Customer cohort analysis for Shopify

Customer cohort analysis follows a group of customers who share a starting event and compares their behavior as time passes. For an ecommerce store, a useful starting point is the month of each customer’s first purchase. You can then ask how many of those customers buy again in later months.

The key comparison is the same age of each cohort: April customers in their first later month versus May customers in their first later month. Comparing lifetime purchases across groups acquired at different times gives older customers more opportunity to return.

This guide uses first-purchase cohorts and repeat purchasing. Membership retention is a separate question: a member can remain subscribed without placing a merchandise order, and a repeat buyer need not be a member.

In Shopify Analytics, open Reports and search for Customer cohort analysis. Shopify’s customer reports documentation describes first-order cohorts, configurable metrics and intervals, and heatmap or retention-curve views. Confirm the selected metric before interpreting a cell.

Shopify’s default report separates first orders from period 0, which captures returning orders in the same period as the first order. Do not assume every Month 0 retention cell should be 100%. Our simplified worked table below starts at Month 1 and omits first-order and Month 0 columns.

The report configuration guide explains cohort definitions and intervals. Record your metric, date range, first-order filters and customer filters so a later comparison uses the same settings.

Before using your own export, confirm how it handles customer identity, cancellations, refunds, test orders and imported history. A customer’s first order in a short export is not necessarily their first-ever store order. Use the store’s reporting definition consistently and document any custom exclusions.

For the worked example, a customer belongs to the calendar month of their first qualifying purchase. Count each customer at most once in each later month, even if they place several orders. Keep the original cohort size as the denominator.

For 100 customers first acquired in April, 20 distinct customers buying in May gives a Month 1 rate of 20%. If those 20 customers place 27 orders, the customer rate is still 20%, not 27%. Revenue, orders and customers are different measures.

Calendar Month 1 is not the same as days 1-30 after each purchase. An April 1 buyer and an April 30 buyer are both measured in May here, although their elapsed time differs. For an exact 30-, 60- or 90-day question, build and label an elapsed-day analysis instead.

Fictional data, complete through August 31, 2026. Each cell shows unique purchasing customers and their share of the original cohort. Only complete calendar months are included. These are illustrative numbers, not Memberply customer results or industry benchmarks.

Read across: April’s Month 1 is May, Month 2 is June, Month 3 is July and Month 4 is August. Its rate rises from 15% to 18% between Months 2 and 3. That is possible because customers can skip a month and return later; this is not continuous subscription survival.

Read down: Month 1 is 20% for April and May, 25% for June and 30% for July. July’s 30% versus April’s 20% is a 10-percentage-point difference, or a 50% relative increase. It is an observation to investigate, not proof that a particular campaign improved retention.

Read the gaps: July’s Month 2 would be September, which is outside this example’s cutoff. “Not observed” is not zero. If you run the report during September, an incomplete September period still should not be compared as a finished month.

The table uses calendar months, so month lengths also vary. Use the same interval convention throughout, and choose an elapsed-day method if an equal number of days is essential to the decision.

Overlap example: if April has 20 buyers in Month 1 and 15 in Month 2, with five customers in both, there are 20 + 15 − 5 = 30 unique buyers across those two periods. The combined rate is 30%, not 35%. This example deliberately excludes any repeat purchases in Month 0.

Weighted example: April, May and June together have 15 + 18 + 16 = 49 Month 2 buyers out of 100 + 120 + 80 = 300 original customers. The pooled rate is 16.33%. The unweighted average of 15%, 15% and 20% is 16.67%, which answers a different question. July is excluded because its Month 2 is not observed.

A missing value can also reflect a reporting or data issue. Investigate the reason before labeling it as unobserved, and keep that separate from a genuine completed period with zero repeat buyers.

Use the RFM analysis guide to choose relevant customer groups for action. Use ecommerce metrics to keep reporting definitions consistent. You can combine methods, but label each denominator and observation window.

Someone making a second purchase may need a different follow-up from a first-time buyer. Cohort analysis helps locate the pattern; individual purchase history is still needed to choose the actual audience.

For practical messages, see the post-purchase email guide. Use actual purchase history, channel eligibility and stop rules rather than sending to everyone in a cohort indiscriminately.

Where practical, randomize eligible customers into a test and holdout group within the same cohort. Predefine the outcome and time window. Comparing a campaign launched in July with an April cohort alone does not isolate its effect. Small cohorts need particular caution; a handful of customers can move the percentage substantially.

A membership-focused analysis can group people by when they joined and separately track active membership, benefit usage and merchandise purchasing. Define the joining event and treatment of trials, pauses, cancellations and reactivations before interpreting the result.

Keep first-purchase cohorts and membership-join cohorts distinct. Do not classify customers as members at acquisition merely because they joined later: that uses future information and can make the comparison misleading. Current membership status alone does not reconstruct historical status.

Members and non-members may differ before joining, so stronger member repeat purchasing is not automatically caused by the program. Compare similar histories and observation periods, and use a suitable experiment where feasible.

Memberply provides membership tiers and benefits. This guide does not describe a native Memberply cohort report. Use Shopify reporting or a separately prepared analysis, and verify that the historical data needed for a membership comparison exists. The customer segmentation guide covers useful audience distinctions.

1. Define entry. Use a documented first-purchase or membership-join event and preserve the original group.
2. Choose the metric. Separate unique buyers, orders, revenue, contribution and active memberships.
3. Set the time convention. Record calendar or elapsed-day intervals, timezone and data cutoff.
4. Check completeness. Exclude incomplete and unobserved periods from like-for-like comparisons.
5. Check the population. Document identity handling, order exclusions, product mix and acquisition filters.
6. Choose one next test. Record the hypothesis, audience, cost and outcome window before changing the experience.

Store the definitions alongside the ecommerce KPI worksheet. Keep observed values separate from forecasts; projections are estimates, not completed customer behavior.