How to Read a Cohort Analysis for a Shopify Store

A retention cohort table groups customers by the month of their first order and shows, for each group, how many ordered again in each month that followed. Each row is one group, each column is the time since that group's first order, and each cell is the share of the group that came back. This guide explains how to read rows, columns and diagonals, how to recognise a weak acquisition month, how to cut cohorts by first product or channel, and the mistakes people make most often.

What does a retention cohort table show?

It shows whether customers return, and how that changes depending on when they were acquired.

The layout is nearly always the same:

• Rows: one per cohort, usually labelled by first-order month, with the number of customers in the cohort beside it.
• Columns: months since first order. Month 0 is the month of the first order, month 1 is the month after, and so on.
• Cells: the percentage of the cohort that placed an order in that month, often shaded so that darker means higher.

Because newer cohorts have had less time, the filled cells form a triangle. The January row may have twelve columns of data and the November row only two. Nothing is missing; those months have not happened yet.

How do you read across a row?

A row is the life of one group of customers. Read it left to right to see how quickly they fade.

Take an example with made-up round figures. Say the March cohort has 500 customers. In month 1, 60 of them order again, in month 2 it is 45, and in month 3 it is 40. The row reads 12%, 9%, 8%.

Two things to take from a row:

• The first drop. The step from month 0 to month 1 is usually the steepest, since many customers only ever buy once.
• Where it flattens. In the example the row settles at around 8%. A row that levels off means a core of customers has formed a habit. A row that keeps sliding towards zero means the product is not being rebought.

How do you read down a column?

A column compares every cohort at the same age, which is the fair comparison.

Reading down answers the question a row cannot: is retention getting better or worse over time? If month 2 retention climbs steadily as you move down the column, newer customers are returning more than older ones did at the same stage. If it falls, something changed in who you are acquiring or in what they receive.

Diagonals have a meaning too. Cells on the same diagonal fall in the same calendar month. A dark diagonal stripe usually marks an event that reached everyone at once, such as a sale or a big email campaign, and should not be read as a lasting improvement in any one cohort.

How do you spot a bad acquisition month?

Look for a row that is paler than its neighbours from month 1 onwards. A large cohort with low retention is the classic sign of customers bought with a discount or an offer that attracted people with little interest in buying again.

A simple routine:

1. Find the largest cohorts by customer count. These are usually your promotion months.
2. Compare their month 1, month 2 and month 3 cells with the rows above and below.
3. If a large cohort is clearly weaker, note what was different that month: the offer, the channel, the product being pushed.

This matters for spend: customers who never return have to repay their acquisition cost on the first order alone. Our guide to calculating customer lifetime value turns these rows into a value per customer.

Why cut cohorts by first product or channel?

Because the month a customer arrived is often less telling than how they arrived and what they bought first.

• By first product. Some products are doorways to a second purchase and others are one-off buys. Grouping customers by the first item in their first order shows which products create repeat customers. Those deserve the front of the store and the ad budget, even if their first-order margin is thinner.
• By acquisition channel. Customers from a referral, a search for your brand, or a paid social campaign may repeat at quite different rates. This requires that orders carry reliable source data, which depends on consistent campaign tagging.

What are the common misreadings?

Most errors come from forgetting what a cell is a percentage of.

• Ignoring cohort size. A cohort of 40 customers where six return shows 15%; one more or fewer moves it by several points. Read the customer count before the percentage.
• Comparing incomplete periods. The most recent column of each row often covers a month that is still in progress. Treat the edge of the triangle as provisional.
• Mixing per-period and cumulative tables. Some tables show the share who ordered in that month; others show the share who have ordered again at any point so far. Cumulative figures can only rise. Check which you are reading.
• Confusing customer retention with revenue retention. A table can count returning customers or the sales they bring. They can tell different stories if returning customers spend more per order.

A related but different figure is covered in our piece on returning customer rate.

Where do you get a cohort table?

You can build one from an order export, which is a chore to repeat each month. The guide to what Shopify's built-in reports leave out explains why merchants usually want this view sooner or later. Marka Metrics provides retention cohorts, customer segments and predicted lifetime value on its Growth and Pro plans, built from your Shopify orders.

A short checklist

1. Confirm what each cell means: per month or cumulative, customers or revenue.
2. Read the cohort size before the percentages.
3. Read rows for the shape of decay and columns for change over time.
4. Flag large cohorts with weak month 1 to month 3 retention.
5. Treat the newest cell in each row as provisional.
6. Recut by first product or channel when a month looks unusual.

A cohort table rewards ten minutes of careful reading each month. If you would like the cohorts drawn from your orders and kept up to date for you, Marka Metrics does that inside the Shopify admin.