Multi-Touch Attribution & Marketing Mix Modeling (MMM)Playbook3 min readUpdated September 2026

Reading Your Cohort Tables to Find Expansion Revenue

A cohort table looks intimidating the first time you see one, rows of signup months against columns of revenue over time, but it answers one specific question well: which customers, grouped by when they joined, actually grow their spend with you over time, and which shrink. Here's how to actually read one, worked through with a simple example.

Set Up the Table Correctly First

Group customers by the month they signed up, not by the month you're currently reporting on. Each row is a cohort; each column is a month since signup, showing that cohort's revenue relative to its starting point. If your table groups by calendar month instead of months-since-signup, you're comparing cohorts at different points in their lifecycle, which makes the whole exercise misleading. Getting this structure right the first time saves you from rebuilding the table later once someone notices the numbers don't line up.

A Worked Example: Two Cohorts, Two Different Stories

Say your January cohort's revenue is still at 100 percent of its starting value six months later, while your March cohort has grown to 140 percent over the same window. That gap is the entire point of the exercise: something changed between January and March, in your onboarding, your pricing, or the type of customer you were acquiring, that made March's cohort expand instead of stay flat. Without the table laid out this way, that 40-point difference stays invisible, buried inside a single blended revenue number that looks fine on its own.

How do you work backward to find the cause?

Once you spot a gap like that, check what actually changed: did you launch an upsell prompt in February, change your onboarding flow, or start acquiring through a different channel that brings in customers with a different profile? Cross-reference the cohort data against your own product and marketing changelog rather than guessing, since the cause is almost always something you can point to on a specific date, not a mystery that needs a fresh investigation each time.

Segment Cohorts by Acquisition Channel, Not Just Signup Date

A single monthly cohort can hide two very different stories if it blends customers from several channels. Splitting the March cohort by whether customers came from paid search, referral, or a partner integration might show that the referral-sourced customers are the ones actually driving the expansion, while the paid search customers are flat. That distinction changes where you'd invest acquisition budget, and it's the kind of finding a blended cohort table can never surface on its own, however carefully you read it.

Turn the Finding Into a Marketing Action

Once you know which segment expands and why, the marketing response is specific, not general: double down on the acquisition channel that brings in expanding customers, adjust messaging to attract more of that profile, and consider whether the onboarding change that helped March's cohort should roll out to everyone, or whether it only worked for that specific channel's customers. Put a number on the action too, a target for how much of next quarter's acquisition budget shifts toward the channel that's proven it can produce expansion, rather than leaving the finding as an interesting observation nobody acts on.

To read a cohort table for expansion, follow this sequence:

  1. Group customers by signup month and show each cohort's revenue by months since signup, relative to its starting point.
  2. Compare cohorts at the same age to find one that expanded while another stayed flat.
  3. Check your product and marketing changelog for what changed around the dates that separate the two cohorts.
  4. Split the standout cohort by acquisition channel to see which source is driving the expansion.
  5. Shift budget or onboarding changes toward what worked, after confirming that a few large accounts are not skewing a small cohort.

What is the most common misread of a cohort table?

It's tempting to credit a single marketing change for a cohort's expansion when several things shifted around the same time. Before reallocating budget based on one gap, check whether the expanding cohort also happened to include a couple of unusually large accounts that could be skewing the average on their own. A cohort table built from a small customer base is more sensitive to a handful of outliers than one that gets averaged across hundreds of accounts, so treat a striking result from a small cohort as a hypothesis to test further, not a settled conclusion. Waiting one more month of data before committing budget is usually a cheap way to avoid chasing a fluke.

As a decision rule, ask whether the pattern holds in at least two consecutive cohorts before changing budget. One strong cohort is a lead worth following; two or three in a row is evidence. In the meantime, you can run a small test, such as shifting a modest slice of spend to the promising channel for a quarter while the rest of the plan continues unchanged. That keeps the cost of being wrong low and turns a striking table into a decision you can defend if someone asks how you reached it.

Executive Capability Standard

What Good Looks Like

A working cohort analysis practice can name, for the most recent quarter, which specific customer segment expanded and what marketing or product change is the likely cause.

Building The Capability (5-Stage Skill Ladder)

1. Learn:Build a basic cohort table in a spreadsheet using your last twelve months of signup and revenue data, grouped by months since signup.
2. Do Manually:Compare two cohorts by hand and cross-reference the gap against your product and marketing changelog for that period.
3. Delegate:Have a data-minded marketing or finance hire own the monthly cohort review and flag findings to the team.
4. Automate:Use analytics tooling that segments cohorts by acquisition channel automatically, so the channel-level story doesn't require manual slicing every month.
5. Buy:Bring in a data or growth analyst if cohort patterns keep surfacing questions your internal team doesn't have time to dig into.

How to Get Started

Frequently Asked Questions

How often should I review cohort tables?

Monthly is enough for most small and mid-sized companies. Reviewing more often than that usually just shows noise, since a cohort needs a few months of data before its expansion or contraction pattern becomes clear.

What if every cohort looks roughly the same?

That's useful information too. It means whatever's driving expansion or contraction isn't tied to a specific change you've made, and the pattern is more likely structural, tied to your product category or pricing model as a whole.

Do I need special software to build cohort tables?

A spreadsheet works fine for a company with a manageable number of customers. Dedicated analytics tools help once you have enough volume and channel variety that manual segmentation becomes tedious, but they're not required to get started.

About the numbers

This guide doesn't quote a sourced benchmark. Figures in it are estimates or general guidance, so check them against your own numbers.

Related Guides