How to Measure Customer Loyalty and Identify Which Customers Stop Returning

Before measuring, define what returning means

In a business with repeat sales, customer loyalty cannot be summed up in a single figure. A customer may buy several times, remain active for a certain period, or generate a significant share of revenue. To find out whether your relationship with customers is strengthening or weakening, it helps to combine indicators and apply consistent criteria.

Start by defining three things: who counts as a customer, what behavior counts as a repeat purchase, and how long you will observe it. A monthly purchase may be typical for a frequent-consumption retailer, but not for a durable-goods retailer. If a store considers someone inactive after 30 days without a purchase, even though its normal cycle is six months, it will record artificially high churn.

Choose a time window that matches the purchase cycle and keep it consistent when comparing periods. If the cycle varies considerably across categories or customer types, calculate the indicators separately instead of imposing a common threshold.

Four indicators for understanding customer relationships

Repeat purchase rate. Measures what proportion of customers in an acquisition cohort made a repeat purchase within a defined follow-up period. A simple formula is:

Customers in the cohort whose first purchase occurred in the acquisition period and who made another purchase within the follow-up period ÷ customers in that cohort whose first purchase occurred in the acquisition period × 100

For the comparison to be valid, observe all customers in the cohort for the same follow-up period. This indicator shows whether new buyers return. For fast-replenishment products, a short period may be informative; for occasional purchases, a longer observation period may be needed.

Retention. Indicates what share of an initial group is still buying or meets the activity criterion at the end of the period. If the business tracks customers at the beginning, new customers, and customers at the end, it can calculate retention as follows:

(Customers at the end − new customers during the period) ÷ customers at the beginning × 100

The formula avoids counting customers who arrived during the period as retained customers. It is essential to specify what “still active” means—for example, having made a purchase within the last six months. In businesses without an ongoing contractual relationship, retention depends on this definition and does not necessarily mean that the customer has canceled.

Customer churn. In subscription models, this usually measures the proportion of paying customers who cancel or do not renew during a period, relative to the number who were customers at the start. For non-contractual sales, there is no observable cancellation: someone who goes longer than a set period without buying can be classified as churned, but that period is an analytical rule, not proof that the customer will not return.

Purchase frequency and value. Supplement the rates with the number of purchases per customer and the associated amount over a comparable period. Two groups may show a similar proportion of repeat buyers and still differ in frequency or spending. In subscriptions, customer churn is also distinguished from revenue churn: losing one large account can affect revenue differently from losing several small accounts.

Do not interpret a single figure as a universal rating. Rates vary by industry, business model, category, and purchase cycle. For a small or medium-sized business, it is usually more useful to track how its own indicator changes, using stable definitions, and compare it across relevant segments.

Use cohorts to identify when behavior changes

A cohort is a group of customers who share a defined characteristic. In an acquisition cohort, everyone is grouped by the date of their first purchase—for example, by month or quarter—and the proportion who make another purchase in subsequent periods is observed. This lets you compare groups with the same amount of follow-up time: the second-month retention of a recent cohort should not simply be compared with that of a cohort observed for a year.

You can also create behavioral cohorts, such as buyers in a particular category, customers who used a specific channel, or people who purchased a particular offer. The comparison may reveal where to investigate: perhaps people who bought a certain product return sooner, or a recent cohort has fewer repeat purchases than earlier ones.

That contrast alone does not prove that the product, channel, or offer caused the difference. Cohorts may have received different campaigns, have different profiles, or have purchased during seasons with particular patterns. Track these factors and treat the comparison as a signal for asking questions, not as automatic attribution.

A practical process for a small or medium-sized business

1. Set the unit of analysis. Decide whether you will measure individuals, households, or business accounts, and how you will handle duplicates, shared purchases, and customers who return after a period of inactivity.
2. Define activity and time windows. Specify what counts as a repeat purchase, how long each period lasts, and when you will consider a customer inactive. Adapt the criterion to the normal cycle for each product or service.
3. Calculate a baseline. Measure repeat purchase, retention or churn, frequency, and amount using data from the same interval. Record the definitions so the calculation can be repeated.
4. Compare segments and cohorts. Where relevant, separate categories, channels, or acquisition groups. Make sure each group has had enough time to show the behavior you are measuring.
5. Investigate the differences. Review the purchase journey, service, availability, price changes, and campaigns. If possible, ask customers who stopped buying; a metric identifies the pattern, but does not by itself explain the reason.
6. Test an action and measure again. For example, you can test a replenishment reminder or an improvement to customer service with a defined group and compare the result with a comparable group. Keep the period and criteria consistent so you can interpret the change cautiously.

Turn indicators into decisions without overinterpreting them

If repeat purchasing declines, first check whether the time window, assortment, acquisition channel, or customer mix has changed. If overall retention remains stable but worsens for a specific cohort, look for what sets that group apart before extending a response to the entire customer base. If purchase frequency falls while the number of active customers appears stable, the problem may not show up in a traditional churn rate.

The goal is not to pursue the highest percentage at any cost, but to identify which customers return, when they do so, and which patterns warrant investigation. Measurement helps guide commercial tests and assess how they change over time; on its own, it does not establish why someone stopped buying or guarantee that an action will win them back.