Step-by-Step Guide to Cohort Analysis & Reducing Churn Rate

Originally posted on 7/15/2022

Churn rate cohort analysis is a type of analysis that calculates and gives you the information you need to increase user retention. You're able to analyze user life cycles and identify when users churn or stop using your product. From there, you can learn about your users' behavior and investigate the reasons that lead to churn. Once you understand why users churn, you're ready to take action to reduce the rate of churn and keep people engaged with your product. Retention has become the metric most teams track today, and AI-assisted analysis makes cohort work faster—Amplitude AI Agents can surface insights and help you act on them.

Understanding the relationship between churn rate & cohort analysis

A cohort is a group of users who share a start point or a behavior, like the week they signed up. Churn rate cohort analysis is a method that tracks how each group's churn changes over time, so you can see when and why users leave.

Most companies see a small percentage of churn as healthy and normal. But a large number of users leaving your product is a cause for concern. Cohort analysis helps you reduce churn rates because it allows you to investigate why groups or segments (cohorts) of users leave your app. If you can figure out what's making people leave, you can take action to avoid high customer churn rates.

Cohort analysis also lets you understand which factors lead to high engagement and customer retention. If you focus on retention and providing value to existing customers, you'll often get a stronger return than if you focus purely on customer acquisition. Research from Bain & Company, cited by Harvard Business Review, found that increasing retention by 5% can raise profits 25% to 95%. You've already invested time, money, and effort into acquiring customers, so it makes sense to try and keep them on your platform instead of only focusing on reaching new users.

What are the main types of cohorts

Cohorts come in a few common types. Each one groups users differently, so you can look at churn from more than one angle.

• Acquisition cohorts: Group users by when they joined, like everyone who signed up in a given week or month.
• Behavioral cohorts: Group users by the actions they take or skip. Amplitude Product Analytics captures how users interact with your product across the full journey, so you can build these cohorts from real behavior.
• Demographic cohorts: Group users by traits like age, location, or device.
• Revenue or plan-based cohorts: Group users by what they pay or which plan they're on, such as free versus paid.

5-step guide to completing a cohort analysis that reduces churn

These steps show you how to conduct a cohort analysis that will help you identify when and why your users are leaving, so you can experiment with solutions to reduce churn.

1. Group users based on when they joined your app

Start your analysis by grouping users based on when they joined your app. "Date joined" is a type of acquisition cohort—you'll use behavioral cohorts later in your analysis. Rather than looking at churn rates across all users, this will let you see when users churn in relation to the customer lifecycle.

Create daily, weekly, or monthly cohorts, depending on your type of app and your product usage interval. For example, if you run a meditation or fitness app, it'll be useful to look at daily acquisition and usage. If it's an app people might use less frequently (like an investing app), you'll probably use monthly or weekly cohorts.

The time frame you include in your analysis will also depend on what you want to find out. You'll look at daily retention or churn over the course of a few weeks to see short-term changes. To spot longer-term trends, you'll look at the monthly retention rate over the course of a year.

When you've established the cohort you want to analyze, create a cohort table in Excel or use analytics software like Amplitude. In a cohort table, there are rows for each cohort and columns for each day, week, or month. "Day zero" is the day that the user joined the platform. In the cells, display either the churn rate or the retention rate.

If you were tracking cohorts by day, your “joined July 11” cohort might show a retention rate of 100% in the day zero column, 10.8% in the day one column, and 4.87% in the day two column. Try exploring this data for free by using Amplitude’s self-service demo, then navigating to this example cohort table.

If you were tracking cohorts by day, your "joined July 11" cohort might show a retention rate of 100% in the day zero column, 10.8% in the day one column, and 4.87% in the day two column. Try exploring this data for free by using Amplitude's self-service demo, then navigating to this example cohort table.

Here's how to read a cohort table. Each row is one cohort, grouped by when they joined. Each column is a period since joining, and each cell shows the retention rate for that cohort at that age.

Read across a row to see how one cohort's retention falls over time. Read down a column to compare different cohorts at the same age—for example, how each group looks two days after joining. A spreadsheet handles acquisition-date views like this one. A digital analytics platform like Amplitude also tracks behavioral cohorts across the full journey, so you can group users by what they do, not just when they joined.

2. Identify when users in each cohort churned

If you can identify the point at which users churn, you can spot patterns in the user journey. You'll be able to investigate further to discover why they churn at that moment.

Take a look at your cohort analysis chart and find points when many users churn. To visualize the retention curve and easily spot when people drop off, convert your cohort table into a chart.

When you set up a Retention Analysis chart in Amplitude, you can select “Retention View” to see a graph of the retention curve. Try it yourself using Amplitude’s self-service demo.

