Which Creative Is Working and Which Is Not: The Diagnostic Checklist

Key Takeaways

• The broken metric tells you the broken element. Each metric isolates one stage of the creative's job, so the metric that breaks first points directly at what to fix.
• Read the funnel top-down. CPM, then hook rate, then hold rate, then CTR, then conversion rate — because a break upstream contaminates every metric below it, so the first break is the real problem.
• A low hook rate is a hook problem, not a creative problem in general. If people scroll past in the first seconds, nothing after the hook matters — fix the first frame before anything else.
• A good hook with a low hold rate means the middle fails — the pacing, the payoff, the content between the hook and the ask lost them. The hook wrote a cheque the body did not cash.
• A good hold with a low CTR means the ask fails — the call to action, the offer clarity, the reason to click. They watched and were not moved to act.
• A good CTR with a low conversion rate usually is not the creative at all — it points to the landing page, the offer, or an audience-message mismatch. Do not keep editing a creative that already did its job.
• Judge nothing on too little data. Every one of these metrics is noise on small samples, so the checklist only works once each metric has enough data to be real.

1. The Short Answer: Read the Metrics as a Funnel

To know which creative is working and which is not — and why — stop looking at the metrics as a scorecard and read them as a funnel. Each metric isolates one stage of the creative's job: getting served (CPM), stopping the scroll (hook rate), holding attention (hold rate), earning the click (CTR and engagement rate), and driving the action (conversion rate). Read them top-down, and the first metric that breaks tells you exactly which part of the creative is failing.

This is the diagnostic move that turns a wall of numbers into an answer. A creative with a good hook rate and a poor hold rate is not 'a bad creative' — it is a creative whose hook works and whose middle fails, which points you at the pacing and payoff, not the opening. A creative with a great hold and a poor click-through is not 'a bad creative' either — it is one that held attention and failed to convert that attention into an ask, which points you at the call to action. The metric that breaks names the element to fix.

The critical rule that makes this work is to read top-down, because a break upstream poisons everything downstream. A creative with a broken hook has a low hold rate, low CTR and low conversion rate too — but not because the middle, the CTA and the offer are bad; because almost nobody got past the hook to experience them. The first break is the real problem, and the metrics below it are contaminated. Diagnose the first break, fix it, and re-read. This guide gives you the funnel, an interactive diagnostic tool, a per-element checklist, and the honest test of when the problem is not the creative at all. It pairs with the [Meta Ads creative strategy](/resource/blogs/meta-ads-creative-strategy) work on making better creative and the [automated creative testing guide](/guides/automate-meta-creative-testing-n8n-guide) on testing it at scale.

• AEO Quick Answer: read CPM, hook rate, hold rate, CTR and conversion rate as a funnel; the first metric that breaks names the failing element.
• The broken metric tells you the broken element — each metric isolates one stage of the creative's job.
• Read top-down: an upstream break contaminates every metric below it, so the first break is the real problem.

2. The Core Principle: The Broken Metric Names the Broken Element

The whole checklist rests on one principle worth stating clearly, because most creative analysis ignores it: each performance metric corresponds to a specific stage of what the creative has to do, so a metric's failure is a specific diagnosis, not a general verdict.

The mistake most people make. They look at a creative's headline result — its cost per acquisition or its return — decide it is 'good' or 'bad', and either scale it or kill it. This tells them what happened and nothing about why, so they learn nothing transferable: they cannot say what made it work or what to change, and their next creative is another shot in the dark. A single outcome metric is a verdict without a diagnosis.

The diagnostic alternative. The intermediate metrics — hook rate, hold rate, CTR — are not just steps toward the outcome; they are a diagnostic sequence. Each one isolates a different part of the creative, so reading them in order localises the failure to a specific element. This is exactly how a doctor works from symptoms to a diagnosis rather than declaring a patient 'unwell' — the pattern of symptoms points to the specific problem. The pattern of creative metrics points to the specific element.

Why this makes you better over time. When you diagnose the specific failing element rather than judging the whole creative, you learn something transferable: this hook style fails, this pacing works, this CTA converts. That knowledge compounds into better creative, whereas 'that one was bad' teaches you nothing. Diagnostic reading is what turns creative testing from a slot machine into a learning system, which is the entire point of testing.

The stages and their metrics, in order. Delivery (does the auction serve it efficiently?) shows in CPM. Hook (does it stop the scroll?) shows in hook rate or thumbstop rate. Hold (does it keep attention?) shows in hold rate and average watch time. Click (does it earn the action?) shows in CTR and engagement rate. Convert (does the click become a customer?) shows in conversion rate. Each stage, each metric, each a specific diagnosis — and the rest of this guide walks them one by one.

