Cohort analysis: what it is and how to do it step by step
FLYDE Glossary · November 9, 2026 · 4 min read
Cohort analysis groups customers by when they first did something, usually their first purchase, and follows how each group behaves over time. It tells you whether retention is improving or getting worse, and why.
What a cohort is
A cohort is a group of customers who share the same starting point. The most common is the acquisition cohort: everyone who bought for the first time in January is the January cohort; February's first-time buyers are the February cohort, and so on.
Unlike looking at the total, it separates two things the average mixes up: how many new customers arrive and how long they stay. If a campaign brings in lots of customers who never come back, total sales can rise while real retention collapses.
How to do a cohort analysis
- Choose the entry event. Almost always the first purchase. For subscriptions, the sign-up.
- Choose the period. Months for most businesses; weeks if purchases are very frequent.
- Choose the metric. Share of customers who buy again (retention), revenue per customer or orders per customer.
- Build the table. One row per cohort and one column per period since entry (month 0, month 1, month 2...). The result is a triangle: older cohorts have more columns than recent ones.
An example (illustrative), showing the share of each cohort that buys again in each month:
| Cohort | Month 0 | Month 1 | Month 2 | Month 3 |
|---|---|---|---|---|
| January | 100% | 24% | 18% | 15% |
| February | 100% | 27% | 20% | |
| March | 100% | 31% |
Each new cohort repeats more than the previous one in month 1: something that changed between January and March (onboarding, product, acquisition channel) is working.
How to read the table
- Across rows: how each cohort fades. A sharp drop from month 0 to month 1 points to the first experience.
- Down columns: whether new cohorts retain better or worse than old ones at the same point in their life.
- Diagonally: calendar effects. If every cohort rises at once in November, that is Black Friday or El Buen Fin, not better retention.
Common mistakes
- Mixing channels: a January cohort acquired with heavy discounts is nothing like one acquired through referrals. Split them by acquisition channel.
- Counting the same customer twice because they bought in store and online with different details. Without merging their records, retention looks lower than it is.
- Drawing conclusions from very small cohorts: with 40 customers, a couple of purchases move the percentage by several points.
Cohort analysis in FLYDE
Cohort analysis is only reliable if each customer is one person even when they buy in store, online and by phone. FLYDE merges those records into a single profile, so cohorts are built on real customers rather than loose emails or cards. With FLYDE Brain you can ask in plain language: "show me the 3-month retention of the customers acquired each month this year".
Frequently asked questions
What is a customer cohort? A group of customers who share the same starting point, usually the month of their first purchase. The group is followed over time to see how many buy again and how much they spend.
What is cohort analysis used for? To measure real retention, separate from the volume of new customers; to compare the quality of customers each channel or campaign brings; and to estimate what a customer will be worth over time.
What is the difference between cohorts and segments? A cohort groups customers by when the relationship started and is followed over time; a segment groups them by traits or behaviour at a given moment, as in an RFM analysis. They work well together: cohorts by channel or by segment.
Can you do cohort analysis in Excel? Yes, with a pivot table: rows with each customer's first-purchase month, columns with months elapsed and, as the value, the number of distinct customers who buy. It gets complicated when customer data comes from several sources that are not merged.