Glossary
What is cohort analysis?
Cohort analysis groups customers by when they first bought and tracks each group forward, which separates a retention problem from a recruitment one in a way a single overall figure cannot.
An overall repeat rate can improve for bad reasons: a business that loses half its new customers sees the ratio rise because the denominator shrank. Cohorts fix that by asking a narrower question: of the customers who first appeared in March, how many came back within ninety days, and then asking it again of April, and May. If that figure holds steady while the overall rate falls, the business is recruiting more customers rather than keeping fewer, which calls for a completely different response.
What is the simplest useful cohort report?
One row per month: how many customers first bought that month, and what share of them returned at all within a window suited to the trade. Ninety days works for most businesses; a garage needs a year.
That single column is enough to answer the question most owners have, which is whether last quarter's change came from new customers behaving differently or from existing ones drifting away.
Do you need special software for this?
No: a spreadsheet and the customer list of enrolment dates and visit dates, which belongs to the business and can be returned on request to [email protected], is enough for a business with a few hundred customers. The analysis is a count, not a model.
What you do need is the underlying record, which is the part a cash-and-card business without a loyalty programme does not have. Cash visits are invisible and most small tills do not identify customers at all.
How often is it worth running?
Quarterly for most businesses, and never more often than one purchase cycle. Reading a cohort before its members have had a realistic chance to return produces a figure that only measures how recently they joined.
The exception is the first few months of a new programme, when the question is enrolment rate rather than retention. Judge a programme on how many people join until a full reward cycle has passed.
See also
- Purchase frequency: Purchase frequency is how often a customer buys within a period, and its inverse, the typical gap between one customer's visits, is what sets every other number in a loyalty programme.
- Retention rate: Retention rate is the proportion of customers from one period who return in the next, and even a small increase compounds substantially over a year.
- Lapsed customer: A lapsed customer is someone who used to visit regularly but has not returned within an expected window, commonly 45 days for a café (the Loonine default, adjustable from 1 to 365 days in the app), or two visit cycles for an appointment business.
- Churn: Churn is the rate at which customers stop returning, and in a local business it is usually silent: customers rarely announce they have left, they simply stop appearing.
Further reading
How to measure repeat customer rate (and what to do with it)
The formula for repeat customer rate, the window to measure it over, what a good figure looks like by industry, and the two numbers that are more useful.
Birthday and seasonal rewards: worth it, or a distraction?
When a dated reward earns its place, what to collect and what not to, and why the seasonal message usually matters more than the reward attached to it.