Every business that retains customers should be running a cohort retention chart. Most are not. The ones that are usually pull the chart once a quarter, glance at it, and put it back. That is a waste. The cohort curve is the single highest-signal artifact in a retention-led business, and if you know how to read it, it tells you what to do next.
The chart, in one paragraph
Group every customer by the month they made their first purchase. For each cohort, plot the percentage of original customers still active (or still purchasing, or still subscribed, depending on your business model) at month one, month two, month three, and so on. You end up with a chart where every line is a cohort and every line trends down over time. The shape of the lines is the entire story.
Three shapes that matter
Real cohort curves take one of three shapes. Knowing which shape your business has is more useful than knowing your blended retention rate.
- 01Decay to zero. Every cohort trends downward and approaches zero. This is a transactional business that depends entirely on new customer acquisition to grow. The customers you acquired this month will be effectively gone by month twelve. Common in single-purchase commerce, low-stakes services, and most low-AOV ecommerce.
- 02Decay to a floor. Cohorts trend down, then flatten out at a non-zero asymptote. The percentage of customers who survive past month six is the same percentage that survives past month thirty. This is a business with a loyal core. The asymptote is the most important number you have, because that floor is what compounds.
- 03U-curve. Cohorts drop, hit a floor, and then climb. Rare and exceptional. Usually only seen in businesses with a strong network or community effect. If you see this in your data, the right strategy is investment in whatever is driving the recovery.
The asymptote is the business
If your cohort curve flattens at a non-zero floor, that floor is the single most valuable number in your business. A cohort that flattens at 12% means that for every 100 customers you acquire, 12 will still be customers in year three. Your true LTV is calculated against that 12%, not against the average. A business with a 12% asymptote on a $400 yearly customer is worth dramatically more than a business with no asymptote and a $400 average first-year revenue.
Blended retention numbers tell you what already happened. The cohort curve tells you what is about to happen.
What the curve tells you to do
Reading the curve is one thing. Using it is another. A few decisions the curve makes for you:
- If newer cohorts are decaying faster than older ones, your product or acquisition channel has degraded. Investigate the most recent change before assuming a market problem.
- If newer cohorts retain better, something you changed is working. Find the change and double down before the effect washes out.
- If the asymptote is high but the early decay is steep, your onboarding is broken. The customers who survive month three turn into loyalists; the ones who don't never had a chance.
- If the early decay is shallow but the asymptote is zero, your product retains attention but not loyalty. There is a wall around month nine to month fifteen where you are losing the people who initially loved you. Find the wall and figure out why.
How often to look
Monthly. Always monthly. Quarterly is too slow to catch a degradation; weekly is too noisy to see the trend. Print the chart on the first Monday of every month, compare against the previous three months, and ask one question: has the shape changed? If the shape has changed, that is the operating priority for the month. If the shape has not changed, you have permission to focus on something else for thirty days.