A single retention rate mixes customers acquired three years ago with customers acquired last month and says nothing about the direction of the business. A cohort chart separates them: each row is a month of new customers, each column is how many are still active after one, two, three months. Read across a row to see how a group ages; read down a column to see whether newer groups are doing better.

Define active for your business

For subscriptions: still paying. For e-commerce: ordered within a window that matches the product's purchase cycle. For software: logged in and did the core action. For services: has an open engagement or a booked appointment. The definition should be one a system can compute nightly without judgement.

Build it automatically

From the clean customer and event tables: cohort by first-purchase or signup month, and for each subsequent month, the share still active. A scheduled query produces the table; a dashboard renders it as a triangle with colour. Nobody should build this in a spreadsheet more than once.

Read it for decisions

Rows that flatten early are good — customers who stay past month three tend to stay. Newer rows sitting above older ones at the same age mean recent changes helped; below means they hurt. A sudden drop in one column across all rows is an event — a price change, an outage, a competitor. Cut the chart by acquisition source and by first product to find which customers stay.

Pair it with revenue

The same triangle with revenue per cohort instead of share active shows whether the customers who stay are spending more. A cohort that shrinks in count but grows in revenue is a business finding its best customers; one that grows in count and shrinks in revenue is buying volume.