Every retention model breaks at $5M ARR. The model itself is fine — the assumptions underneath it stop holding. Acquisition shifts from founder-network to paid. The cohort that loved you in year one is no longer the cohort you're acquiring in year three. The dashboard does not register this because the metric definitions never updated.
What to build before the break
A cohort-level retention view, not an aggregate one. A per-channel curve. A clear separation between activation and retention, because they are the same problem at different timescales but they are owned by different people in most orgs.
The most useful retention metric is rarely the most reported one. For consumer products it is often something idiosyncratic — second-week behavior on a specific surface — that nobody outside the company would think to track.
"Activation and retention are the same problem looked at on different timescales. Most teams own them in different rooms."
If you wait until the cliff is in the dashboard, you are six months late. The lever you needed to pull was upstream, in onboarding, two quarters ago.
Dispatch Newsletter
One operator-grade breakdown every Tuesday. SEA tech, AI economics, execution frameworks. No fluff.
