ROAS is never read in isolation. Every verdict is cross-checked against order count and campaign duration, so a lucky four-order weekend can't fake a scale signal.
Ads get matched to the exact product they're selling — not by campaign name, by what's actually in the creative and what's actually moving in the store.
Separates "the hook fails" from "the hook works but the content after it loses people" — two different problems, two different fixes.
Every report remembers the last three. When the same ad keeps showing up, ROASted says so — consistent winner, or a slow decline you'd otherwise miss.
Not "make more videos like this." A scene-by-scene brief for the next one, built from what's actually working in your account.
Optional signal from the Meta Ad Library when it's useful — offered as context, never forced into a verdict that doesn't need it.
Ask a follow-up question and get an answer grounded in your actual numbers, not a generic marketing tip.