Why platform growth slows before anybody admits it

Sign-ups are up nine per cent, the dashboard is green, and the team reads it as continued momentum. The composition changed underneath: the people joining this quarter needed four times the spend to reach and leave twice as fast. The total is still rising.

What everyone sees

Growth is read off aggregate counts, which treat every additional user as equivalent to the last. They are not equivalent. Adoption proceeds through populations in order of how ready they were, so the remaining pool is systematically harder than the one already converted, and a flat total can hide a collapsing marginal return.

What is actually happening

Rogers’ synthesis of diffusion research established that adopter categories differ systematically — early adopters requiring less persuasion, less social proof and less support than the later majority, who adopt only after the innovation’s uncertainty has been resolved by others. Bass’s growth model formalises the arithmetic: adoption is driven by a shrinking pool of remaining non-adopters, so the same conversion effort produces fewer additions as the pool depletes, and the curve bends before the absolute numbers fall. The two together predict exactly the pattern that surprises operators. The slowdown appears first in cost per addition and retention of recent cohorts, both of which are averaged away in the headline figure for several quarters.

Why it stays hidden

The turn hides because the metrics that would show it are cohort-level and the metrics on the wall are cumulative. Total users only ever rises. Blended retention improves as the large early cohorts age well, which makes the worsening recent cohorts invisible inside a number that is getting better.