"Our repeat rate is 30%" sounds like an answer but tells you almost nothing. Cohort analysis shows whether customers acquired in different months actually behave better or worse over time, which is the only way to know if retention work is paying off.
What a cohort is
A cohort is a group of customers who made their first purchase in the same period, usually a month. You follow each group over the months after that first order and record how many of them buy again.
How to build a simple cohort table
- Export orders with customer ID, order date and order value.
- Label each customer with the month of their first order.
- For every later month, count how many customers from each cohort placed an order (and, optionally, how much revenue they generated).
- Divide by the cohort size to get a percentage.
An illustrative result (made-up numbers to show the shape, not benchmarks):
| First-order month | Customers | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| January | 1,000 | 18% | 12% | 10% |
| February | 1,100 | 20% | 14% | 12% |
| March | 1,200 | 24% | 17% | — |
Read it row by row. Here, each newer cohort returns more often than the one before it. That pattern, not the blended average, is evidence that something you changed is working.
Read the shape of the curve
- Steep early drop, then flat: normal. The customers who stay after the first reorder tend to keep ordering.
- A bump at a specific month: often your product's replenishment point. If the bump arrives later than the product's days of supply, your reminder timing is probably late.
- Continuous decline with no floor: customers are not forming a habit. Look at onboarding, product education and subscription uptake.
Split subscribers from one-time buyers
Subscribers and one-time buyers behave very differently, so analyze them separately. A healthy blended number can hide a subscriber base that is shrinking, or a one-time group that never returns. The same applies by product and by acquisition channel.
A simple LTV formula
LTV ≈ average order value × gross margin % × expected number of orders per customer
Use observed cohort data for the number of orders, and be conservative about extrapolating. Compare the result with your customer acquisition cost to get LTV : CAC and the payback period — how many months of gross profit it takes to recover the acquisition spend.
Common mistakes
- Using blended averages that mix old and new customers.
- Counting revenue instead of gross profit when judging payback.
- Comparing a young cohort against an older one before it has had time to reorder.
- Ignoring seasonality and promotions that pull in one-time deal seekers.
Once you can see cohorts, you can test changes properly. Pair this with the replenishment flow and check the whole system against the retention playbook.
Tehseen Arbab
Retention and lifecycle marketer for DTC supplement and coffee brands. I led retention for a supplement brand whose monthly run-rate grew from about $500K to $700K, and I work in Klaviyo, Skio, Recharge, Meta Ads and Shopify.