Samaipata’s 3-step/5' cohort analysis (Part 1):

First steps to a cohort analysis

By Álvaro González San Pedro

Welcome to Samaipata Ventures’ blog, a VC fund specializing in market-place and disruptive e-commerce models in Europe. If you don’t want to miss out on any quality information, follow us here on Medium, Twitter or suscribe to our newsletter.

If you’ve never heard about cohorts before, or if you have but you are still unsure of what they mean, this post is for you. If you’ve already used cohorts, but want to deepen your knowledge, jump to Part 2: Run a full stack cohort analysis from scratch (…).

Before getting to cohorts, let’s start from the very beginning. You want your business to escalate, and perhaps even more important, you want investors to believe in it. One condition is absolutely sacred then: at some point in time, the value you can gain from a user during his lifetime (customer lifetime value- LTV) has to exceed the cost of acquiring this new user (Cost of Acquisition or CAC). Otherwise the business will never be profitable. It’ll require external funding eternally and sooner than later investors will grow tired of wasting their money on something that might never pay off (and unfortunately we all know that investors are not exactly known for their patience waiting for results).

What does this mean then? First, that acquiring new users is not always as profitable as we might think; particularly if they don’t end up covering what it initially cost to acquire them. Let’s work with an example; if you spend 100 on marketing and you manage to acquire 10 new users, each has a CAC of 10. If this new user is not expected to generate at least 10 euros of contribution margin during his lifetime as a user (LTV); then the effort might not make sense. Indeed, sometimes it might be wiser to work on retaining already acquired users, than acquiring new ones. So how can we know what to do when?

Let’s do a bit of easy maths. The CAC is quite straightforward to calculate (normally, total marketing cost divided by the number of newly acquired users), so let’s have a look at a client’s future profitability (LTV). The LTV can be calculated up to different periods: up to 12 months (what you make from this client in 12 months), 24 months, etc. The most aggresive managers/investors will target +36-months paybacks in order to boost growth, whilst the more conservative prefer shorter paybacks, sometimees even lower than 6 months. To calculate the LTV just multiply the number of orders that a customer is expected to make over his/her lifetime as a user –say in 36 months– (this is recurrence) and multiply it by your average contribution margin per order. Simply put, LTV = recurrence during a customer’s lifetime x contribution margin.

Calculating your margin is also straightforward; recurrence might be a tad more complicated and this is where cohorts (finally!) come in. In statistics, a cohort is a group of people who share a common characteristic; thus, a cohort could be a group of customers who are from the same city, who are female, who were born the same year, etc. A cohort analysis then is the study of how a certain group of people who share a common characteristic behave across time.

Applied to our field, one of the most important factors that could explain changes in customer behaviour across time is the date in which the customer was acquired: people that become users in similar dates could have reacted similarly to a specific acquisition strategy from the company (for instance, vouchers) or felt the need to use the service/good in the same season, etc. This is important to know. Thus, it has become a standard to define cohorts as the group of people who become first time buyers (FTB) in a given date (usually a given month, sometimes week). For instance, all the new users you gain in September 2016 make up a cohort.

You are now ready to do your own cohort analysis, using our 3-step/5' template. But first, read part 2: “Run a full stack cohort analysis from scratch (…) where you’ll find the Excel file for you to download. Don’t worry, the template we are sharing is as intuitive as possible and you’ll just need 5 minutes to complete the 3 steps by yourself! Stop bothering your tech team with things you can do on your own and let them focus on the hard stuff.

What are you waiting for to try it out? If you have any further doubts with interpreting or updating the data, we’ll be delighted to help out. Email us! If you have any problems downloading the file please contact us and we’ll send it you.