Definition:

Cohort analysis is a technique that groups users, customers, accounts or other units that share an event or characteristic during a defined period and compares how their behaviour develops over time.

A cohort might consist, for example, of people who registered during the same week or made their first purchase in a particular month. The entry criterion and starting point must be defined, because they determine who belongs to the group and when its development begins to be measured.

A cohort is not just any segment. In market segmentation, membership may depend on current attributes and change over time; cohort analysis usually preserves a group formed around a shared event or period so that its trajectory can be compared.

How cohort analysis is performed

The analysis begins with a question and a metric linked to it. The procedure can be summarised as five decisions:

Define the objective: specify whether the analysis will study retention, repeat purchases, revenue, activation, feature use or another behaviour.
Set the entry condition: choose the event that adds each unit to the cohort, such as first registration, first purchase or the start of a subscription.
Choose the granularity: group by day, week, month or another interval suited to the cycle being observed.
Set the observation window: decide how many periods will be measured and which condition will count as a return or outcome.
Compare cohorts: display the values by elapsed time since entry and examine differences between comparable groups.

A cohort table commonly displays groups in rows and elapsed periods in columns. Each cell represents a cohort’s result at a comparable stage of its cycle, not necessarily on the same calendar date.

The definition of a user, event and period depends on the tool and the web analytics implementation. Two platforms may produce different values even when they use similar labels.

Applications of cohort analysis

The technique can be used for digital products, subscriptions, ecommerce and other services with repeated activity. These five applications are common:

Retention: observe the proportion of a cohort that returns or remains active after one or more periods.
Activation and conversion: measure how long each group takes to complete a relevant action or conversion.
Revenue and cumulative value: compare purchases, renewals or revenue generated by cohorts during an equivalent window.
Acquisition: study the development of groups acquired through different campaigns, channels or landing pages.
Product changes: compare cohorts exposed to different versions, prices, onboarding processes or conditions.

These comparisons can reveal differences, but a better result does not by itself prove that the channel, campaign or change caused it. Cohorts may also differ in size, composition, seasonality and context.

Interpretation and limitations

Recent cohorts have had less observation time than older ones. Cells that cannot yet be observed must be distinguished from zero values, and comparisons should use equivalent windows to avoid misleading conclusions.

The results also depend on measurement quality. Changes in consent, identifiers, devices, tags or event definitions can alter the series. The analysis represents the data collected, not necessarily every person who used the product or service.

Google Analytics 4 includes a cohort exploration that can define inclusion and return criteria and use daily, weekly or monthly granularity. A tool’s options shape the report, but they do not replace a clearly defined question or a review of its limitations.

When an effect must be attributed to an intervention, experiments or other methods may also be required. Cohort analysis describes differences and change over time; it does not establish causality automatically.