Cohort analysis – a simple guide

Cohort analysis – a simple guide
Cohort Analysis thumbnail
Image courtesy of DataHero

Cohort analysis is most probably the first slightly more advanced analytics technique I have come across during my experience, besides controlled experiments, which really makes a lot of sense and turns out to be simpler than expected. Recently I got reminded of its power in managing a marketing channel portfolio, so thought of sharing how much you can gain from it.

What need can it answer?

Let’s say you want to know how valuable customers acquired through different channels are. You look at how much money till today customers on average generate acquired from PPC or Display or Social. But this measure is flawed.

  • When did you run your last campaign on these channel?
  • How long are these customers in your portfolio?
  • Do you count with churners?

All of these make it difficult to answer this question right and issues similar to these ones

  • What is your retention of customers
  • Do they purchase again or upgrade to a higher service level?
  • Is your channel performing better or worse over time?

So what cohort analysis is?

The concept is that you pick a reference date which all your customers have gone through and measure performance on how much time has passed since then.

The official definition of cohort actually is:

A cohort is a group of people who share a common characteristic over a certain period of time.

An example: what percent of your customers purchased again 1-2-3 months since their first purchase. This way you can compare people coming from 2010 and 2015 campaigns because your analysis does not rely on specific dates, but on time passed from the same reference events.

These reference events could be:

  • Acquisition day
  • Day of newsletter sign up
  • First purchase or the 5th purchase as well
  • Time of first ad click
  • Day of sending the first inquiry email
  • Or anything which have a clear timestamp

Cohort Analysis Tabular

There are many ways to look at it, but basically you can start seeing how customers behave and add value to your base throughout their lifetime. You can then go a segment this information not only basedon when they were e.g.: acquired, but also based on channels or any other attributes you feel important to manage your marketing activity portfolio.

How to best use it

This is where it gets really interesting, The whole point is to not get “stuck” with comparing different generations of customers, but as mentioned above, use different attributes as well. This is how cohort analysis can actually really start driving value for you.

Refering back to the question at the beginning of the post, and a similar example from the Mode Analytics blog, you can for example start comparing the value of customers acquired through different channels. Pairing that with the cost of acquisition you can set up your channel portfolio to drive the most revenue and/or profit to your organization.

You will also be able to validate this portfolio set up time and again, since you will have the new generations of customers giving you the same kind of information which will enable you to refresh your cohort analysis. Based on this fresh data you will be able to recalibrate your portfolio.

In conclusion

While cohor analysis is not the simplest analytics technique ever, once you start using the information coming from it, I’m certain your boss(es) will really start loving you as you can prove that you are getting the most out of your channel selection and that you are able to manage that portfolio with state-of-the-art methodology.

So go ahead, give it a try and learn more about cohort analysis and its benefits for marketers.


Leave a Reply