Jake Ballinger

Jake Ballinger - FP&A Writer

September 2022 - min to read

Cohort Analysis: Tell the Secret Story Behind the Numbers

Cohort Analysis: Tell the Secret Story Behind the Numbers

We can all agree that discovering better insights is a good thing.

And everybody wants to be better at forecasting.

(Of course, how you present your insights and forecasts also matters.)

In this article, you'll learn about a tactic that does all those things: cohort analysis.

It's even better than that, actually: it tells you more about how specific groups of your customers behave. 

So if you're interested in learning how to create more ideal customers, keep reading. 

Key Takeaways

  • Cohorts are customer segments grouped by a shared trait.
  • We perform cohort analysis to learn more about customer behavior so that we can strategically encourage the behaviors we want to see.
  • Acquisition Cohorts are grouped by join date/time as customers.
  • Behavioral Cohorts are clustered by product use behaviors. 
  • It's best to begin with the end in mind before starting cohort analysis. 

Contents

  1. What is Cohort Analysis?
  2. Acquisition Cohorts vs. Behavioral Cohorts
  3. 6 Steps to Perform (A Simple) Cohort Analysis
  4. What Doesn't Cohort Analysis Tell Us?

What is Cohort Analysis?

Cohort Analysis is when you examine a group of customers who share a specific trait.

Companies perform cohort analysis to better understand their revenue. 

How does this work?

Because cohort analysis lets companies study a specific group of customers. This means they can identify patterns in that group's behavior. 

An example: understanding why a group churns at a higher rate than average.

Or how they can do better expansion into a specific market. 

When we understand our customers, we can better retain and acquire more customers like them.

(This is sometimes called "customer retention analysis," by the way.) 

But first, let's define a cohort.

What is a Cohort? 

A cohort is a group of customers who share a common characteristic.

This could be: 

  • Location  
  • Size 
  • Join time
  • Industry 
  • Plan type
  • Time with the company
  • Join circumstance

...or any number of factors. 

We can account for these in Cube using dimensions

Why Perform Cohort Analysis?

We perform cohort analysis to understand how different customer segments behave.

Splitting customers into cohorts helps you better understand their behavior as a group. This means you can also learn how to increase or decrease certain segments of customers. 

In other words, it helps you identify trends. 

These trends could be (and this list is not exhaustive):

  • Churn rates
  • Usage patterns
  • Expansion rates

It's important to identify trends at the cohort level because they could get washed out at a higher level.

(Or they could be magnified.)

Here's an example: 

Company A loses 5% of their monthly revenue to churn. They want to know why, so they perform a cohort analysis.

They then find that 80% of their churn comes from companies in Cohort B, who are all SMBs.

So now they know to focus their retention efforts on SMBs.

Telling the Narrative Behind the Numbers

Cohort analysis lets you understand your revenue from a few different perspectives.

This means that, when you're going to tell the story of the company's numbers, you can use cohort analysis to stay honest about the numbers while also giving stakeholders a more clear picture of the reality.

For example, say your company is seeing high churn rates.

You have a hypothesis that the free trial you offered last quarter is the cause of those churn rates, so you perform a cohort analysis and strip out customers who joined as part of that free trial.

And then you see that your churn rates are actually on par for what you'd predicted.

Now the numbers look better and you've identified the cause of the problem.

Keep What Works and Drop What Doesn't

On the flip side, say you have a specific vertical that's doing astoundingly well. 

You want to replicate that success.

...right? 

So you analyze that cohort and look for the factors that led to the current circumstances. 

Then, when you're doing scenario planning and budgeting, you can try to reproduce those factors. 

You can also do this when looking at cohorts that don't perform well. 

And then you can make the necessary adjustments in your scenarios and budgets. 

Acquisition Cohorts vs. Behavioral Cohorts

There are two big ways to split customers into cohorts: acquisition cohorts and behavioral cohorts. 

Acquisition cohorts are cohorts based on join time. This kind of cohort helps you identify the who and when.

Here are some examples: 

  • Joined within 3 months of some big event
  • Joined when we offered a free trial
  • Joined in Q3 of last year

This kind of cohort is useful for when you want to determine at what point in the lifecycle your customers tend to churn. 

