Cohort Analysis: A Step-by-Step Guide With an Example

A cohort analysis shows you how groups of customers behave over time, instead of lumping everything together. TL;DR: by grouping users by the moment they came on board, you see whether your retention is genuinely improving or whether growth is merely masking new sign-ups that leak straight back out of the back door. In this article you get a hands-on, step-by-step method to run a cohort analysis yourself, with an example you can replay right away.

Many B2B teams look at one big total: number of active accounts, total revenue, number of logins. That number can climb nicely for months while your product actually holds on to customers less and less well. A cohort analysis strips that illusion away.

What exactly is a cohort?

A cohort is a group of users or customers you bundle together based on a shared starting moment. The most common form is the acquisition cohort: everyone who became a customer in the same month forms one cohort. After that, you track each cohort separately over time.

The difference with a plain average lies in the fairness of the comparison. If you look at all customers together, you are comparing newcomers with people who have been around for two years. That skews everything. In a cohort analysis you compare apples with apples: the January cohort after three months against the February cohort after three months.

Broadly speaking, there are two kinds of cohorts:

• Acquisition cohorts: grouped by moment of entry. Good for measuring retention and recurring behaviour.
• Behavioural cohorts: grouped by an action people took, for example everyone who activated a specific feature. Good for seeing which behaviour correlates with sticking around longer.

For most growth questions you start with acquisition cohorts. They are the easiest to build and deliver insight immediately.

Step-by-step: how to run a cohort analysis

Step 1: pick one sharp question

Never start with the data, start with the question. A cohort analysis without a question turns into a dashboard nobody looks at. Good questions are, for example: do the customers we win this quarter stay longer than those from last year? Or: are we seeing that an improved onboarding lifts first-month retention?

The more concrete the question, the easier everything else gets. You immediately know which cohort you need and which event you have to measure.

Step 2: define your cohort

Determine the starting moment that puts customers into a cohort. In B2B that is usually the month an account was created or the contract started. Choose the period deliberately: monthly cohorts work well with a normal inflow, weekly cohorts only if you have enough volume per week to see reliable patterns.

Watch the group size. Cohorts of a handful of accounts produce erratic numbers you should not draw conclusions from. If you have little volume, use broader periods (a quarter instead of a month) so every cohort carries enough weight.

Step 3: choose the event that means “active”

Retention is measured against an event. What counts as “this customer is still active”? For a SaaS product that could be a login, a completed core action or a paid invoice. For a service provider it could be a repeat purchase or a renewed contract.

Pick the event that sits closest to real value. A login says little if someone logs in and clicks away again straight away. A core action that correlates with the value of your product says a lot more. This is the same logic as choosing a North Star metric that measures real customer value: measure the moment the customer gets value, not the surface-level movement around it.

Step 4: build the time grid

Now you set two axes against each other. Vertically you have the cohorts (entry period), horizontally the time since entry: month 0, month 1, month 2, and so on. In each cell you put the percentage of the cohort that was still active in that period.

Month 0 is almost always 100 percent, because that is the moment of entry itself. The interesting information sits in the cells after it: what percentage is still there after one month, after three months, after six months.

Step 5: fill in the retention table

Calculate the percentage per cohort and per period. A simple example to make it tangible. Suppose 100 accounts come on board in January. After month 1 there are 70 still active, after month 2 there are 55, after month 3 there are 48. Then the January cohort row looks like this: 100 percent, 70 percent, 55 percent, 48 percent.

Do the same for the February cohort, the March cohort, and so on. Line the numbers up in a table or colour them as a heatmap, so patterns jump out visually. The exact numbers above are an illustration, not a benchmark: your figures depend entirely on your product and market.

Step 6: read the table in two directions

This is the step most teams skip, and it is exactly where the insight sits.

• Read vertically (per column): compare the same period across cohorts. Does the March cohort hold on to more customers after three months than the January cohort? Then your retention is improving over time, often a sign that onboarding or the product got better.
• Read horizontally (per row): follow one cohort through time. Where is the biggest drop? If most customers fall away between month 0 and month 1, your problem sits in onboarding. If it only slips later, the problem lies in ongoing value or engagement.

The combination of both reading directions tells you not only whether there is a problem, but also where in the customer journey it sits.

A cohort analysis example that changes your conclusion

Suppose your total number of active accounts rises every month. Management is happy. Then you build a cohort analysis and see that every new cohort holds on to fewer customers after three months than the one before it. The total curve only rises because you keep bringing in more new accounts, not because you retain them better.

That is exactly the kind of distortion a headline number hides and a cohort analysis exposes. Without one, you would keep pushing harder on acquisition. With one, you know you first have to close the back door before more inflow makes any sense. That is a fundamentally different growth decision.

Common mistakes

• Cohorts that are too small: with few accounts per period the percentages are erratic. Use broader periods or wait until you have enough volume.
• Measuring the wrong event: a login that represents no value gives you an overly rosy picture. Measure the action that genuinely correlates with value.
• Reading in one direction only: looking only vertically or only horizontally leaves half the insight on the table.
• Attaching no decision: a cohort analysis that stays stuck in a dashboard without anyone taking action is wasted effort.

From cohort to growth decision

A cohort analysis is not a goal in itself. It is a diagnostic instrument inside a larger system. Improving retention touches onboarding, product, content and the way you qualify leads. Cohort data tells you where it leaks; the fix sits in the execution that follows.

That is why cohort analysis works best as part of an approach in which acquisition, conversion and retention are steered as one whole. That is the core of growth marketing as a system that connects loose tactics into one growth engine: you do not optimise a single channel, but the whole path from first contact to loyal customer. Anyone who steers only on inflow without looking at cohorts keeps filling a leaking bucket.

If you want to dig into what happens before the cohort, read how to attract and qualify the right accounts with demand generation. And to understand why the choice of your event matters so much, the piece on the North Star metric helps you pick the right action to measure against.

Want a team to set this up for you?

Cohort analysis is powerful, but only once it sits structurally in your growth management and leads to decisions. As a growth marketing agency, Customer Impact builds that measurable growth engine for B2B companies in the Benelux: from retention diagnosis to the interventions that genuinely improve the numbers. See what our approach to B2B growth marketing can do for you.

Ready to take your retention and growth seriously? Get in touch and we will look together at where the biggest win sits in your customer journey.