How to analyze Meta Ads creative performance
Creative analysis is about locating the failure, not scoring the ad. A video has three places it can lose someone: the opening that fails to stop the scroll, the body that fails to hold them, and the call to action that fails to move them. Each produces a different metric signature and each needs a different fix, so a single CTR number tells you something is wrong without telling you what to do about it.
To analyze creative performance, separate:
- Hook rate — are people stopping at all
- Hold rate — do they stay once they have stopped
- CTR — do they act once they have watched
- CPM — what the audience costs, as context for all three
- The conversion metric matching the campaign objective
- Format — image, video, carousel and catalog behave differently
- The pattern repeating across the winners
Match the signature to the fix
Four profiles cover most video creative. Find the one that matches and the fix is specific rather than 'test new creative'.
| Hook | Hold | CTR | What is wrong | The fix |
|---|---|---|---|---|
| Good | Poor | Average | The opening works; the body loses them | Keep the hook. Rebuild from 3 seconds on — value earlier, more visual variety |
| Average | Average | Average | Nothing stands out anywhere | Fix the hook first. If they do not stop, nothing else matters |
| Poor | Any | Any | People are not stopping at all | Rebuild the opening completely against your best-performing hooks |
| Good | Good | Poor | They watch it all and do not click | The CTA. Check comments first, then strengthen the ask |
The analysis
Step 1
Set the comparison set
Restrict to the ads carrying the first 90% of cumulative spend, filtered to active entities, and establish the account's own 90-day averages for hold rate and CPM. Creative metrics are only meaningful relative to what this account normally achieves.
Step 2
Compute the video funnel correctly
Hook rate is 3-second views divided by video plays. Hold rate is thruplays divided by 3-second views. The 3-second figure comes from the video view action — not from the 25%-watched metric, which measures something different and will quietly give you a wrong hook rate.
Our working thresholds: hook rate at or above 40% is good and under 25% poor; CTR at or above 1.25% is good and under 0.65% poor. Hold rate is judged against the account's own 90-day average, because it varies enormously by content length and format. These are our numbers, not an industry benchmark.
Step 3
Classify each ad into a failure profile
Using the table above, place each ad in the spend set into one of the four profiles. This is the step that turns a creative report into a production brief, because each profile points at a specific part of the asset rather than at the asset as a whole.
Step 4
Analyse by format, because the levers differ
Formats fail differently and expose different controls.
- Single image — no funnel to split. High CPM with weak return means an expensive audience that is not converting; low CTR with weak return means the creative is not landing, so check comments before assuming it is the concept.
- Single video — apply the four profiles above.
- Advantage+ catalog — Meta does not report conversions per product, so use spend and CTR as the proxy. Weak products are removed from the set rather than optimised within it.
Step 5
Check comments before blaming the creative
An ad with good watch-through and poor CTR sometimes has a comment section working against it. Negative sentiment, unanswered questions about shipping or sizing, or complaints visible under the ad all suppress clicks regardless of how good the creative is. This takes a minute and saves a pointless rebuild.
Step 6
Find the pattern across the winners
Individual results are less useful than what repeats. Look across the strong performers for the common angle, the common format, the kind of opening that recurs, and the copy structure. Three winners sharing a hook style is a brief; one winner is an anecdote.
- Which selling point appears in more than one winner?
- Is one format carrying the account?
- What kind of opening appears in the ads with strong hook rates?
- Does the copy lead with a problem or a benefit, and which wins here?
Step 7
Turn the diagnosis into the next variation
What you build next depends on what worked. If an ad performs well, vary the copy first, then the angle, then the format — smallest change first, so you learn something from each test. If it performs poorly, the profile tells you which part to rebuild rather than starting over.
A worked example
Illustrative example. The numbers below are made up to show the method. They are not a client account and not a benchmark.
Four video ads in the spend set, with the account's 90-day average hold rate at 31%.
| Ad | Hook rate | Hold rate | CTR | Profile |
|---|---|---|---|---|
| Ad A | 46% | 18% | 0.9% | Good hook, poor hold |
| Ad B | 31% | 30% | 0.8% | Average everything |
| Ad C | 19% | 34% | 0.7% | Poor hook |
| Ad D | 48% | 41% | 0.5% | Good hook and hold, poor CTR |
All four have weak CTR, and all four need different work. A single CTR column would have produced one undifferentiated verdict — "the creative is not working" — and four wasted rebuilds.
Ad A's opening is its strongest asset. Hook rate 46% against a hold rate well under the account's 31% average: keep the first three seconds and rebuild everything after them.
Ad C is the opposite. The content holds people better than the account average once they stop, but only 19% stop. The body is working; the opening is throwing away the audience.
Ad D is not a creative problem at all. It stops people and holds them better than anything else in the set, then fails to convert attention into clicks. That is a CTA and offer question — and worth checking the comments before touching the creative.
What not to conclude too quickly
Creative analysis goes wrong in predictable ways:
- Judging on CTR alone. It compresses three different failures into one number.
- Computing hook rate from the 25%-watched metric. It measures something else and produces a confidently wrong figure.
- Comparing hold rate to a published benchmark. It varies hugely with content length — your own account average is the only useful reference.
- Ranking creative without a volume floor. Below 3–5 conversions, the ranking is noise.
- Assuming poor CTR means poor creative. Check the comments and the offer first.
- Applying ROAS to video view or awareness campaigns. Judge each against its own objective.
- Treating a catalog ad like a single creative. There is no per-product conversion data, so spend and CTR are proxies rather than answers.
How to automate creative analysis
Doing this properly means pulling ad-level video metrics, computing two derived rates per ad, holding them against a 90-day account baseline, classifying each ad into a profile, and then reading the creative itself to find what the winners share. Repeated every time new creative enters rotation.
GoMarble connects creative attributes — hooks, formats, angles — back to performance, so the output is which patterns are working rather than a table of rates.
The shape of it
- Connect the ad account
- Ask for a creative review, or schedule one
- Hook, hold and CTR are computed per ad against the account baseline
- Each ad is classified so the fix is specific, not generic
- The recurring pattern across winners comes back as a brief
The deliverable is what to make next, not a scoreboard of what you already made.
Creative analysis checklist
- ☐ Restrict to the first 90% of spend, active only
- ☐ Establish the account's 90-day hold rate and CPM averages
- ☐ Compute hook rate from 3-second views over video plays
- ☐ Compute hold rate from thruplays over 3-second views
- ☐ Do not use the 25%-watched metric for hook rate
- ☐ Classify each ad into one of the four profiles
- ☐ Apply the right method per format
- ☐ Use spend and CTR as proxies on catalog ads
- ☐ Check comments before concluding the creative is wrong
- ☐ Apply a 3–5 conversion floor before ranking
- ☐ Identify the repeating angle, format and hook across winners
- ☐ Choose the variation type that matches the diagnosis
Questions people actually ask
What is a good hook rate on Meta Ads?
What is the difference between hook rate and hold rate?
Why is my CTR low when the video performs well?
How do I analyse Advantage+ catalog ad creative?
How many creatives should I test at once?
Where this method comes from
The four diagnostic profiles, the hook and hold formulas, the format-specific decision framework and the variation strategy on this page come from the creative analysis methodology GoMarble's agents run against live Meta accounts — including the explicit instruction to compute three-second views from the video view action rather than the 25%-watched metric, and to judge hold rate against an account-level 90-day baseline. Benchmark figures are ours rather than industry standards.
Related
Get the brief, not the scoreboard.
Connect Meta and GoMarble connects hooks, formats and angles to performance — so you know what to make next.