Cohort analysis for startups 101
As part of moving my blog to medium, I will repost a selection of the old posts. This post was written in June 2015 and originally appeared at nikos.vc.
Most of the times a VC fund is considering an investment in a startup it will ask for a cohort analysis. Such an analysis is comparing different groups of users throughout time. As a company is building its product, it is constantly testing new strategies. Customers that start using an application in the first week after its launch will have a completely different experience from users that start using an application later in time. Cohort analysis aims to help entrepreneurs (and potential investors) to understand whether a company really improves throughout time. Lets look at a specific example and assume that we are analysing an e-commerce site. This website has 500 new buyers every month. In the majority of the pitch decks that are shared with VCs you would find a figure like the one below.
The first impression from the figure above is that this specific e-commerce business is developing very well. The reality is though that these numbers don’t necessarily say a lot. Even looking at the average money spent per customer doesn’t give a clear picture.
The reason that these figures don’t give a clear understanding of the development of the business is that it is mixing revenues from clients that shopped for the first time in January, with revenues from clients that shopped for the first time in May. Lets look at two different scenarios regarding how the revenues per customer cohort could be in reality. Both scenarios would result in a figure like the one above. But as will be made clear, these represent two totally different scenarios regarding the development of the company.
Scenario 1
When rearranging the data, we would end up with a figure like the one below, which is a typical figure used in a cohort analysis.
From the table above we can observe the following:
- Clients that bought for the first time in January, spent on average $10 in January, $8 in their second month (February), $3 in their third month etc. Clients that bought for the first time in February, spent $11 in their first month, $8 in their second month, $7 in the third etc.
- We can observe a trend that clients spend on average more money during their first month throughout time. This would probably mean that either the quality of the products or user experience is improving.
- Clients that bought for the first time in January are spending only a small amount 5 months later.
Scenario 2
When rearranging the data in a similar way as before we end up with the following table.
This table is showing a different story.
- Clients who started buying in January are still spending a lot of money.
- Clients who started buying in February are spending less money than in Scenario 1.
- Clients of March and February are almost at the same spending levels.
- There is a trend of clients spending less money on average during their first month throughout time. Revenues remain at high levels because January clients are loyal customers. Still, there is no clear improvement for new clients.
Cohort analysis would give entrepreneurs and potential investors a clear overview of a company’s progress. One could create this type of analysis for revenues, users, average revenue per user, expenses or any other metric. Since it is an analysis that VCs like a lot, entrepreneurs that are fundraising should be aware of it. You can find more information on cohort analyses, the correct way to read and interpret the and some templates in the following links.
- The ultimate cohort analysis cheat sheet (https://chartmogul.com/blog/2015/02/the-ultimate-cohort-analysis-cheat-sheet/)
- Excel template for cohort analyses in SaaS (http://christophjanz.blogspot.nl/2013/10/excel-template-for-cohort-analyses-in.html)
- Great book on lean analytics (http://leananalyticsbook.com)
