How to Use Cohort Analysis and Prevent Churn

Customer acquisition in SaaS is like pursuing a love interest; you want to impress, pull out all the stops so they understand your offering, let them research the competition — but ultimately get him or her to invest in the relationship. It has to be mutual.

If acquisition is the pursuit and retention is the relationship, then churn is the divorce. Losing a customer is tough, gut-wrenching, and painful for any SaaS business. The time, money, and effort you’ve put into acquiring and retaining the customer is wasted.

As with any break up, it’s important to know why that person chose to part ways.

  • What were you doing wrong?
  • Why did they leave you?
  • How can you prevent this from happening again?

In business this is where you need cohort analysis.

What Is Cohort Analysis?

In relationships you don’t compare your whole entire roster of girlfriends or boyfriends as one collective; you never say “every single partner I’ve had left be because of XYZ.” You segment them individually or into groups. Cohort analysis works in exactly the same way.

Cohort analysis looks at groups of customers, opposed to your customer base as a whole. The analysis examines customer behavior and allows quick, valuable insight into customer retention and the overall health of a business. Cohort analysis helps business owners identify trends and take action.

There are two main types of cohort analysis for SaaS businesses: acquisition and behavioral. Acquisition cohorts are groups that allow you to investigate when they left your service.

Behavioral cohorts are groups that allow you to investigate why they left your service.

Using both these cohorts together gives insight into the who, what, why, and when of your churn rates.

Evaluate Your Churn Rate Using Cohort Analysis

Churn and cohort analysis go hand in hand. By having clear cohorts, you can begin to investigate your churn rate on a deeper level.

When do customers leave your business?

Acquisition cohorts allow you to look at the “when” part of your churn rate. Analyzing this information will show insight into the timeframe of churn from both newly acquired and long-standing customers.

By measuring this data, you’ll be able to implement a strategy to decreases your churn rate and predict when customers might be more inclined to leave your business. A very simple example could be if you’re seeing that many of your cohorts are churning in December, you could put out an incentivized offer to retain them during the holiday season.

If you start seeing new customers dropping off in weeks three to four of their subscription, you can start hypothesizing why this is.

  • Is the software too complex?
  • Are their needs being met?
  • Do they experience technical difficulties at this stage?

Knowing or hypothesizing this information will allow you to test changes and monitor your churn rate.

How do customers engage with your business?

Let’s say you notice that customers who are retaining with your SaaS model engage with app functions such as social or in-app mail.

  • Are you able to correlate these behaviors with customers who’ve dropped off earlier?
  • Is there a way to engage the fast-leavers earlier in the subscription sign up?

Find the features of your app that keep customers interested and engaged. Analyzing what keeps a customer interested is a key component of reducing churn.

Segment the right cohorts

Segmenting your cohorts into acquisition and behavioral is the first step into understanding how you can reduce churn. Delve deeper by further separating your cohorts.

  • Are there different actions people take to segment them further?
  • How about looking at age or gender?
  • Is there a correlation between job sector and churn?

Cohort analysis is about testing and investigating with an overall aim to reduce churn. Think about performing A/B tests to find out more about what works and what doesn’t. Don’t be afraid to try new things. When you find a method to improve retention using cohort analysis, apply the same model time and again.

Originally published on the now defunct Control blog.