Skip to content

Cohort analysis

| Article: 2 min read | Customer retention

Cohort analysis splits customers into groups (cohorts) by a shared trait, for example the month of their first purchase, and compares how they behave over time. It shows what the overall average hides.

In plain words

A cohort is a group of people who share a trait or experience in a given period. Wikipedia gives an example: online store customers who registered in the last two weeks and made a purchase (Wikipedia). For a store, the most useful trait is the month of first purchase.

The benefit is that overall reports mix new and old customers. In the example in the Wikipedia article, the overall figures hid a problem that affected only part of the users, and the cohort breakdown revealed it (Wikipedia).

What it looks like

A table: rows are months of first purchase, columns are "month 0, 1, 2, 3…". Each cell shows the share of the cohort's customers who bought again, or their accumulated margin. a16z describes it like this: the first month is 100% of the cohort, then the percentage of those who keep buying (a16z).

Example (illustrative numbers)

The numbers are made up for the example. A clothing store.

Cohort

Customers

Bought again within 3 mo.

Bought again within 6 mo.

January

1,000

12%

19%

March

1,400

9%

14%

The March cohort is bigger but comes back less. The cause may be the sale used to attract it, or a new ad channel.

Tools

Google Analytics 4 has a "Cohort exploration" report: you choose what puts a user into a cohort (first visit, purchase) and which action keeps them in it, the granularity (day, week, month) and breakdowns. There are limits: up to 60 cohorts and the top 15 values of a breakdown (Google Analytics Help). For orders and repeat purchases, CRM data is more accurate.

Common mistakes

  • Cohorts that are too small. In a group of 20 people the percentages jump around at random.
  • Comparing cohorts of different ages. January with "six months of life" cannot be set next to a September cohort that is one month old.
  • Not separating channels. One cohort may combine completely different buyers.
  • Duplicate customers because of different phones and emails.

FAQ

What is cohort analysis in plain words? Comparing groups of customers by month of first purchase to see how their actions change.

What is it for in an online store? To see retention, the share of repeat purchases and LTV separately for each period and channel.

How do I do a cohort analysis? Formulate the question, choose a metric, define the cohorts and compare them over time. These are the stages of the method as described (Wikipedia).

Does Google Analytics have cohort analysis? Yes, GA4 has a Cohort exploration report in the "Explore" section.

Sources

Share

Related

  • Customer surveys

    | Article: 2 min read

    Customer surveys are short questions to customers about their experience (why they bought, what they couldn't find, how delivery went) that give a store what numbers don't show: reasons. Analytics shows where people stop, a survey helps you understand why.

  • RFM analysis

    | Article: 2 min read

    RFM analysis is a way to segment customers by three attributes: how recently they last bought (Recency), how often they buy (Frequency) and how much they spend (Monetary). It shows who to write to first and who has, it seems, already been lost.

  • Customer churn

    | Article: 2 min read

    Customer churn rate is the share of customers who left a group over a given period. The opposite measure is retention.

  • Customer service and replies to shoppers

    | Article: 2 min read

    An online store's customer service is everything that happens between a shopper's question and its resolution: answers before purchase, order status, changes, returns, complaints. It is part of the product: a shopper judges a store by how their issue was…