Most metrics answer the question "how did my account do this month". Cohort Analysis answers a different and more valuable question: what does one new customer do in the months after you acquire them. That number tells you what a customer is really worth, how long it takes to earn back their acquisition cost, and how aggressively you can afford to advertise. This article explains how the table works.

What a cohort is

A cohort is simply all the customers whose first purchase happened in the same month. Everyone who bought from your brand for the first time in January forms the January cohort. Everyone whose first order came in February forms the February cohort, and so on.

The important rule: a customer belongs to one cohort, permanently. When a January customer comes back and reorders in February or March, that repeat purchase counts toward the January cohort, because January is when you acquired them. Repeat orders never start a new cohort.

This rule is what makes the analysis possible. By keeping every customer attached to the month you acquired them, the table can show you how the value of an acquired customer builds over time, instead of blending new and returning customers into one average where that story disappears.

What the columns 0, 1, 2, 3 mean

The columns are not calendar months. They count months since the first purchase:

Month 0 is the month of the first purchase itself. This is what a customer is worth on day one.
Months 1, 2, 3, 4, 5, 6 and beyond are the months after the first purchase. This is where repeat purchases land.

Because the count starts from each cohort's own first month, the same column means a different calendar month for each row: for the January cohort, month 2 is March; for the February cohort, month 2 is April.

Values are cumulative. Each column includes everything up to that point, first order included. If the January cohort shows $46 revenue per customer in column 3, the average January customer generated $46 in total through their first four months: the first order plus all repeats so far. Reading along a row, the number can only grow.

Empty cells at the bottom right are normal. Younger cohorts have not lived that many months yet: a cohort acquired three months ago cannot have a month 6. For the same reason, avoid judging your most recent cohorts. Their repeat purchases simply have not had time to happen. Give a cohort two to three months before drawing conclusions.

The views and what each one tells you

Open the LTV & Subscriptions report and go to the Cohort Analysis page. The same table can display different metrics:

LTV - Units: how many total units the average customer has purchased.
LTV - Revenue: how much total revenue the average customer in the cohort has generated.
LTV - Profit: the same logic in profit, with advertising included in the acquisition cost. This view shows whether a cohort started negative (ad spend exceeded first-order margin) and in which month it turned profitable.
LTV - Margin: the profit view expressed as a percentage instead of a dollar amount.
Active Customers: how many of the cohort's customers have come back to reorder each month.
Retention Rate %: percentage of active customers
Monthly Cohort Revenue: the cohort's total revenue by month, rather than per customer.
Cumulative Cohort Revenue: total revenue generated by the cohort from day 0 till today.

Filtering cohorts

You can filter the table by product (parent or child ASIN) and by order type (Subscribe and Save versus non-Subscribe and Save). Comparing the Subscribe and Save cohorts to the regular ones for the same product shows how much subscription behavior changes LTV, and whether pushing customers toward subscribing is worth the investment.