Journal · Cohorts
Cohort charts that mislead finance
Retention curves are the most persuasive pictures in subscription revenue analytics. They are also the easiest to stage. Finance does not need fewer charts. It needs the entry rule printed above the chart.
Month zero is a policy, not a timestamp
Product analytics often treats “signed up” as month zero. Finance cannot. A trial that never converted, a logo that signed and paused before first invoice, and a partner-billed seat that never touches your gateway do not belong in the same opening column. When they do, month-one retention looks heroic because the weak logos were never really in.
On the Recurring Desk we require a written entry rule before anyone draws. Example: “Month zero is the month of first recognised recurring invoice greater than zero, excluding courtesy credits.” Ugly sentence. Useful chart.
Silent removals
Some warehouse jobs drop logos that churned in month one from the original cohort so the curve “focuses on real customers.” That is not analysis. That is editing. If you want a survivor cohort, say so in the title. If you want an acquisition cohort, leave the dead logos in and let the curve fall. Boards can survive a falling curve. They cannot survive a curve that was quietly rewritten.
Revenue versus logo cohorts
A logo cohort can look stable while revenue from the same logos shrinks through seat loss. Plotting only logos is how CS teams get praised in the same quarter finance misses budget. Subscription revenue analytics insists on both pictures, same entry rule, same period. If you only have room for one, print revenue. Logos can live in the appendix.
Small Bangkok samples
Several of our case companies have cohort sizes that would embarrass a consumer app. Twenty logos is not a statistical argument. It is still a management argument if you annotate sample size on the axis. We have seen teams hide n=14 under a gradient area chart. The gradient was not the problem. The missing n was.
If your cohort is small, write names (internally) and talk about contracts. A chart that pretends to be a law of nature is the misleading one.