For six months, I managed a Meta Ads account for a portfolio of lifestyle brands with a combined total budget of €15,000 across the six months. We hit 3.2× ROAS consistently. This is what I learned — not the headline numbers, but the uncomfortable truths underneath them.
Most Creative "Testing" Is Not Testing
When I inherited the account, there were 47 active ad variations across 12 campaigns. The previous manager called this "creative testing." It was not. It was creative hoarding.
Real creative testing requires a hypothesis. We believe this headline will outperform that one because of X. What I found instead was 47 ads running simultaneously, with no clear structure, no control variables, and no way to attribute performance to specific creative decisions.
We cut to 8 active variations across 3 campaigns. Performance improved by the second week. Not because the new ads were better — because we could finally see what was working.
The First Three Seconds Are the Only Seconds That Matter
Meta's own data says users decide whether to stop scrolling within 1.7 seconds of encountering an ad. Our data confirmed it. Across every campaign, video ads that opened with a static, high-contrast frame outperformed ads that opened with motion — because motion took time to register, and we didn't have time.
The implication for creative is brutal: your best line can't be in the middle of the copy. The visual hook has to earn the read. The first frame has to earn the next second. Nothing else matters until you've won the first three.
I started briefing creative with a single constraint: "What is the first frame? If that's all they see, have we communicated anything?" It changed everything.
Audiences Are Hypotheses, Not Targets
Every audience you build in Meta is a hypothesis about who your customer is. The mistake most brands make is building the audience they want — 25–40, high income, interested in sustainability — and then optimising the ad to perform within that audience.
The better approach is to build three or four structurally different hypotheses — interest-based, lookalike, broad with creative targeting — run them in isolation, and let the data tell you which hypothesis was closest to correct. Then narrow.
On one campaign, the highest-performing audience was 45–55 year old women. The brand had no idea this demographic existed in their customer base. The audience they'd been targeting — 28–35 year olds — was their aspiration, not their reality.
ROAS Is a Lagging Indicator. Watch the Middle.
The campaigns that sustained 3.2× ROAS over six months weren't the ones I optimised for ROAS. They were the ones I optimised for landing page conversion rate and add-to-cart rate — the metrics that sit in the middle of the funnel, between the click and the purchase.
ROAS tells you what happened. Mid-funnel metrics tell you why. An ad with a high CTR and a low conversion rate is sending people to a broken landing page. An ad with a low CTR and a high CVR is targeting precisely but under-reaching. You can't diagnose or improve the machine if you only watch the output.
The data doesn't tell you what to do. It tells you where to look. The decision is still yours.
Fatigue Is Real and Faster Than You Think
Creative fatigue — the point at which an audience has seen an ad enough times that performance drops — happens faster than most brands prepare for. In our account, for warm audiences, fatigue typically set in between days 14–21.
This means you need a creative production pipeline, not just creative assets. The brands that maintained performance weren't the ones with the best single ad. They were the ones with a system for consistently producing new creative.
Industrial design taught me that systems outlast objects. A good manufacturing process produces consistent quality at scale. A good creative process produces consistent performance at scale. The asset is the output, not the asset.
What I'd Tell My Earlier Self
Spend less time on the ad and more time on the landing page. Build your creative testing as a scientific experiment, not a playlist. Watch mid-funnel metrics weekly and ROAS monthly. And always — always — know what success looks like before you spend the first euro.
€15,000 across six months is an expensive classroom. But the curriculum is unambiguous: attention is scarce, data is noisy, and the best campaigns are the ones that are honest about both.