Customer retention, and the three metrics people confuse for it
In short
Customer retention measures how many customers from a defined starting group remain active in a later period. State what active means, which cohort you follow and how long you observe it. Retention rate divides by the starting cohort, repeat purchase rate by customers with any order in the window, and revenue retention by the cohort's baseline revenue; report them separately.
What customer retention actually measures
Customer retention is the share of customers you had who are still customers at the end of a period. That is the whole definition, and it hides the two decisions that determine the answer: which customers you started with, and how long the period is.
Changing the starting group or observation window changes what the rate describes. Compare companies or periods only after aligning those definitions. For a non-subscription store, define active through a purchase window suited to the category; a customer with no purchase this month is not necessarily lost.
The three metrics people call retention
They answer different questions and routinely disagree.
| Metric | What it asks | Where it misleads |
|---|---|---|
| Retention rate | Of the customers I had, how many are still here? | Says nothing about how much they spend |
| Repeat purchase rate | Of all customers, how many bought more than once? | Rises when you acquire fewer new customers |
| Revenue retention | Of last period's revenue, how much came back? | Can look healthy while customer count falls |
An aggregate repeat purchase rate can rise when fewer new customers enter the denominator, even if existing customers behave the same way. It can also fall if repeat purchasing weakens. A lower advertising budget by itself does not determine the direction.
If you report one number to a board, report revenue retention and say what the customer count did. If you report one number to a marketing team, report repeat rate by acquisition cohort, because that is the one a campaign can move.
How to compute each one
Define the observation window before counting. For a store, it should reflect the purchase cycle and the question you need to answer.
- Customer retention rate: customers from a fixed starting cohort who meet your activity rule in the follow-up window, divided by all customers in that starting cohort. New arrivals do not enter the denominator.
- Repeat purchase rate: customers with at least two orders in the stated window, divided by customers with any order in that window. If you instead include purchase history from before the window, name that definition explicitly.
- Revenue retention: revenue in the follow-up period from a fixed starting customer cohort, divided by that same cohort's revenue in the baseline period. Use comparable periods and revenue definitions.
A lifetime repeat purchase rate can fall when new one-time buyers enter the base. Revenue retention can exceed 100% if growth among retained customers outweighs lost revenue. Neither measure establishes that a retention campaign caused the change.
Break the results down by acquisition cohort when sample sizes allow, and compare cohorts at the same age. For the product workflow, see Loyalz retention measurement.
Why your retention number is probably measuring your acquisition
New-customer volume affects an aggregate repeat purchase rate. It does not directly change the denominator of a correctly fixed retention cohort.
Separate the two in your report. Show how many new customers arrived, how established cohorts repurchased and whether channel mix changed. When comparing cohorts, allow each the same time to buy again.
A campaign may bring customers with different repeat-purchase patterns. Following CAC payback by cohort helps show the financial consequence without treating an attribution report as proof of causation.
What no retention metric can tell you
Descriptive metrics show what happened. They do not isolate the effect of a loyalty programme, email flow or win-back offer.
Programme members may already have been more likely to buy again before joining. Comparing them with non-members mixes that selection effect with any programme effect. A randomised control group can separate these more reliably when the experiment is large enough and implemented correctly. Other causal methods require additional assumptions; a simple before-and-after comparison leaves competing explanations unresolved.
The loyalty campaign guide covers experiment design. Report reward costs, uncertainty and whether the observation window is complete. A test can be inconclusive, and withholding an ineffective offer need not cost revenue.
For basket-size decisions, see average order value and margin and the other retention guides.
Common questions
What is a good customer retention rate?
Compare rates only with the same activity definition, cohort age and observation window. Product category and purchase cycle matter. Your own comparable cohorts provide a useful baseline; a carefully matched external benchmark can provide context.
What is the difference between retention rate and repeat purchase rate?
Retention follows a fixed starting group into a later period. Repeat purchase rate measures the share of buyers with repeated orders under a stated window and history rule. New buyers can change the latter without changing the behaviour of an existing cohort.
How do I know if my loyalty programme is working?
Where feasible, randomly assign eligible customers to an offer or control group before the campaign. Compare contribution per assigned customer after reward costs, over the same complete window. Member-versus-non-member averages alone cannot isolate the effect.
Should retention be measured on customers or on revenue?
Both, and they can move in opposite directions. Revenue retention above 100% with a falling customer count means a shrinking group of customers is spending more — which is a real result and also a concentration risk worth naming out loud.
See it on your data
On a twenty-minute call we connect one ad account and one month of your orders, and work out the gap between what the platform claims and what your shop took. Profit and POAS stay blank until you enter your margin and costs per order.
Book a demo