top of page

Repeat Purchase Cohorts: Test Whether New Customers Actually Come Back

Writer: DIGITALTHROB
DIGITALTHROB
Oct 30
3 min read

Returning-customer sales this month do not tell you whether last month’s newly acquired customers came back. Fix the customer group at its first purchase, give each person the same amount of time to reorder and follow the resulting orders. That lets you compare observed repeat behaviour with the assumption used to fund acquisition.

A coral gift card and mint ribbon sit beside a cobalt gift box and amber accounting tokens.

Start with one acquisition group

A report showing returning-customer revenue cannot tell you whether the customers acquired by last month’s advertising are returning. That revenue may come from a much older customer base. Build a cohort from customers with a first accepted purchase in a defined period, using a consistent customer identifier.

Record exclusions such as staff orders and tests. Explain how cancellations, guest checkouts and customer records that cannot be matched are handled. If identification is incomplete, keep the limitation visible; quietly treating unmatched orders as new customers can distort both acquisition and repeat rates.

Give every customer a fair amount of time

Define the window relative to each customer’s first accepted purchase: for example, a second accepted order within the following 60 days. Report a cohort as fully observed only when every included customer has had those 60 days. If recent customers are still waiting to mature, show that partial group separately rather than treating its lower repeat count as a fair comparison.

The new versus returning customer breakdown remains useful for a period-level view. A cohort report answers a different question by keeping the original customer group fixed and allowing the observation time to develop. Retain both definitions rather than merging the figures.

Calculate people and money separately

Suppose a fictional cohort contains 100 first-time customers, and 18 place at least one further accepted order within the chosen 60-day window. The observed repeat-customer rate is 18%. If those people place 27 repeat orders, the order count is 27, not the number of returning people.

The fictional 18 returning people generated 27 repeat orders, so repeat orders per returning customer are 27 ÷ 18 = 1.5. That is different from the 18% repeat-customer rate. Add net revenue and contribution from those 27 orders in separate columns; a heavily discounted reorder can increase order frequency while adding little money toward acquisition costs.

Look for an offer or product mix explanation

Split a cohort only where the data volume and business decision justify it. First product, acquisition offer or delivery experience may help explain a pattern. Avoid slicing a small group into dozens of cells and presenting the resulting differences as reliable channel rankings.

Check whether a repeat order genuinely belongs to the original group and whether a refund reverses the supposed first purchase. Document the reconciliation date. The campaign platform and the customer order system may answer different attribution questions, so use customer records for the repeat history and label acquisition-source uncertainty.

Use the evidence to revise the acquisition assumption

Compare observed contribution with the repeat value used in the original acquisition plan. If the repeat assumption is not appearing, reduce the amount you rely on or extend the observation period where appropriate. Do not keep spending against an optimistic forecast simply because a few loyal customers exist.

A repeat cohort is most useful when it changes a decision: an acquisition ceiling, a first-order offer, a product experience fix or a retention experiment. Record the next review date and preserve earlier snapshots. That makes later learning traceable without pretending the first estimate was known from the start.

If acquisition decisions depend on customers buying again, explore our Meta Ads management service and use a consistent customer-outcome view alongside campaign reporting.

 
 
 

Comments


Discover clics solution for the efficient marketer
More clics

Never miss an update

bottom of page