I’ve been asked how to analyse cohorts by a couple of companies recently, so I thought I’d distill my thinking here into a blog post.
Most companies show their cohort analysis in the form of a table like the one below. This is the format that comes from most popular analytics packages.
Whilst these tables are helpful (and believe me, I’ve read a lot of them) they are a lot more useful if combined with margin data and shown in a chart like the one below:
CM1 in the chart title stands for Contribution Margin 1 – i.e. the contribution from the average customer in the cohort to covering the cost of marketing and the central overheads of the business. For marketplaces and ecommerce companies that means revenues (net of VAT) less any discounts or rebates, the cost of the physical item, delivery costs and the cost of returns.
This view is useful for a few reasons:
- It’s easy to see that even the oldest cohorts are still improving over time and that the new ones are more valuable than the old ones (you can see this from the table too, but it’s harder).
- You can see the lifetime value (LTV) of each cohort – the more mature cohorts are nearing £80, whilst the newest is nearer £70. Young companies can extrapolate these lines to estimate the ultimate LTV.
- You can work out how long it will take to pay back varying customer acquisition costs (CAC). The dotted red line shows that all cohorts would have paid back a £60 CAC after three months, but that the most recent cohort would have paid back a £72 CAC in the same period. CAC and payback period are key inputs into the financial model which works out how much cash a company will burn each month at a given growth rate, and therefore whether a company can get past key revenue milestones before they need to raise their next round.
This chart is most useful for companies like ecommerce businesses and marketplaces where customers make repeat purchases on irregular schedules. You can also use it for with subscription businesses (including SaaS) but in these situations calculations based on churn rate might be simpler and more effective.