# ROAS Didn't Break Your Campaign. It Just Told You Last. > ROAS didn't break your campaign. It just told you last. By the time the ratio drops, the hook has already decayed, frequency has crept up, and CPMs have been quietly inflating for a week and a half. Here's the actual order things break in, and why watching ROAS first means you always find out too late. - **Author**: Mohit Goyal - **Published**: 2026-09-12 - **Category**: Creative strategy - **URL**: https://marxx.ai/posts/why-roas-drift-is-a-symptom-and-what-actually-starts-breaking-first --- The first time a campaign's ROAS cratered on me, I did what everyone does: I panicked and rewrote the ad copy. The tenth time it happened, I didn't touch a single headline. I went and looked at what had broken nine days earlier, because by then I knew ROAS is always the last one to find out. Here's the thing nobody puts on the dashboard: ROAS isn't a metric. It's a rumor that's had two weeks to travel. ## It's a ratio, and ratios lie by omission Revenue over spend. That's it. But stuffed inside that one number are four separate things, each with its own timeline and its own bad day: - **CPM**: what the auction is charging you for a thousand eyeballs - **CTR**: whether anyone actually cares enough to click - **CVR**: whether the click turns into money - **AOV**: how much the sale was worth once it landed These don't move together, and they definitely don't move on the same clock. CTR can be quietly bleeding out for a week while a slightly-better AOV covers for it. Frequency can be climbing while CPM hasn't caught up yet. ROAS averages all of it into one calm-looking line, right up until it isn't calm anymore. You weren't watching a decline. You were watching a delay with a good poker face. ## What actually happens, in order I've sat through enough of these post-mortems to know the sequence by heart at this point. It's almost always the same order, just compressed or stretched depending on spend level: **Day 0-3, the hook goes first.** The opening beat, first three seconds of the video, first glance at the static, stops earning the stop. This is invisible on a normal CTR chart because raw day-over-day CTR is noisy as hell. A real decay looks exactly like a bad Tuesday for the first several days. **Day 2-5, CTR actually starts trending down.** Not a bad day, a slope. You only see this if you're looking at trend, not averages. Most people aren't. **Day 4-8, frequency creeps up.** The algorithm's still trying to hit your budget, and reach has stalled, so it leans harder on the same people. Rising frequency next to flat reach isn't "scaling." It's the platform running out of new audience to hide the fatigue in. **Day 6-10, CPM starts climbing.** This is the one everyone reacts to, and it's the one that's least useful to react to, because it's downstream of everything above it. Engagement dropped, quality score followed, and now you're paying more for the exact same eyeballs. Fighting CPM directly here is like bailing water without finding the leak. **Day 8-14, conversion rate softens.** This is where people usually blame the landing page, or the season, or "the algorithm's being weird." Sometimes that's true. More often, the traffic mix has quietly shifted: fatigued creative pulls in more low-intent scroll-through and less of the audience that actually buys. **Day 10-21, ROAS finally shows up to the meeting.** By the time it crosses whatever threshold trips your alert, the creative's been sick for a week and a half to three weeks, and you've been paying inflated CPMs against a fatigued audience that whole time. That's the part that used to genuinely bother me: how much money gets spent in the gap between "something's wrong" and "the dashboard admits it." ## Why good marketers still miss this Not because they're not paying attention. Because the account is built to hide it. Most dashboards treat ROAS as the headline and CTR, frequency, and CPM as the fine print you check after ROAS already spooked you. That's backwards. Those three aren't the footnotes, they're the whole story, told early, if anyone's reading. And the tooling doesn't help. A 7-day rolling average is, by design, built to lag. That's the entire point of smoothing: it makes noise disappear, and it makes early signal disappear right along with it. You need something closer to a robust trend line, a slope estimate that isn't spooked by one noisy day, if you want to catch the hook decaying while it's still just a hunch and not yet a headline. ## The order I actually watch now, and so should you 1. **Hook retention / scroll-stop rate**: always first. Always. 2. **CTR trend line**, not the raw number: the first honest warning you get. 3. **Frequency against reach**: rising with flat reach means fatigue, not momentum. 4. **CPM**: a symptom, not a cause. Trace it back before you try to fix it. 5. **CVR**: softens last, gets blamed for problems it didn't cause. 6. **ROAS**: the final word, and the least useful place to start looking. ## What I'd actually change Stop making ROAS the thing you check first thing in the morning. Make hook retention and CTR slope the thing you check first, per creative, per day, with an alert on the trend, not on a threshold you've already blown through by the time it fires. If those two are healthy and ROAS is still soft, fine, go look at targeting, offer, or the calendar. That's a real possibility. But if the hook's decaying, you already know what needs to be refreshed, and you knew it a week before the ratio would've told you to panic. ROAS drift was never the disease. It's just the receipt that finally showed up. --- - [More Creative strategy articles](https://marxx.ai/posts/category/creative-strategy) - [All articles](https://marxx.ai/posts)