Cohort Analysis: What It Is and How to Run It by Hand
By Alva Chew · Updated
Definition Cohort analysis groups customers by a shared starting point, such as the month they joined your loyalty program, and then compares how each group behaves in the months that follow. It shows whether newer members stay as active as older ones. You can do it by hand in a spreadsheet with three columns: member, join month and points earned each month.
What is cohort analysis?
Cohort analysis is a way of comparing groups of customers who share a starting point. In a loyalty program the starting point is usually the month a customer joined. Everyone who joined in January is the January cohort, and everyone who joined in February is the February cohort. You then follow each cohort for the same number of months and compare them.
The method comes from the social sciences, where researchers follow groups of people born or enrolled in the same period. A shop owner uses the same idea on a smaller scale.
Why does it beat one overall number?
A single customer retention rate blends everyone together. If you ran a street promotion in February and it brought in many people who never came back, your overall figure falls, even though your January members behaved exactly as before.
Cohorts keep the groups apart. You can see whether members who joined after a change, such as a new reward, stay active longer than members who joined before it. You can also see how fast activity fades in the first few months, which is when most members drift away.
What data do you need?
Three columns are enough:
| Column | Example |
|---|---|
| Member | Member 0147 |
| Join month | 2026-01 |
| Points earned in each month after joining | Month 1: 60, Month 2: 0 |
Points earned are proportional to spend, so they stand in for money. If you give 2 points per US$1, divide points by 2 to get US$. The earn rate page explains how to set that ratio.
Gimify does not include cohort reports. If you want this view, you build it yourself from whatever record of join dates and points added you keep.
How do you run it by hand?
- Sort members into cohorts by join month.
- For each cohort, count how many members earned points in month 1, month 2, month 3 and so on after joining.
- Divide each count by the size of the cohort to get the percentage still active.
- Add up the points for the cohort each month and divide by 2 to get spend in US$.
- Place the cohorts in a table with months-since-joining across the top, so each row reads left to right.
What does a worked example look like?
Take a café that gives 2 points per US$1. The January cohort has 40 members and the February cohort has 50. The figures below are invented for illustration.
| Cohort | Size | Active in month 1 | Active in month 2 |
|---|---|---|---|
| January | 40 | 22 (55%) | 16 (40%) |
| February | 50 | 20 (40%) | 15 (30%) |
Now add spend. January members earned 2,640 points in month 1, which is US$1,320, or US$60 per active member. February members earned 2,000 points, which is US$1,000, or US$50 per active member. In month 2 the January cohort earned 1,760 points (US$880, US$55 each) and the February cohort 1,500 points (US$750, US$50 each).
February brought more sign-ups but kept a smaller share of them, and each active member spent less. That is a prompt to ask what was different in February. Perhaps the sign-ups came from a one-off event rather than regular customers. The table does not tell you the cause, only where to look.
What are the limits of cohort analysis?
Small cohorts give shaky percentages, so avoid reading too much into one month. Seasonality also matters, because a cohort that joins before a holiday may earn more simply because of the calendar. Compare cohorts from the same season where you can. For splitting members by recency, frequency and spend rather than join date, see RFM analysis. To judge the whole program, the guide to measuring loyalty program success covers the wider set of measures.
How Gimify fits in
Gimify records points per member across every outlet on one balance, so earned points give you a spend proxy for each person. The free plan covers one outlet with unlimited members. The spreadsheet work for cohorts is yours to do.
Questions owners ask
What is cohort analysis?
A method that splits customers into groups, called cohorts, by a shared starting event such as the month of joining. You then track each group over the same number of months and compare them side by side.
What is a cohort in a loyalty program?
All members who joined in the same period, usually a calendar month. Everyone who signed up in March is the March cohort.
Why not just look at overall retention?
An overall figure mixes old and new members. If a quiet month brings in many one-time visitors, the average drops even though your regulars have not changed. Cohorts keep the groups apart.
Do I need special software for cohort analysis?
No. A spreadsheet with one row per member, a join month and points earned per month is enough for a small shop. Count active members and total points for each cohort in each month.
How many members do I need for cohort analysis?
Enough that one person does not swing the percentage. With fewer than about 20 members in a cohort, treat the percentages as rough and compare two or three months together.
Read next
Start free
Gimify is free for one outlet with unlimited members and no time limit. Unlimited outlets are one flat monthly price.
See plans