Cohort AnalysisWatching customers age

Why averages hide what matters#

Suppose monthly revenue is rising. Good news, apparently. But it could be rising because you are acquiring many new customers who buy once and vanish, masking a steady decline in how long customers stay. The total looks healthy right up until acquisition costs rise and the leaking bucket is exposed.

Cohort analysis prevents that illusion. By following each group of customers separately from their starting month, it shows whether newer customers behave better or worse than older ones. That is the question behind retention, lifetime value, product-market fit and the real return on acquisition spend.

Totals tell you how big the bucket is. Cohorts tell you whether it leaks.

Step 1: choose the cohort and the behaviour#

A cohort needs a defining event and a time unit. The most common choice is acquisition cohort by month: everyone whose first purchase, first paid subscription or signed contract happened in a given month. Alternatives include cohorts by first product bought, acquisition channel, campaign or first-order discount.

Then choose the behaviour to track. Retention (share of the cohort still active, or who purchased again, in each later period) is the usual start. Revenue or gross profit per customer, cumulative over time, is the basis for lifetime value. Orders per customer suits repeat-purchase businesses.

Step 2: build the cohort table#

A cohort table has one row per cohort and one column per period since acquisition: month 0, month 1, month 2 and so on. Each cell shows the behaviour for that cohort in that period. Read across a row to see how one cohort ages; read down a column to see whether newer cohorts do better or worse at the same age.

The triangle shape is normal: recent cohorts have not been around long enough to fill later columns. You can build this in a spreadsheet from an order export (customer ID, order date, order value) with a pivot table, or in SQL against your order or CRM database. GA4's cohort exploration offers a web-behaviour version, but for revenue and retention your own transaction data is more reliable.

Keep the first version simple: monthly acquisition cohorts and a single behaviour such as repeat purchase. Add channel and offer splits only once the basic table is trusted. A cohort table that tries to show everything at once is as unreadable as the averages it was meant to replace.

Step 3: read the retention curve#

Plot each cohort's retention over time and look at the shape. Almost every business sees a sharp drop after the first period, as one-time buyers fall away. What follows matters more. A curve that flattens means a core of customers stays; that flat level is the foundation of the business. A curve that keeps sliding towards zero means customers do not find lasting value.

Then compare cohorts. Are newer cohorts' curves above or below older ones at the same age? Improvement means product, onboarding or targeting is getting better. Deterioration often accompanies rapid scaling: as acquisition expands into broader audiences or deeper discounts, the customers won are less well matched.

Read the curve in the units that match your purchase cycle. A grocery or personal-care brand might look at monthly repeat; a furniture or electronics brand should expect long gaps between purchases and may track referrals, accessories or service usage instead.

Step 4: compare cohorts by channel and offer#

This is where cohort analysis becomes a marketing tool. Split cohorts by acquisition channel, campaign or first-order offer, and compare their retention and cumulative value. A channel with a higher cost per customer may still win if its customers stay longer; a channel with cheap customers may lose if they never return.

Discount cohorts deserve special attention. Customers acquired with a deep first-order discount often behave differently from those who paid full price. If their repeat rate and value are lower, the true cost of the discount includes that lower lifetime value, not just the margin given away on the first order.

Step 5: use cohorts to estimate real LTV#

Cumulative gross profit per customer, by cohort, is the most defensible way to estimate lifetime value. Plot it over time: where older cohorts' curves flatten, you can see what customers are actually worth after a year or two. Use that, rather than an assumed lifetime, in your CAC and LTV calculations.

It also gives you real payback: the month at which a cohort's cumulative gross profit per customer exceeds the CAC paid to acquire it. Some cohorts pay back quickly, some slowly, some never. That is far more informative than a single average payback figure.

Common mistakes#

Two further cautions. Make sure customer IDs are stable: a customer who checks out as a guest with a different email may appear as two people and inflate churn. And account for seasonality: a cohort acquired during a festive sale may behave differently from one acquired in a quiet month, for reasons unrelated to quality.

Making cohorts a habit#

Refresh the cohort table monthly and put a simplified version in front of leadership quarterly. It is one of the few analyses that speaks to both marketing efficiency and long-term business health, and it tends to change the conversation from 'how many customers did we get?' to 'how many good customers did we get?' That shift sits at the heart of retention marketing and of any credible north star metric.

Published by Fabulous.Media, a network of specialist marketing agencies. Updated 9 October 2026. Platform features change often; check current official documentation before acting on platform-specific detail.