The "kill it under 3x ROAS" rule almost killed my best campaign

The "kill it under 3x ROAS" rule almost killed my best campaign

Everyone repeats it: if an ad set isn't doing 3x on day one, cut it. I followed that gospel for two years. Then I tracked cohorts instead of days.

The setup
Nutra offer, native traffic. Day-one return on ad spend sat at 1.4x. Textbook says kill.

The move
Instead of cutting, I pulled the postback data and matched conversions back to the day the click happened, not the day the sale posted.

The numbers
That 1.4x ad set was actually doing 2.9x once delayed conversions (rebills and 3-7 day lags) landed. I'd been killing winners because the network attributed sales to the wrong calendar day. Roughly $4,100 spent, $11,900 returned over 14 days — but only $5,700 of it showed inside the first 72 hours.

The lesson
ROAS thresholds measured on same-day data punish offers with any conversion lag. The rule isn't wrong; the measurement window is.

What I'd do differently
Set your kill rule on a 7-day attributed cohort, not the dashboard's "today" number. I now wait for at least 50 clicks AND one full conversion cycle before any ad set lives or dies. Illustrative figures, but the lag is real on every rebill offer I've run.