Cohort analysis
Cohort analysis is grouping customers by when they first bought (or signed up), then watching each group over the months that follow. A cohort is simply a batch of people who started in the same period, such as everyone who first bought in March. It shows whether this month's new customers spend like last month's. One big sales number hides that difference.
Theodor Lindfors, Founding Marketer ·
How cohort analysis works
Pick the grouping: acquisition month is the usual one, but channel, campaign, first product, or discount used all work. Pick what you follow: repeat purchase rate, revenue per customer, retention, or payback. Then plot each cohort by months since acquisition so groups of different ages are comparable.
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. The drink brand advertises on Meta, TikTok, and YouTube.
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. In January, the drink brand acquired 1,000 new customers. In month 1 those 1,000 customers spent $48 each on average (a first order plus a few extras). In month 2 those 1,000 customers spent $22. In month 3 those 1,000 customers spent $18. That is the January cohort's curve.
Let's take a drink brand as an example. They sell a low-sugar sports drink for active women over 50, mostly from their own website. In March, the drink brand ran a 30% off TikTok push and acquired another 1,000 customers. Month 1 spend is $36 per customer. Month 2 spend is $8 per customer. Month 3 spend is $6 per customer. Same company, same product. The March cohort is cheaper to impress and quicker to disappear. A blended "average customer spends $40" number would have hidden that, because January still props up the average.
Why cohort analysis matters
Cohort analysis is how you find out whether growth is healthy. Rising CAC (customer acquisition cost: what you spend to get one new customer) is only a problem if the customers are not getting more valuable to match. Cohorts show both sides on the same chart. If March customer acquisition cost is higher and March repeat spend is lower, a drink brand is buying worse customers, not just paying more.
Cohort analysis also gives LTV (lifetime value: total spend from a customer over time) a shape instead of a single number, so you can see how long payback actually takes rather than assuming a lifetime you have never observed.
How to read a cohort chart
Compare cohorts at the same age. The March cohort has not had time to repeat, so its lifetime number will always look small in April. Read down the same column (month 2 for January vs month 2 for March), not across rows of different maturity.
Look for the break. If cohorts from a particular month all curve lower, ask what changed then: a discount push, a channel mix shift, a new offer. For a drink brand, the 30% off TikTok campaign is the finding, not a mysterious drop in brand love.
Common cohort analysis mistakes
- Comparing an immature cohort to a mature one and calling it a decline.
- Extrapolating a lifetime from three months of data.
- Cohorting by acquisition date only, when the interesting split is channel or first product.
Lemonado
Your AI co-worker for marketing
Lemonado is your marketing team's AI co-worker. It connects to your whole stack and does the work end-to-end: reporting, campaign checks, and analysis.
In-house teams and agencies use it so people spend attention on decisions, not busywork.