We will explore and then recreate the three most common types:

But first… an important question…

“Cohort” is a fancy ass word for a “group” of users or customers with similar characteristics. The most common filter to create a cohort is by start date (e.g., new customers in July of 2024).

Why? Because it puts everyone on equal footing in terms of how long they've had to adopt the product and either stick around, expand, or bounce.

Other common cohort cuts include:

Both of these cohort cuts are frequently referenced during public company earnings calls for SaaS and fintech companies, as it provides context as to how the base is maturing.

Furthermore, cohort data is typically linked to revenue but can also be linked to activity metrics like logins, transactions, or downloads.

There are three common variants of cohort charts.

Keep in mind that if you run a consumer business where it’s difficult (or even impossible) to “expand” a customer (like Netflix or Strava), you will look at cohorts on a gross retention basis.

On the other hand, if you run a business that is predicated on landing and expanding a customer over time (e.g., Notion, Snowflake), you will be able to look at cohorts on both a gross and net dollar retention basis.

Let’s take a look:

This chart has dates on both the left and top. The left is the cohort we are measuring, organized by start date. Next to it are the absolute number of users or customers that originate from that start date bucket.

Across the top you’ll see periods. These are usually in months, but can be in hours, days, or weeks, depending on how frequently you expect someone to utilize your product.

So in essence, you have two date measurements on the same chart. One is fixed (start date… you never get a “new start date”) and the other is relative.

Within the box we include the % retained from one period to the next. If the box is blank, that means it’s in the future and hasn’t happened yet.

Reviewing the data, you can see that only about a quarter of the users in each cohort even make it out of the first period. And only ~5% are around after 24 months.

That’s why it’s important to really hone in on the user activity you are trying to measure and how it corresponds to your business model’s desired outcome.

The color coding is your typical heat map to better visualize which cohorts are outperforming relative to their peers.

Ok, moving on, here is perhaps the most important cohort chart to understand…

(Keep reading for breakdowns on cohort charts 2 and 3, as well as a link to the TEMPLATE to build your own)

Here’s the gist. Think of it as the rate of decay. Each cohort starts at 100%. And then you measure the rate of fall off.

The goal is to “flatten” the downward curve as fast as possible. Get to horizontal.

In consumer businesses it’s normal for companies to bleed subscribers in the first three months. Users might be trying something new, and don’t know if they like it yet. But ideally it becomes habit forming, and they decide to stick around. That’s when the curve flattens out, as the tire kickers fall off and the true fans stick around.

Let’s look at some examples:

What you’ll notice is there is always some decay in the long tail for streaming services. This is typical for content based businesses who’s users will binge through all the shows possible, and retire their subscription after a show ends. For example, there was a steep drop off in HBO Max’s subscriber base after Game of Thrones ended. That’s why it’s so important for them to keep investing in new content, and for Netflix to keep giving Jason Bateman roles that are all kinda similar to Jason Bateman in Ozark.

Blue Apron’s data is total ass. Not only do they have a very rapid fall off in the first few months, but it never flattens out. In fact, under 15% of their customers stick around for a full year, which means they are a subscription service that needs to find nearly 85% of their subscribers each year. At that point their business model is closer to transactional than recurring.

Notice something happening at the end points? The curve looks like an awkward smile. It swings back around. That means the company is reactivating users at a rate faster than they are losing them. Their retention actually gets BETTER over time. The smile retention curve is rare, and something investors salivate over. It’s a sign of STRONG PMF.

You build this chart when you’re in a business that expects customer spend to go up over time. It’s essentially a graphical representation of your net dollar retention rate.

Below is a chart showing Slack ARR by cohort. Each colored “layer” of the blob cake represents a year that customers started. And the X axis measures the absolute time that’s passed. You can see that the blobs for each color take up an increasing amount of surface area.

What you want to see is:

What you can see in the example above is a falloff starting sometime around April 2016. The glob goes down when a customer churns entirely, ceasing to contribute to the cohort, or shrinks, due to downsell or partial product churn.

Those are the three most common types of charts companies use to visualize cohort retention. As a reminder, you can reswizzle the first two to be driven by dollars or by absolute customer count. Hell, you can even do them based on activity metrics like downloads and logins. Go wild! Speaking of that…

You can build your own cohort charts and tables using this link.

Version up and put in your own data.

Wishing you a cohort retention chart that smiles back,

CJ