• Each metric corresponds to a specific stage of the creative's job, so its failure is a specific diagnosis.
• A single outcome metric (CPA, ROAS) is a verdict without a diagnosis — it teaches nothing transferable.
• The intermediate metrics are a diagnostic sequence, like symptoms pointing a doctor to a diagnosis.
• Diagnosing the specific failing element produces transferable learning that compounds into better creative.

3. The Creative Attention Funnel

Here is the funnel in full, the backbone of the whole diagnostic. Every ad creative has to move a person through five stages in order, and each stage has a metric that measures whether the creative cleared it. Use the interactive tool above to walk your own creative through it: mark each stage strong or weak, and it identifies the first break and the element to fix.

Stage one: delivery. Before anyone sees your creative, the auction has to serve it, and how efficiently it serves shows in CPM — the cost per thousand impressions. A high CPM means the platform is charging you more to show the creative, which often signals low relevance or quality (the algorithm deprioritises creative people ignore) or simply a competitive, expensive audience. Delivery is the stage before the creative even gets its chance, and CPM is the earliest warning sign.

Stage two: the hook. In the first seconds, the creative has to stop the scroll — earn the viewer's attention against everything else competing for it. This shows in the hook rate (often measured as three-second video views over impressions) or thumbstop rate. A low hook rate means the opening fails: people scroll past before the creative says anything. This is the highest-leverage stage, because a failed hook means nothing downstream gets a chance.

Stage three: the hold. Having stopped the scroll, the creative has to keep the attention it caught — carry the viewer from the hook through the content to the point of the ask. This shows in the hold rate (retention deeper into the video, or average watch time). A good hook rate with a low hold rate means the creative grabbed attention and then lost it: the middle is slow, the payoff never comes, the content does not deliver on the hook's promise.

Stage four: the click. The creative has to convert held attention into action — give the viewer a reason and a way to click. This shows in CTR (click-through rate) and engagement rate. A good hold rate with a low CTR means the creative held attention but failed to move it to action: the call to action is weak, unclear or absent, or the offer is not compelling enough to act on.

Stage five: the conversion. The click has to become a customer. This shows in the conversion rate. But this stage is where the creative's job ends and the landing page and offer begin, so a good CTR with a low conversion rate often means the problem is downstream of the creative entirely — which is a crucial diagnostic distinction covered later. The creative got them to click; something after the click failed.

• Delivery — CPM: how efficiently the auction serves the creative; high CPM signals low relevance or a competitive audience.
• Hook — hook rate / thumbstop: does the opening stop the scroll; the highest-leverage stage.
• Hold — hold rate / watch time: does it keep the attention it caught through to the ask.
• Click — CTR / engagement rate: does it convert held attention into action.
• Convert — conversion rate: does the click become a customer; often points downstream of the creative.

4. Metric by Metric: What Each One Actually Tells You

Each metric deserves a precise reading, because a metric misread is a diagnosis misdirected. Here is what each tells you, what good and bad look like, and why.

CPM (cost per thousand impressions). What it tells you: how expensive it is to get your creative shown, which reflects the auction — your relevance and quality as the algorithm judges them, and the competitiveness of your audience. A rising or high CPM can mean the algorithm is deprioritising your creative because people ignore it (a creative-quality signal), or simply that you are targeting an expensive, contested audience (not a creative problem). Read CPM as context: it flags whether a delivery or relevance problem is inflating your costs before the creative even performs.

Hook rate (three-second views over impressions, or thumbstop rate). What it tells you: whether the opening stops the scroll. This is the purest measure of the hook, because it captures only the first moments, before anything else in the creative has a chance to matter. A low hook rate is an unambiguous verdict on the opening — the first frame, the first line, the first movement failed to earn attention. It is the single most diagnostic metric because it is the least contaminated: nothing downstream affects it.

Hold rate (retention deeper in, or average watch time relative to length). What it tells you: whether the creative keeps the attention the hook caught. Read it always relative to the hook rate — hold rate in isolation is contaminated by the hook (a creative that hooked the wrong people has low hold for reasons of audience, not content). A good hook rate with a poor hold rate is a clean signal that the middle fails; a poor hook rate makes hold rate unreadable, because too few people got past the hook to measure.

CTR (click-through rate) and engagement rate. What they tell you: whether the creative converts attention into action. CTR measures clicks; engagement rate measures broader interaction (likes, comments, shares). Read CTR relative to hold rate — a good hold and poor CTR isolates the ask (the CTA, the offer's pull), while a poor hold makes CTR unreadable. Engagement rate is a softer signal: high engagement with low CTR can mean the creative resonates emotionally but does not drive the specific action, which is useful for brand and a warning for direct response.

