Cohort analysis

What it means

Cohort analysis follows groups of customers who started in the same period and compares their behaviour over time. A cohort might be all customers acquired in January, the next those from February. For each, it shows how many are still buying after a month, six months or a year, and how much they spend.

Summary figures such as total customers or total sales can hide a problem. If a company is winning new customers fast, the totals grow even if new customers leave faster than older ones did. Cohorts reveal whether retention is improving or getting worse.

The result is usually a table with cohorts as rows and months since start as columns. From it come churn, customer lifetime value, NRR for subscriptions, and insight into which product or marketing changes affected customer behaviour.

How it works in practice

For companies with a subscription model or repeat purchases, cohort analysis is one of the most important reports for management and investors. It shows whether investing in customer acquisition pays off because customers come back, or whether the company is pouring water into a leaking bucket.

The input is invoicing or online shop data with each customer’s first purchase date. When linking it to the books, we make sure revenue in the analysis matches revenue in the accounts, net of credit notes and cancelled orders.

Cohorts can also be split by acquisition source, such as advertising, referrals or search. It often turns out that a cheaper channel brings customers who leave quickly, and a more expensive one customers with higher total value.

What to watch out for

A common mistake is drawing conclusions from the latest cohorts that do not yet have enough history. A cohort from two months ago cannot show annual retention. Always compare the same period since start.

The second risk is small cohorts. If a cohort has a few dozen customers, one large customer changes the whole picture. With small numbers it is better to combine cohorts, for example by quarter.