When you set up a Retention Analysis chart in Amplitude, you can select "Retention View" to see a graph of the retention curve. Try it yourself using Amplitude's self-service demo.

Oftentimes, product and growth teams will make a list of "problem" moments—points in the user lifecycle when users leave, for example, after 10 days or after the first month.

3. Analyze other cohort types to investigate churn further

After you've looked at churn in relation to the acquisition date, creating groups of users based on their behaviors, acquisition channels, or demographics will help you understand them further. It brings you one step closer to discovering the reasons why users leave.

Use behavioral cohorts to group users by actions they have or haven't taken. You might look at users who read reviews before purchasing or those who make an in-app purchase within the first three days of joining the platform. In each case, you'll want to compare the cohort who does take the action with those who don't to see if there are any links to churn. Once you spot a behavioral cohort that's likely to churn, Amplitude Activation can sync that cohort into your campaign and engagement tools, so you can act on the insight in one workflow.

In addition to using behavioral cohorts, group users by type of acquisition to identify which channels lead to churn and when. For example, you might notice that users from paid ads churn faster than those who convert from your newsletter. Another way of segmenting users is by demographics—look for patterns among different age groups or locations.

For consistency, look at different cohort types in the same time frame as the acquisition cohort analysis you did. Investigate one customer cohort at a time, so you can clearly see which factors are contributing to churn. You'll start to draw conclusions about why cohorts vary in the next step.

4. Hypothesize why churn happened for each cohort

Now that you've gathered data on different cohorts, you can theorize about why people are churning. By building and testing out hypotheses, you can discover what's really causing churn.

Based on the "problem moments" you spotted from your investigation into when people churn, along with your analysis of the different behavioral cohorts, pick out the cohorts with the highest churn rates. Then brainstorm what could be causing churn.

For example, from the acquisition date cohorts:

• Seasonal changes impact app usage—a lot of users joined during the holidays, but they churn in February because they don't have enough free time to use the app.
• Users drop off after two days because the onboarding process is too long and complicated.

From the behavioral cohorts:

• Users who sign up for the "basic" plan churn because they don't have access to enough features.
• Users who don't activate reminder notifications end up churning because they forget to use the app and don't get value from it.

From the acquisition channel cohorts:

• Users who sign up from the Instagram ad churn quickly because the ad doesn't accurately portray the product, so there's a discord between customer expectations and the product.

5. Test your hypotheses and experiment with solutions

Before implementing any product changes, you'll need to check whether your hypotheses are correct. Be prepared for the fact that many different factors contribute to churn, so this stage will involve some experimentation.

First, check your hypotheses by conducting A/B testing. To test out your example hypotheses, you could:

• Simplify your onboarding process.
• Add more features to your "basic" subscription tier.
• Update your marketing campaigns to set more accurate customer expectations.

Experiment with different solutions to optimize your platform. For example, simplifying the onboarding process might stop users from dropping off after two days, but another solution could be to give users the option to skip onboarding altogether. If the adjustments you make reduce churn, you're ready to roll them out to your whole app. If none of your hypotheses are correct or none of your solutions work, go back a step and continue your cohort churn analysis.

Cohort analysis FAQ

What is churn cohort analysis?

Churn cohort analysis is a method for grouping users into cohorts and tracking how each group's churn changes over time. It shows you when users leave and helps you investigate why. From there, you can take action to keep more of them engaged.

How do you analyze cohort results?

Read across a single cohort to see how its retention or churn changes over time, then compare cohorts at the same age to spot patterns. Next, compare acquisition, behavioral, and other cohort types to find the factors tied to churn. Investigate one cohort at a time so you can see what's contributing to churn.

What is the difference between segmentation and cohort analysis?

Segmentation groups users by shared traits, like demographics or acquisition channel. Cohort analysis groups users by a shared start point or behavior, then tracks how those groups change over time. For churn work, cohort analysis adds the time dimension that shows when users leave.

How do you do cohort retention analysis?

Group users by when they joined, then build a cohort table or retention chart that shows how each group's retention changes over the days, weeks, or months after joining. Read across a row to follow one cohort over time, and down a column to compare cohorts at the same age. Then dig into behavioral cohorts to understand what drives the differences.

Continue your customer churn analysis

Cohort analysis is a process you will go through more than once. Customer churn rates change over time, so keep tracking cohorts and regularly conducting cohort analysis to spot patterns in user behavior—that way, you can take action to keep your customer retention rates high. Treat it as an ongoing, retention-first habit rather than a one-time check.

Try conducting your own churn rate cohort analysis today using Amplitude's free starter plan, or check out our Mastering Retention playbook to learn more about fixing a leaky bucket.

References

• The Value of Keeping the Right Customers. Harvard Business Review.
• User onboarding examples worth stealing. Appcues.