Ad creative metrics contradict each other constantly, and the reason is almost always the denominator. Two numbers describing the same advert can point in opposite directions because one divides by people and the other divides by times shown.
Once you know which is which, the contradictions stop being confusing and start being informative.
This is for you if your creative reporting contains several numbers that seem to disagree about the same advert.
The six ad creative metrics worth knowing
Each one answers a different question, and the denominator is the question.
Click-through rate. Clicks divided by impressions. Measures whether the advert prompted action, per time shown.
Cost per click. Spend divided by clicks. Measures what attention costs, and moves with auction pressure as much as with creative quality.
Frequency. Impressions divided by reach. Measures how often the same people are seeing it, and it is the number that quietly distorts several of the others.
Hold rate. Views meeting a retention threshold, divided by impressions. Measures whether video kept people.
Cost per thousand impressions. Spend divided by impressions, times a thousand. Measures what the auction is charging you to be seen, which is partly creative and largely market.
Cost per result. Spend divided by outcomes. The only one that pays anybody.
Why the denominator decides the answer
Reach counts people. Impressions count times shown. That distinction produces most of the confusion in creative reporting.
As a campaign runs, frequency rises. The same people see the advert repeatedly, so impressions grow while reach does not.
Every metric divided by impressions therefore falls automatically, whether or not anything about the creative changed. Click-through rate declines, hold rate declines, and the report looks like creative decay.
Some of it is decay. Some of it is arithmetic. Separating the two requires looking at frequency alongside, and almost nobody puts frequency next to the metric it is distorting.
The practical rule is to check frequency before diagnosing a declining rate. If frequency climbed and reach is flat, the denominator moved and the creative may be fine.
The account that proves traffic metrics lie
This is the clearest example I have, and it cost about a thousand dollars to acquire.
On a US flag and banner ecommerce account, one ad set targeted a neighbouring state. It produced the highest reach in the entire account at 17,086, and the most link clicks at 542.
Every traffic metric said it was the winner. Reach, clicks, engagement, all top of the account.
It returned 0.25 on ad spend. Three purchases at roughly $341 each, about $1,027 spent before it was dropped. It was the only genuine loser in the account.
The creative was not the problem and neither was the audience’s interest. The client manufactures locally and delivers same day in its home state, so people in the neighbouring state saw an offer that was not really available to them. Interest was high, fulfilment was not.
No traffic metric could have told me that. Only revenue did.
The ad creative metrics hierarchy that prevents wrong decisions
Three tiers, and keeping them straight matters more than any individual number.
Cost per result decides. It is the only metric attached to money, and it overrules everything above it.
The three Meta rankings explain. Quality, engagement rate and conversion rate separate a creative problem from an audience problem from an offer problem, which is the diagnostic step.
Everything else explains the explanation. Video metrics, click-through rate, cost per thousand. Useful when something needs understanding, not useful as targets.
Most bad decisions I see from creative reporting come from inverting this, where an upstream metric with no reference point overrules an outcome that is working. Agreeing the hierarchy in writing before anybody reads a report is what prevents that in the moment.
Where ad creative metrics comparisons quietly break
Two conditions have to match or the comparison measures something other than the creative.
Placement is the big one. Feed, stories and reels produce completely different rates for identical creative, so an account-level average describes none of them. Compare within a placement.
Time is the second. Two creatives run in different weeks were exposed to different auction pressure, different competitor activity and different seasonal demand. That is a comparison of weeks wearing the appearance of a comparison of adverts.
Both get violated routinely, usually because the report defaults to an account-level view and nobody changes it.
Which ad creative metrics you can stop watching
Most of them, most of the time.
For a typical account, cost per result plus the three rankings plus one retention measure for video covers the decisions you actually make. The rest are diagnostics, worth pulling when something needs explaining.
Watching everything continuously has a specific cost beyond time. More numbers on a dashboard means more opportunities for one of them to move enough to prompt an intervention, and most of those interventions are noise-driven.
I would rather review four numbers weekly and understand them than twelve and react to them.
The exception is a period where something is genuinely wrong. Then pull everything, because the diagnostic value is real and the answer usually sits in a metric you do not normally look at.
How ad creative metrics should appear in a report
One layout decision removes most of the confusion, and it costs nothing.
Put the deciding number first and the diagnostics after it, visually separated. Most reporting presents twelve numbers as a grid of equals, which invites whoever reads it to pick whichever one supports the point they wanted to make.
Then put frequency immediately beside any rate, so a falling percentage can be read against a growing denominator in the same glance.
And label the period and placement at the top rather than leaving them implied. A comparison across placements is a different claim from a comparison within one, and the reader cannot tell which they are looking at unless somebody says.
None of that is sophisticated. It is the difference between a report that gets interpreted correctly and one that produces a monthly argument about which number is real.
The absent benchmarks, and what to do about it
Several of these have no published benchmark from any source with a disclosed method.
That includes hook rate, hold rate and thumbstop rate, where every figure circulating comes from vendor content citing other vendor content. It also includes fatigue frequency thresholds, where Meta analysed 2,439 Brand Lift campaigns and deliberately published no threshold at all.
So for those metrics, comparison against an external standard is not available. Compare your creatives against each other, on matched conditions, and treat any quoted target as unsourced until somebody shows you otherwise.
That constraint is less limiting than it sounds. The question you actually have is which of these adverts to keep running, and relative comparison answers it completely.
What to change this week
Three steps.
Put frequency next to any rate metric you track, so you can tell a declining creative from a growing denominator.
Then write down which single number decides in your account, and share it with whoever reads the reports, before the next review rather than during an argument.
Then check that your creative comparisons are within a placement rather than across all of them.
Hook rate has no published benchmark and hold rate tells you more cover the two video metrics in detail, fund the ad set and read what the algorithm chose is the position underneath, and statistical significance in ad testing covers how large a difference has to be before it is real. Getting the conversion right is what makes cost per result trustworthy, which is web development work as much as advertising. To have your creative reporting reviewed, book a teardown.