I'm Brett Slatkin and this is where I write about programming and related topics. Check out my favorite posts if you're new to this site. You can also contact me here or view my projects.

18 November 2012

Why cohort analysis?

(from Cohort Visualizer, released today)

A cohort is a group of people who share a common characteristic or experience within a defined period (Wikipedia). It's used a lot in health to track how different groups of patients respond to disease and medication. It can be used in business to track progress in the funnel.

For software and websites, cohorts are useful because they let you measure the impact of your product changes over time. Simple example: Using cohorts you can see the conversion rate of new users from two months ago and compare it to new users of today. Ideally, this would let you judge if your software is getting better over time, users are getting happier, etc. Sometimes you'll see that things are getting worse.

Cohort analysis can apply to more than just users. For example, you could treat a set of articles on a blog as the source dataset; the levels of traffic or reshares could be mutually exclusive states; and you could treat common tags as a group type, or author as a group type. Cohort Visualizer would let you drill down and compare all of those groupings pretty easily.

One of the most interesting things about cohort analysis is the graphs change over time. If you take a snapshot of your cohort data today and then take another snapshot two months from now, you'll see that the older cohort bars have changed. This happens because users who signed up two months ago remain active and continue to make progress in your funnel over extended periods. It's useful to save your cohort datasets after you collect them, so you can compare to the past.

Other things to read to understand the motivation and method behind this:
© 2009-2016 Brett Slatkin