Net retention is the number, and it is getting harder to read

It traded at 137.01 this morning, -3.12% on the session. I want to write about disclosure rather than the business, because across this whole cohort the disclosure has been quietly moving.

Why net retention carries so much weight

For a subscription business, net revenue retention answers the only question that matters: does last year's customer base spend more this year, after churn. Above a certain level, growth compounds without acquiring anyone new. Below it, the company is running to stand still.

It is also the single number that most determines what multiple the market applies. Which creates an obvious incentive.

What has been changing

Not fraud, and I want to be careful about the accusation. Definitional drift:

• Which customers are in the cohort, and the revenue threshold to qualify
• Whether the figure is trailing twelve months or a point in time
• Whether it is disclosed as a precise number, a band, or "above" a round figure
• Whether it appears in the release at all, or has moved to the appendix

Each change is defensible on its own. The pattern across a whole sector is harder to read charitably, and the direction has been towards less precision rather than more.

Why I am not positioned

A metric whose definition moves cannot be compared to its own history, and comparison to its own history was the entire point.

I do not know whether retention here is genuinely deteriorating or merely being reported less precisely. Those two things look identical from outside, and only one is a reason to avoid the shares.

Being flat while I cannot tell the difference is not indecision. It is the correct response to an information problem.

What I am watching

Restored precision, or enough quarters on the new basis to establish a comparable series. Either resolves it. Until then the software cohort is one I read closely and own sparingly, which is an uncomfortable place to be given what the businesses themselves are like.