What is Cohort analysis?
Reviewed by CA Apurav GargLast reviewed
In short
Cohort analysis groups customers by when they first bought, usually by month, and tracks what each group does afterwards: how many come back, how often, and how much contribution margin they bring.
What a cohort is
A cohort is a group of customers who share a starting point. In ecommerce it is usually the month of the first order. Each cohort is then tracked month by month: how many customers ordered again, how many orders and how much revenue and contribution margin they produced.
Averages across all customers hide change. If retention is getting worse, new customers keep the averages steady for a while. Cohorts show the change as soon as it starts.
How to build one
- Export orders with a customer identifier (a customer ID, phone number or email), the order date and the order's net revenue. Remove cancelled, RTO and fully returned orders.
- Find each customer's first order date and assign them to that month's cohort.
- For each cohort, count the customers who ordered again in month 1, month 2 and so on after their first month, and add up the revenue and CM2 of those orders.
- Show the result as a triangle: one row per cohort, one column per month since the first order.
A spreadsheet pivot table is enough at first. Keep the definitions fixed, so cohorts from different months compare fairly.
What to look at
- Repeat rate curves: how fast each cohort drops and where it levels off. A curve that flattens means there is a core of loyal customers.
- Cohort LTV: cumulative CM2 per customer, by month. Compare it with the CAC paid to acquire that cohort to see the real payback.
- Differences between cohorts: a cohort from a sale, a new channel or a new product often behaves very differently. Tag cohorts by acquisition channel or first product to see why.
Limits
- Cohorts need your own customer data. Marketplaces usually do not share the buyer's identity, so most brands can build cohorts only for their website and app sales.
- Young cohorts have little history. Do not judge a cohort on one month.
- Customers who use different phone numbers or emails can be counted twice. Clean the identifier first.
Investors and lenders often ask for cohort tables, because they show whether growth comes from customers who stay.
Example in rupees
Customers grouped by the month of their first order on a brand's website. The table shows the share of each group that ordered again in each later month:
| First order in | Customers | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| July | 1,000 | 18% | 12% | 10% |
| August | 1,200 | 15% | 9% | – |
| September | 1,500 | 11% | – | – |
The September group is the largest but comes back least. Checking what changed shows that most September customers came from a heavy-discount campaign. The brand acquired more customers that month, but fewer of them stayed.
Measured in money: the July group of 1,000 customers brought ₹4,00,000 of CM2 on first orders and ₹2,80,000 more on repeat orders over three months, so ₹680 of CM2 per customer so far. The figures are illustrative.
Related terms
Work it out for your business
Want help applying this to your own accounts?
We handle bookkeeping, reconciliation, GST and TDS for D2C and marketplace sellers, and can walk you through how this works in your business.
General information only, not professional advice. Rules and rates change — check the official source before you act.