Cohort Analysis - Chapter 3

What is Cohort Analysis?

In marketing speak, a cohort is a group of customers who all joined at the same time. “Joining” in this case can mean either subscribing to your email list or making a first purchase, and “at the same time” can mean in the same week, month, or quarter.

Regardless of how exactly you define the cohort, the key is that you are following a single group of customers and watching how their behavior changes over time. It’s similar to the kind of longitudinal tracking that’s common in medical research: A well-known example is The Framingham Heart Study, a cardiovascular research project that has been taking place in Framingham, MA since 1948.

Most importantly, in a Cohort Analysis, you are comparing how different cohorts behaved at a comparable point in their lifecycle.


Why cohorts?

So how would our retailer move from an ARPU approach to a cohort-based approach? Perhaps they could look at the cohort of customers who made a first purchase in March 2012 and the cohort of customers who made a first purchase in March 2013. For each group, they could track average revenue per customer over the first six months of their being customers. Here’s what it might look like:

Average Revenue Per User


Month 1

Month 2

Month 3

Month 4

Month 5

Month 6


March 2012 Cohort








March 2013 Cohort








First of all, you’ll see that average revenue per customer drops off pretty steeply between month 1 and month 2. That’s because every customer in the cohort is making a purchase in month 1 - that’s how the cohort is defined! - but only some customers are going on to repeat in the following month.

But we can also see that once we normalize for the amount of time that a particular group of customers has been around, the retailer’s customer retention seems to actually be getting worse over time. The group of customers who made their first purchase in March 2012 spent an average of $27/customer in their second month, while the group of customers who made their first purchase in March 2013 spent only $19/customer on average in their second month. This could mean that fewer customers are repeating, or that repeat customers are spending less on every transaction.

This provides a very different picture from the ARPU snapshots we saw earlier. Whereas ARPU doesn’t control for the age of the customer base - and is therefore very sensitive to changes in the mix of customers - cohort analysis provides an “apples-to-apples” comparison of different customer groups at the same point in their lifecycle.


Back to Custora U Library