Meta creative refresh: rotate on frequency or on cost drift?

Frequency alone is a bad refresh trigger. A frequency of 4 on a small retargeting audience can be perfectly fine if CTR and cost per result hold steady, while a frequency of 2 on a broad prospecting ad set can already be wearing out, because the same hook keeps landing on the same tired scrollers. Picture a dental clinic running one before-and-after reel for six weeks. Frequency creeps up slowly, but the hook stops stopping thumbs and cost per booking climbs 40 percent against its own first-week baseline. That's the signal to act on. Frequency just tells you where to look. Compare the result trend to the ad set's launch numbers and swap the creative when the cost moves, not when some number crosses a magic line. What trigger do you use?

Waiting for cost to move is the slow way to catch fatigue. CPA is usually the last number to react when a creative tires, and it also shifts for reasons that have nothing to do with the ad: auction pressure, a competitor bidding up, a learning phase reset. If you only swap once CPA is up 40%, you've already paid for weeks of weaker delivery.

The earlier signal is the creative's own engagement. When CTR or hook rate drops against its launch week and the audience hasn't changed, the creative is wearing out. I'd use that as the trigger and let cost confirm it.

Hook rate is a better trigger than CTR for video, mostly because it's the closest thing to a pure creative signal. CTR gets dragged around by placement mix and by how many times people have already seen the ad. Still, I'd split the diagnosis before swapping anything. If hook rate drops while frequency climbs, the opening has worn out, and a new hook on the same body might be enough. If hook rate was already low in week one, fatigue was never the problem, and rotating in a fresh ad just gives you another weak one. Which of those does your launch-week baseline look like?

Splitting the diagnosis makes sense, and I'd add a second number: hold rate, meaning the share of people still watching at around 10 seconds. If hook rate holds but people drop off around second ten, the body is what's tired, and a new opening will just send more people to the same dead spot.

Compare within each placement too. Reels usually posts a higher hook rate than Feed, so a blended number can make a healthy ad look like it's sliding.