Conversion rate (CVR). What it tells you: whether the click becomes the outcome you want. This is where the creative hands off, so CVR read against CTR is the key handoff diagnostic: a good CTR with a poor CVR points past the creative to the landing page, the offer, or an audience-message mismatch, because the creative did its job (it earned the click) and something after the click did not. A poor CTR makes CVR unreadable, because too few clicked to measure conversion. Never blame the creative for a conversion problem when the click-through was strong.

• CPM: how expensive it is to be shown — reads as relevance/quality and audience competitiveness, before the creative performs.
• Hook rate: the purest, least-contaminated metric — an unambiguous verdict on the opening.
• Hold rate: read relative to hook rate; a good hook with poor hold cleanly isolates the middle.
• CTR / engagement: read relative to hold; a good hold with poor CTR isolates the ask.
• Conversion rate: read against CTR; a good CTR with poor CVR points past the creative to the landing or offer.

5. The Diagnostic Procedure: Read Top-Down, Isolate the First Break

The metrics only diagnose correctly when read in the right order, because of contamination. Here is the procedure that produces a reliable diagnosis rather than a confused one.

Step one: check you have enough data. Before reading anything, confirm each metric is based on enough data to be real — a hook rate on two hundred impressions is noise, a conversion rate on three clicks is meaningless. The entire procedure is invalid on thin data, so this check comes first, always. More on the data threshold later, but never skip it.

Step two: read from the top. Start at delivery (CPM) and read down: CPM, hook rate, hold rate, CTR, conversion rate. Do not jump to the outcome metric and work backward; read the funnel in order, because each metric's validity depends on the ones above it being sound.

Step three: find the first break. The first metric that is meaningfully below benchmark is the diagnosis. If the hook rate is the first thing that breaks, the hook is the problem — full stop, for now — and you do not diagnose the hold rate or CTR yet, because they are contaminated by the broken hook. The first break is the real, actionable problem; everything below it is downstream noise until the first break is fixed.

Step four: fix the first break, then re-read. Fix only the first broken element, relaunch, and read the funnel again. Now a different, genuine problem may surface downstream — the one that was hidden behind the first break. Diagnosis is iterative: fix the first break, re-read, fix the next real break, re-read. Trying to fix everything at once, based on a contaminated reading, changes things that were not actually broken and muddies the next diagnosis.

The contamination principle, made concrete. Imagine a creative with a hook rate at half of benchmark, and a hold rate, CTR and conversion rate all also poor. The novice reads this as 'everything is broken' and rebuilds the whole creative. The diagnostician reads it as 'the hook is broken, and the rest is unreadable until it is fixed' — because the low hold, CTR and conversion are almost certainly just the consequence of few people getting past the hook, not independent failures. Fix the hook, and the downstream metrics often recover on their own, revealing that they were never broken. This is why top-down, first-break reading is the whole method: it stops you from fixing symptoms of a single upstream cause.

• Step 1: confirm enough data — the procedure is invalid on thin data.
• Step 2: read from the top (CPM, hook, hold, CTR, CVR), not from the outcome backward.
• Step 3: the first metric meaningfully below benchmark is the diagnosis; metrics below it are contaminated.
• Step 4: fix only the first break, relaunch, re-read — diagnosis is iterative.
• Contamination: a broken hook makes everything downstream look broken; fix it and they often recover on their own.

6. Element Checklist: The Hook

The hook is the first seconds, and it is the highest-leverage element because a failed hook makes everything else irrelevant. It shows up in the hook rate. Check these when the hook rate is your first break.

The first frame or first line. Does it earn attention instantly, before the viewer decides to scroll? A hook that takes two seconds to get going has already lost, because the decision to keep watching or scroll happens almost immediately. Check whether the very first moment — the opening frame of a video, the first line of copy, the top of a static — is doing work or warming up.

Pattern interrupt and relevance. Does the opening stand out from the feed (a pattern interrupt) and signal relevance to the viewer (this is for you, about a problem you have)? A hook that blends into the feed or does not signal relevance fails on both counts — it neither stops the scroll nor gives a reason to keep watching. The best hooks do both: they are visually or tonally arresting and they immediately signal 'this is relevant to you'.

Curiosity and tension. Does the hook open a loop — a question, a tension, a surprising claim — that the viewer needs the rest of the creative to resolve? Hooks that give everything away in the first frame have no pull; hooks that open a compelling question or tension create a reason to keep watching. Check whether your opening creates a curiosity gap or resolves it too soon.

