Cohort analysis made simple

Your overall churn rate is an average across every member you have: the brand-new ones still deciding whether you are worth it, and the long-timers who will stay for years. Because it mixes those groups together, it can hide what is really going on. If your onboarding gets worse but a large group of loyal members stays put, the average barely moves. If you have a big month of sign-ups, churn can rise simply because you have more new members, even though nothing about your membership has changed.

Cohort analysis solves this by comparing like with like. It sounds technical, but at heart it is a simple spreadsheet, and it is one of the most useful things a membership owner can build.

What a cohort is

A cohort is a group of members who share a starting point. The most common choice is join month: everyone who joined in March is the March cohort, everyone who joined in April is the April cohort, and so on.

You then follow each cohort over time, recording how many of them are still paying one month after joining, two months after, three months after and onward. Because each cohort is measured from its own start, you can compare the March joiners at month three with the April joiners at month three, regardless of what else was happening on the calendar.

Building a cohort grid

You need a list of members with their join date and, for those who have left, their end date. Most membership software can export this.

1. Add a column for each member's join month.
2. Add a column for how many full months they stayed. For current members, count up to today.
3. Build a grid with one row per join month and columns for month 0, month 1, month 2 and onward.
4. In each cell, count the members from that row's cohort who were still paying at that month. Month 0 is simply the number who joined.
5. Make a second copy of the grid where each cell is divided by that row's month 0 figure, so you can see the proportion retained.

Cells that lie in the future stay empty, which gives the grid its familiar staircase shape: older cohorts have long rows, newer cohorts short ones.

Reading the grid

There are three ways to read a cohort grid, and each answers a different question.

Along a row

Reading across one cohort shows its retention curve: how quickly that group left over time. Most curves drop more steeply in the first few months and then flatten. The point where the curve flattens tells you when members have settled in. If yours never flattens, members are leaving at a steady pace no matter how long they have stayed, which suggests a problem with ongoing value rather than onboarding.

Down a column

Reading down a column compares cohorts at the same age. Are more of your recent joiners still here at month two than your earlier joiners were at month two? If so, something you changed is working. This is exactly the comparison your overall churn rate cannot make.

Around a change

Mark the cohorts that joined before and after a change, such as a new welcome sequence, a price rise or a new course. Then compare their early columns.

A worked example

A hypothetical watercolor painting membership introduced a structured beginner path for new members at the start of May. Here are illustrative cohorts, showing how many members were still paying after one, two and three months:

• February joiners (100): 78 after one month, 66 after two, 60 after three.
• March joiners (80): 62 after one month, 52 after two, 48 after three.
• April joiners (120): 92 after one month, 79 after two, 72 after three.
• May joiners (100): 86 after one month, 78 after two, 74 after three.
• June joiners (90): 78 after one month, 70 after two.

Converted to proportions, the February, March and April cohorts all kept roughly 77% after one month and 60% after three. The May cohort kept 86% after one month and 74% after three, and the June cohort is tracking similarly. The beginner path appears to have made a real difference in the first three months, and the grid shows it cleanly, without being muddied by the long-standing members who make up most of the membership.

Notice what the overall churn rate would have shown: a modest dip, spread across all members, easy to dismiss as noise.

Cohorts beyond join month

Once you are comfortable with the grid, you can group members in other ways:

• By how they found you, to see whether members from search, referrals or a partner stay longer.
• By plan, to compare monthly and annual members, or each level.
• By whether they joined through a promotion.
• By whether they completed a key first step, such as finishing an introductory course.

These cohorts connect retention to decisions you control. Paired with member lifetime value, they show which kinds of members are worth the most over time.

Keeping it honest

Small cohorts are noisy. If only 15 people joined in a month, two cancellations move that row sharply. Where monthly cohorts are small, group by quarter instead. Also make sure your definition of a paying member matches the one you use for calculating churn rate, especially for paused members and failed payments, or your cohort grid and churn figures will disagree.

Your first cohort grid

1. Export members with join dates and end dates.
2. Build the grid for the last twelve months of joiners, counting members still paying at each month.
3. Convert to proportions and look down the month one and month three columns.
4. Mark any changes you made to pricing, onboarding or content, and see whether the cohorts after them differ.
5. Update the grid monthly; it takes only a few minutes once the structure exists.