How to Prove Your Ad Creative Is Driving ROAS

Your ROAS went up last month. Was it the new ad? The audience? A bidding change? Most marketers can’t actually answer that with confidence, because the number on the dashboard doesn’t tell you why it moved. This post walks through a real method for figuring out how much of your results actually come from creative, so you can say so with evidence instead of a guess.

Why the Platform’s Own ROAS Number Doesn’t Answer This

The ROAS number in Ads Manager tells you what happened. It doesn’t tell you why. A campaign can post a great ROAS because the creative is strong, because the audience was already primed to buy, because a competitor paused their ads that week, or some mix of all three. The platform doesn’t separate these out for you, and most reporting stops at the account-level number without digging further.

This has gotten harder, not easier, over the past couple of years. Automated campaign types like Advantage+, and ranking systems like Meta’s Andromeda and GEM (which we cover in a separate post), increasingly decide targeting and creative selection together, inside the same system. That makes it even less clear, from the outside, which lever actually moved your results.

The Metrics That Actually Point to Creative

A few numbers are better proxies for "creative worked" than others. Hook rate (how many people stop scrolling and start watching or reading) and outbound click-through rate are closer signals of creative quality, since they measure the moment someone reacts to what they saw, before targeting or landing page even come into play. Conversion rate and cost per thousand impressions lean more toward targeting: they reflect who saw the ad, not just how they reacted to it.

Watching these side by side is more useful than watching ROAS alone. If hook rate and click-through rate are both healthy but conversion rate drops, the problem is more likely downstream (targeting, offer, or landing page) than the creative itself. If hook rate drops too, the creative is the more likely cause.

Method 1: Controlled Test Cells

The most direct way to isolate creative is to hold everything else still and only change the creative. Run two or more ad sets with the same audience, same budget, and same placement, and change only the creative between them. Whatever difference shows up in results is much more likely to be caused by the creative itself, since nothing else moved.

This sounds simple, but it takes discipline. It’s easy to accidentally change more than one thing at once, like launching a new creative alongside a new audience or a new offer, which makes it impossible to tell which change caused the result.

Method 2: Reading Platform Metrics as Proxies

When a clean test isn’t possible, and sometimes it isn’t, you can still get a reasonable read by watching the right combination of metrics rather than one number alone.

If hook rate and outbound click-through rate are strong but conversion rate and cost per acquisition are weak, that points toward the creative doing its job while something else, like targeting, the offer, or the landing page, is the weak link.

If hook rate and click-through rate are both weak, the creative is more likely the actual problem, regardless of what targeting or bidding are doing.

This method is faster than running a full controlled test, but it’s a diagnostic, not proof. Treat it as a strong hint about where to look next, not a final answer.

Method 3: Correcting for Over-Reported Results

Ad platforms tend to take credit for sales that might have happened anyway. Someone who already planned to buy, and would have clicked through from a Google search or typed the URL directly, still gets counted as a platform-driven conversion if they happened to see and click an ad first. This makes reported ROAS look better than the ad’s real, incremental impact.

Incrementality testing (sometimes called geo testing or holdout testing) deals with this by comparing a group that sees your ads to a similar group that doesn’t, then measuring the actual difference in sales between the two groups. This is the only method of the three that tells you what would have happened without the ad at all, which is really the question "did this ad create value or just take credit for value that already existed."

This method takes more setup than the other two, and isn’t practical to run constantly, but it’s worth doing periodically, especially before making a big decision based on a ROAS number you’re not fully sure about.

Turning This Into a Report Someone Will Actually Trust

A creative-attribution report is more convincing when it shows its work instead of just stating a conclusion. A useful structure: state the ROAS number, then show the supporting metrics (hook rate, click-through rate, conversion rate) next to it, then explain which method was used to check whether creative was really the driver, and note any caveats about what wasn’t tested.

Reporting this way on a regular cadence, monthly is reasonable for most teams, builds a track record over time. A single report showing creative drove a result is a claim. Several months of consistent reporting, using the same method each time, is evidence.