The Doorman Problem: what your retargeting dashboard can't tell you

Their question

A growth lead at a direct-to-consumer brand brings the report every operator has seen: retargeting is the best channel on the dashboard. Eight times return. It gets fed first when budgets move, defended first when budgets shrink.

And yet growth feels stuck. Total revenue isn't moving the way an 8× channel says it should. The uncomfortable question — the one the dashboard is structurally unable to answer — is simple: how many of those people were coming back anyway?

The quiet test

The dashboard can't answer it because the platform selected the audience, served the ads, and then graded its own work. Retargeting reaches, by definition, the people most likely to return — so it books credit for every return that follows an impression.

So Signal ran the only test that settles it: it quietly kept the ads away from a small slice of statistically identical shoppers, and watched what they did anyway. No opinions, no models arguing with models — a held-out group and a register.

What the ledger showed

Most of the held-out shoppers bought regardless. When each conversion was classified — caused by the spend, or merely anticipated by it — the channel's real, incremental return landed near 1.4×, not 8×. The gap between the two numbers wasn't fraud and it wasn't a tracking bug. The dashboard had been counting customers who were already on their way back through the door.

What changed

The budget conversation stopped being theological. Retargeting spend came down by a third — held, not zeroed, because 1.4× is still a return — and the freed money moved to finding customers the brand had never met. Revenue held. Growth stopped feeling stuck, because the money was finally buying new demand instead of escorting existing demand to the register.

You were paying a doorman to open a door people were already walking through. The fix wasn't firing the doorman — it was sending him to the doors that were actually closed.