You check your dashboard. Daily active users are up. Revenue is climbing. On the surface everything looks fine.
But aggregate numbers often hide the truth.
They act as a mask over the specific behaviors of different groups of people.
This is where cohort analysis comes in. It is a method of behavioral analytics that breaks data down into related groups rather than looking at all users as one unit.
Usually these groups, or cohorts, are defined by time. specifically the time they signed up or made their first purchase.
Defining the Term
#Think of a cohort like a graduating class or a vintage of wine.
A cohort analysis tracks the life cycle of a specific group of users starting from their acquisition date.
For example you might track the “January Cohort” separately from the “February Cohort.”
By isolating these groups you can see how long users stick around based on when they joined.
If you only look at total users you might miss that your older users are staying but every single new user you acquire leaves within two days.
Total growth can mask a leaky bucket.
Cohort vs. Aggregate Metrics
#Aggregate metrics are often vanity metrics. They make you feel good but do not help you make decisions.
If you have 10,000 users that sounds impressive.
But if you look at the cohorts you might see that the 1,000 users who joined last week have a retention rate of 5% while the users from a year ago had a retention rate of 50%.
Comparing cohorts tells you the trajectory of your product quality.
Aggregate data tells you the history of your marketing spend.
In a startup you need to know if the changes you are making today are actually improving the product experience.
Comparison is the only way to see this.
When to Use It
#You should use this analysis when trying to determine product-market fit or when measuring the impact of a specific change.
Here are a few scenarios where this is vital:
- Product Updates: You released a new onboarding flow in March. Does the March cohort retain better than the February cohort?
- Marketing Channels: You ran a Facebook ad campaign in April. Did those users stick around as long as the users who came from SEO in May?
- Pricing Changes: You raised prices in June. Did the churn rate for the June cohort spike compared to previous months?
If retention improves from one month’s cohort to the next you are building something sustainable.
If it declines you are burning cash to acquire users who do not value the product.
Questions to Ask Your Data
#Looking at a cohort chart, usually visualized as a triangle or layer cake, allows you to ask difficult questions.
It forces you to confront unknowns about your business model.
Why did the October cohort degrade faster than the September one? Was it a server outage? A bad support hire? A change in the market?
It removes the noise of top-line growth and forces you to look at the health of the relationship between you and your customer.
Keep building but verify that what you build makes people want to stay.