If you’re an online marketer or entrepreneur and don’t understand what a cohort analysis is, PLEASE CONTINUE READING. For everyone else, hopefully this is pretty obvious stuff but a refresher never hurt.

Before we get into the specifics, here’s what everyone’s expert Wikipedia has to say about what a cohort is:

A cohort is a group of people who share a common characteristic or experience within a defined period.

Simple enough, right? So for marketers, a typical cohort can relate to things like Users, Customers, Leads, Visitors etc. Essentially these are all pieces of your marketing funnel that you can compare for different segments of people — April vs. May, Boy vs. Girl, AdWords vs. Bing etc.

For example, a SaaS with a freemium model would have a basic overall funnel that looks something like this:

A more sophisticated funnel might have many more steps, and include things like (and depending on your business):

- Downloads
- Installs
- Clicks
- Repeat Purchases
- Engagement
- Referrals
- Cold Calls

Let’s start with a very basic example where we examine the cohort performance of a typical SaaS business over two periods of time. This could be Years, Quarters, Months, Weeks, Days, Hours – you get the picture.

Example One: What are my conversion rates?

When looking at this data, there are a few things going on. First of all, Visits went down in P2 (top of our funnel). Not a good sign, but also probably not the most critical metric here.

Next, Leads (let’s assume a free trial of some sort) are actually higher in P2 despite P2’s top-of-the-funnel being smaller! What gives? As you can see, P2 has a conversion-to-free of 2.7%. A clear winner right? Not so fast!

The most important metric you want to measure probably relates to your bottom-line. In this example, that would mean Conversions (paying customers). With 227 in P1, so far it’s the clear winner. However, this is not to say that P2 is down and out. Remember, P1 Leads had a head start, and as a result started converting much earlier than their P2 counterparts. It’s just as likely that since P2 has more Leads than P1, eventually they’ll match the conversion rate of P1. You need to watch this over time and compare multiple cohorts.

The Visits and Leads in our example won’t change since that time period is now over, but those same Leads may convert-to-paid at any time, causing your conversion rate to increase.

Example Two (A): How long does it take for my Leads to convert?

Building off our data in Example One, we’ll now start answering the question of understanding how long it takes for a Lead (middle-of-the-funnel) to convert to a paid Customer.

In this example, let’s assume P1 refers to a month – April. In April there were 1,876 new Leads. In that same month there were 134 Conversions (Customers/Upgrades/Subscriptions/Unit Sales – whatever you want to call it). Predictably, conversion was highest in the first month with 71. That most likely indicates that the free trial length was <30 days. If it were >30 days, there probably would be very few conversions in P1.

These conversion counts are great, but any true marketer needs the corresponding percentages. Percentages tell the story. Take a look at the same data in Example Two, only now presenting in %’s:

Example Two (B): How can I predict my conversion rates if there is a long paid lag effect?

As your eyes move diagonally here, you’re able to compare conversion %’s across similar periods of time. If we stick with treating Months as our cohort, then M1 is paid conversion within the first month that a Lead joins that cohort (M2 is the second month). Here are a few takeaways to consider:

• For April (P1), conversion in month one (M1) was 7.1%. By September (P6), this drops to 4.0%. This is a clear indication that there is some sort of problem. Either the top-of-funnel is being filled with less qualified Leads, or something has changed – causing conversion to decline.

• The reverse can said about M5. The average conversion rate in M5 (not shown above) is clearly higher than the average conversion % in M4. This is contradictory to what we would expect. A marketer should ask herself what is happening in that 5th month to cause an increase in conversion %. Whatever it is, do more of it!

We’ve explored the basics and by now you should all be beginners (not dummies) at understanding what’s involved in a cohort analysis. There are some great tools that will do the data collection and reporting for you (Google Analytics, KISSmetrics) but it’s critical that you’re able to comprehend what you’re seeing in such reporting.

Want to know how to layer more sophisticated data such as customer retention /churn, lifetime value, and campaign optimization? Check out the second part of this post.