
Which cohort size should I choose?
Weekly or monthly cohorts are most common. Weekly cohorts catch changes in the product and traffic faster, while monthly cohorts give more stable data when registration volume is low.
What do I need to start cohort analysis?
You need events with a date and user identifier, plus the registration source. Data is usually exported from a tracker or data warehouse, and then a retention or revenue table by cohort is built.