Hook, hold and click: what each one tells you, what a good number looks like, and why ROAS alone picks the wrong winners.
ROAS tells you which ad made money last week. It doesn't tell you why, or which new ad will. So every ad we launch gets three more scores, read in order: hook, hold and click.
Why ROAS isn't enough
A new ad with $600 behind it can post a 4× ROAS from two lucky orders. A week later it's at 1.2×. Early ROAS is noise: there simply aren't enough purchases yet to trust it. The attention numbers settle much faster, usually within two days, and unlike ROAS they tell you what to fix.
We still decide winners on cost per purchase. But we use hook, hold and click to decide what to make next, and to kill obvious losers before they've spent their whole test budget.
Hook
The share of people still watching at three seconds. It measures the first line, the first frame and nothing else. Under 25%, the opening isn't stopping anyone and nothing after it matters. Across our accounts winners average 38%, and the best we've seen this year was 47%.
If the hook is low, we rewrite the first line, not the whole ad. Most of the time the fix is saying the most interesting thing first, out loud, instead of building up to it.
Hold
The share who watch at least half. A high hook with a low hold means the opening promised something the rest didn't deliver, or took too long to deliver it. We rewrite the middle, not the opening.
Hold is also the number that tells you about length. If an ad holds 30% at fifteen seconds and 8% at thirty, it was a fifteen-second ad.
Click
Click-through rate to the store. It's the last thing to look at, because a weak hook or hold drags it down too. When hook and hold are healthy and clicks are low, the problem is usually the offer or the last line: people watched, liked it and didn't know what to do next.
Putting them together
We read the three in order, like a funnel. A low number early explains everything after it, so fixing the earliest weak score gives the biggest jump. Every ad's three scores go into the Monday scorecard next to its cost per purchase, and the next week's briefs start from whichever number was weakest.
Nadia Okafor
Co-founder and head of buying. Ran $2M a month on Meta before Winrate.