Behavioral cohorts are cohorts based on actions and behavior. This kind of cohort helps you identify the why. 

Here are some examples: 

  • Logs in twice a week
  • Doesn't use feature x
  • ≤ 200 FTEs

This kind of cohort is helpful when you have a pattern you want to look for or for when you've identified some trait that isn't based on join time.

6 Steps to Perform (A Simple) Cohort Analysis

In this section, we'll walk you through a simple cohort analysis example.

Just like how there are plenty of ways to fry an egg, this isn't the only way to perform cohort analysis. 

It's not even the most comprehensive way.

But it'll get you started.

Step 1: Begin With the End in Mind

If you're anything like us, you could spend all day looking through data. And while that's valuable for what it could uncover, it might not be the best use of time. 

...especially if you're trying to pull together a few KPI's by EOD.

So it's time to begin with the end in mind. 

This means: decide which metrics you're looking to calculate.

Step 2: Pull the Raw Data

Your next step is to pull, gather, cleanse, and organize your raw data.

The data you're looking at might all be in your CRM or ERP. More than likely, though, you're also going to want to look at the data that lives wherever you keep your product use metrics.

(We know this step isn't fun. That's why we built Cube.)

It's time for the next step once you've organized that data.

Step 3: Identify Your Dimensions and Create Cohort Identifiers

Identifiers are dimensions in your spreadsheet. 

  • Location  
  • Size 
  • Join time
  • Industry 
  • Plan type
  • Time with the company
  • Join circumstance

...and so on. 

What you want to do is create discrete cohorts as a combination of dimensions. 

For example, Cohort A could be companies who joined in the last 90 days, are based in the Pacific Northwest, and are on your lowest pricing tier.

Step 4: Slice and Dice your Cohorts

This is the fun step: pivoting and looking at how the data tells you different stories. 

Now that you've identified your cohorts, you can look at the data in a couple different ways. 

For the above example, let's say you notice that these companies have a really low gross revenue retention rate. 

So you dig into the data a little and you notice that half of them started at a higher subscription tier and have downgraded. Interesting. 

You're immediately curious about two things, now: 

(1) What happens to those companies in the future? Do they become logo churn, or do they stay with you? 

(2) What is it about that group of companies who churned? 

To answer either of those, you create a new cohort, look at the data, and piece together a story. 

Step 5: Gut Check (Analytical Skills)

The data tells you a story: does it make sense? 

This is a gut check phase, and it's important to trust your intuition. 

For example, if you notice that all the companies who signed up during your free trial period 6 months ago have upgraded to a paid plan...that's suspicious. Free trials are notorious for having higher churn rates, after all. 

Step 6: Present Your Findings & Suggestions

Now it's time to present the story of the data.

We're not going to tell you how to present your data to your stakeholders.

However, we did want to give you a strategic tip we've found helpful here at Cube. 

It's called inversion. It works like this: 

Say your goal is to decrease revenue churn. Instead of focusing on strategies to prevent churn, why don't you identify the factors that lead to revenue churn and then stop doing those things?

It's a simple model and it produces simpler results.

Limitations

Cohort analysis is a fantastic tool. But like every tool, it has its limitations.

Cohort Analysis won't tell you the underlying cause behind a trend.

For one, cohort analysis helps people find trends in the data. But it's only as good as the imagination.

In Business B's example earlier, knowing that they need to talk with Customer Success isn't helpful information if it turns out that Finance and Customer Success don't have a good rapport.

In Business A's example, the analyst might have a blind spot or simply not think about their onboarding process as influencing churn.

So Business A could spend time perfecting their marketing and sales qualifications and not see a large change in their overall retention.

Cohort Analysis takes time.

If you don't start with a specific problem you want to solve, it's possible to spend all day searching through data.

Now It's Your Turn

Now you know enough to get started with cohort analysis. 

So we're curious...which part of the business are you looking to optimize? 

Are you going to dive deep into your 3-month acquisition cohort churn? Or are you curious about performing better with SMBs? 

Share this on LinkedIn and tag us to keep the conversation going.