Cohort analysis
Cohort analysis is an analytical method in which users or customers are divided into groups, so-called cohorts, by a shared feature, mostly the time of their first interaction, and observed over a longer period. Rather than looking at aggregated averages, cohort analysis shows how a particular group's behaviour, all users won in January for instance, develops week by week or month by month. Patterns such as retention, churn and repeat purchase behaviour thus become visible that stay hidden in classic reports. The method is a core part of modern online marketing and the basis for well-founded decisions about product, campaigns and retention.
Also known as: cohort analysis, cohort report
What is a cohort and how is it formed?
A cohort is a group of users who experienced a shared event at the same time, most often first registration, the first visit to the site or the first order. The most common form is the acquisition-based cohort: all users won in the same calendar week form a unit whose development is then followed separately. Alongside these are behaviour-based cohorts, grouping users by a shared action, by completing an onboarding or using a particular feature for instance.
The decisive advantage lies in comparability over time. If retention falls overall, it stays unclear whether earlier customers are leaving or new customers are converting worse. Cohort analysis separates those effects cleanly, because each group is looked at on its own. You can thus see whether a product change in March actually improved the retention of the cohorts won after it.
Cohort analysis thus differs fundamentally from a mere snapshot. A single figure such as the total number of active users can look stable although old customers are constantly leaving beneath it and new ones taking their place, an effect that masks genuine problems with retention. Only following defined groups across their life cycle reveals whether a company really grows sustainably or is merely plugging a leak with ever more acquisition.
Reading the cohort grid
A cohort analysis is classically presented as a triangular matrix: the rows stand for the individual cohorts (acquisition weeks for instance), the columns for the time elapsed since the start (week 0, week 1, week 2 and so on). Each cell contains a figure, often the share of users still active as a percentage. Reading a row from left to right shows a cohort's life cycle; reading a column from top to bottom compares different cohorts at the same relative point in time.
In Google Analytics 4 has its own cohort explorer, which builds cohorts by acquisition date and shows metrics such as active users, revenue or transactions. Also Matomo offers a comparable cohort report without the sampling of the large platforms. For custom analysis many teams export the raw data and visualise the grid in Looker Studioto fit it into their Marketing dashboards to integrate.
One typical finding is the retention curve stabilising: in the first days the share of active users falls sharply, then flattens into a plateau. If that plateau is at zero, the product lacks lasting value; a stable plateau points to a solid core of loyal users.
Use cases in marketing
Cohort analysis answers concrete business questions. In e-commerce it shows how many first-time buyers from a promotional period order again later — a direct basis for calculating the Customer lifetime value. In SaaS it makes net revenue retention visible and reveals whether a new onboarding flow reduces early churn. In a campaign context you can compare whether users won through a particular channel are more valuable in the long run.
Concrete measures follow from these insights: a poorly retained cohort can be reactivated through targeted email sequences, a successful channel gets more budget. Cohort analysis thus becomes the link between mere Web analytics and operational control.
The method is particularly valuable for assessing changes. If a company introduces a new onboarding, a price adjustment or a changed product feature, the effect can be read cleanly from the cohorts won afterwards. If their retention rises against earlier cohorts at the same relative point in time, the measure succeeded. This causal reading makes cohort analysis one of the most honest instruments in data-driven Online marketingbecause it doesn't claim success but proves it over time.
Common mistakes and limits
As meaningful as cohort analysis is, it is just as easily misread. A common mistake is choosing an unsuitable granularity: daily cohorts produce noisy figures with small amounts of data, monthly cohorts blur short-term effects. The right resolution depends on the purchase frequency and the number of users and should be chosen deliberately rather than taking the tool's default blindly.
Seasonal effects are another pitfall. A Christmas cohort behaves differently from a summer cohort, so comparing cohorts always has to take the context into account. Mixtures of different channels within one cohort can distort the picture too; here an additional Segmentation by acquisition source, to compare like with like.
Finally, cohort analysis answers the what and when but rarely the why. It shows that a group drops off, not the cause. It is therefore ideally combined with qualitative methods such as user surveys or session recordings, to derive a solid hypothesis and concrete action from the pattern observed.
Data protection and data quality
Because cohort analysis follows users over longer periods, it touches the GDPR directly. What matters is lawful consent via a Consent management, since otherwise data is missing and cohorts stay incomplete. Aggregated, pseudonymised analysis is more privacy-friendly than analysis at individual user level and should be preferred wherever the question allows.
On the technical side, cleanly attributing the starting point is decisive. Cross-device switching, deleted cookies and differing time zones can distort cohorts. Regular Performance audit the Trackingimplementation makes sure the data basis is sound and the decisions drawn from it hold.
Frequently asked questions
What is the difference between cohort analysis and segmentation?
Segmentation divides users by properties such as region or device but mostly looks at them at one point in time. Cohort analysis groups them by a shared starting event and follows that group over time. The two methods complement one another: a cohort can additionally be segmented to make differences within the same acquisition period visible.
How do I build a cohort analysis in Google Analytics 4?
In Google Analytics 4 you open the cohort explorer under Explore. There you choose an inclusion criterion for the cohort, the first session date for instance, and a metric such as active users or revenue. GA4 then builds the cohort grid by weeks or months automatically. You can additionally set a granularity and a calculation type, standard or rolling retention for instance. For deeper analysis, exporting the data and visualising it in Looker Studio is advisable, where the grid can be built into existing marketing dashboards.
What cohort size makes sense?
Cohorts should be large enough for percentages to be stable — a few hundred users per cohort is a rule of thumb. With small data volumes, coarser granularity is advisable, that is monthly rather than weekly cohorts. That avoids individual users distorting the retention curve.
What does the retention curve tell you?
The retention curve shows what share of a cohort is still active after a day, week or month. A steep fall followed by a plateau is typical and healthy as long as the plateau is above zero. If the curve falls permanently to zero, the product lacks lasting repeat value and high churn looms.
Is cohort analysis GDPR compliant?
Yes, provided the underlying data is collected lawfully. Valid consent through consent management is required, and aggregated, pseudonymised analysis is preferable to individual-level analysis. Tools such as Matomo can be configured in a privacy-friendly way and therefore suit cohort reports well.
Related terms
Dividing a heterogeneous market or customer base into homogeneous groups so that messages and measures can be targeted.
A metric showing what share of customers leave within a period — a key indicator of retention and growth.
The expected total value of a customer across the whole relationship — the basis for budgeting and steering.
Collecting, measuring and analysing website data in order to improve content and conversions in a targeted way.
Structured, regular preparation of marketing data and KPIs as a basis for decisions.
A marketing approach that bases strategy, campaigns and optimisation systematically on data and metrics rather than intuition.
Put AI to work for your business?
We help you integrate artificial intelligence into your processes, your marketing and your website — strategically and securely.
Your contact
Stefan
I look forward to hearing about your project and finding the best solution together.
