See if we're a fit

Creative metrics & testing

Ad Creative Metrics That Lie: Which KPIs to Trust

Gabe Hutcheon · · 6 min read

Trust metrics in this order: amount spent > ROAS/CPA > CTR/CPC > hook and hold rate. Hook and hold rate are diagnostic only. do not optimise against them. And reliability rises as you move up the account: ad < ad set < campaign.

Most KPI articles hand you ten metrics with equal weight. That is how brands kill good ads and scale bad ones. Some metrics tell the truth, some lie, and the order you trust them in matters more than any single number. Here is the hierarchy we use.

The reliability hierarchy

MetricTrustUse it for
Amount spentHighestThe most honest proxy: where the algorithm is confidently spending is where it is finding efficiency.
ROAS / CPAHigh (with caveats)Outcome, but only with a 7-day click window and existing customers excluded. Read at ad set or campaign level.
CTR / CPCMediumLeading indicators of interest and creative pull; good for early reads.
Hook / hold rateDiagnostic onlyWhy a creative is or is not working. Never an optimisation target.

Why ad-level ROAS lies

Last-click attribution gives the sale to the final ad the buyer touched. But buying is a sequence: an ad that made someone problem-aware does real work and gets almost no last-click credit. So an upper-funnel ad can show a weak ad-level ROAS while actually driving revenue. Judge ROAS where it is more reliable, at the ad set and campaign level, not ad by ad.

Diagnostic vs optimisation metrics

Hook rate and hold rate are diagnostics. they tell you whether the creative earns and keeps attention, which helps you fix it. They are not goals. Optimise an account toward hook rate and you will make attention-grabbing ads that do not sell. Optimise toward profit, and use hook and hold rate to understand the why. (See hook rate vs hold rate.)

The "don't touch it" rule

If an ad set is hitting its KPI, do not switch off an ad inside it just because its individual ROAS looks weak. That ad may be doing sequencing or upper-funnel work the algorithm is using deliberately, and pulling it can drop the whole ad set. Judge the ad set, not the ad.

Read up the account, not down

Reliability climbs as you move up the structure: a single ad's numbers are noisy, an ad set's are steadier, a campaign's steadier still. When a number surprises you, zoom out a level before you act on it. For how this fits the wider creative system, see performance creative strategy, score your own setup with the creative scorecard, or book a free creative audit to have your account read properly.

Frequently asked questions

Which ad metrics are most reliable?
In order of reliability: amount spent, then ROAS or CPA (with a 7-day click window and existing customers excluded), then CTR and CPC as leading indicators, then hook rate and hold rate as diagnostics only. Reliability also rises as you move from ad to ad set to campaign level.
Why is ad-level ROAS misleading?
Because of last-click attribution and sequencing. An ad that does upper-funnel work gets little last-click credit even when it drives the eventual sale, so its ad-level ROAS understates its real contribution. Read ROAS at the ad set or campaign level instead.
Should I turn off an ad with low ROAS?
Not if it sits inside an ad set that is hitting its KPI. That ad may be doing sequencing work the algorithm is using on purpose. Judge the ad set, not the individual ad, before cutting.

Think creative is your bottleneck?

We brief from $250M+ in tracked ad spend and put first drafts in your hands in 48 hours. Book a free creative audit and we will show you where your account is leaking.

Book a free creative audit