Third-party signal loss hits attribution: 38% of SMB dashboards now misreport ROAS

An audit of 1,200 small-business ad accounts found that 38% of platform-reported ROAS figures diverge from actual revenue by more than 25%. The culprit is cumulative: third-party cookie deprecation, iOS signal loss, and modeled conversions filling the gaps with increasingly optimistic estimates.

The dangerous part isn't the inaccuracy — it's the direction. Modeled attribution consistently over-credits paid channels and under-credits organic and direct, which quietly pushes budget toward the channels the platforms sell.

The gold-standard fix is incrementality testing: run geographic or audience holdouts and measure the actual lift. But holdouts require volume most small businesses don't have. The practical alternative: weight your first-party conversion data (actual orders, signed contracts, booked calls) at twice the value of platform-reported conversions when making budget decisions.

If your dashboard says a channel is your best performer but your bank account disagrees, believe the bank account.