The audience match. Does the hook speak to the specific audience and the belief or problem you are targeting? A generic hook underperforms a specific one because specificity signals relevance — a hook that names the viewer's situation or problem stops the right people, while a broad hook fails to stop anyone in particular. Check whether the hook is calibrated to who you are actually trying to reach.

The format-native execution. Does the hook work in the format and placement it runs in — sound-off if the placement is sound-off, thumb-stopping in a fast feed, legible on a small screen? A hook designed for one context that runs in another (a sound-dependent hook in a sound-off placement) fails not on concept but on execution. Check that the hook is built for where it actually appears, which connects to the placement checklist below.

• The first frame or line must earn attention instantly — no warm-up; the scroll decision is immediate.
• Pattern interrupt plus relevance: stand out from the feed AND signal 'this is for you'.
• Open a curiosity gap or tension the rest of the creative resolves; do not give it all away up front.
• Match the specific audience and their problem — specificity signals relevance and stops the right people.
• Execute native to the format and placement (sound-off, small screen, fast feed).

7. Element Checklist: Content and Copy

The content and copy are the body — everything between the hook and the ask — and they show up in the hold rate (does the body keep attention) and contribute to CTR (does the message build to a compelling ask). Check these when the hold rate is your first break with a good hook.

The payoff on the hook's promise. Does the body deliver what the hook promised? A hook that opens a question or tension writes a cheque the body must cash; if the body wanders, delays, or never resolves the hook, attention drains. The most common hold-rate failure is a great hook followed by a body that does not pay it off — the viewer stays a moment expecting the resolution, does not get it, and leaves.

Pacing and momentum. Does the content keep moving, or does it sag? Attention is fragile, and a slow middle — a long setup, a repeated point, a stretch where nothing happens — is where hold rate drops. Check for the sag: the point in the creative where the pace slackens and viewers leave. Tightening the pacing, cutting the slow stretch, is often the single highest-leverage hold-rate fix.

The message clarity. Is the core message clear and easy to follow, or does the viewer have to work to understand it? Confused viewers leave. Check whether someone seeing it once, at feed speed, gets the message — clarity holds attention, confusion loses it. This is especially important because most viewers give the creative a fraction of their attention, so a message that requires focus to parse loses the distracted majority.

The copy that drives to the ask. Does the copy build a case that makes the eventual click feel worthwhile? Copy is not just information; it is persuasion toward the action. Check whether the content builds desire and a reason to act, so that by the time the ask arrives, the viewer is motivated. Copy that informs but does not persuade holds attention and fails to earn the click, which shows as good hold and poor CTR.

Length appropriate to the message and placement. Is the creative the right length — long enough to make its case, short enough not to lose people? Too long and the hold rate decays as people leave before the end; too short and there is not enough to build to a compelling ask. Check that the length matches what the message needs and what the placement supports, rather than defaulting to a length out of habit.

• Pay off the hook's promise — the most common hold-rate failure is a great hook the body does not cash.
• Keep pacing and momentum — find and cut the sag where viewers leave.
• Make the message clear at feed speed — confused viewers leave; clarity holds.
• Copy should persuade toward the ask, not just inform — informing without persuading shows as good hold, poor CTR.
• Length appropriate to the message and placement — not too long (decay) or too short (weak ask).

8. Element Checklist: Format, Size and Placement

Format, size and placement are about the creative's physical fit for where it runs, and they show up across the funnel — a wrong format can hurt CPM (poor delivery), hook rate (not built for the placement) and hold (uncomfortable to consume). Check these when metrics are weak across the board or when a creative underperforms in specific placements.

Aspect ratio and dimensions for the placement. Is the creative the right shape for where it runs — full-screen vertical for stories and reels, square or vertical for feed, the right dimensions for each surface? A creative in the wrong aspect ratio is letterboxed, cropped awkwardly, or shows dead space, all of which hurt performance. Check that each placement gets a creative built for its dimensions, not one format stretched across all of them.

Placement-appropriate design. Does the creative work in the specific placement's context — sound-off where sound is off by default, thumb-stopping in a fast feed, with safe zones respected where interface elements overlay the creative? A creative that ignores the placement's realities (text in a zone the interface covers, a sound-dependent message in a sound-off feed) fails on execution regardless of concept. Check the creative against the actual placement, not a clean preview.

Per-placement performance. Are you looking at performance by placement, not just in aggregate? A blended metric hides that a creative crushing it in one placement is bombing in another, and the fix (exclude the bad placement, or make a placement-specific version) is invisible in the aggregate. Check placement breakdowns — this is one of the most common hidden problems, a creative averaging out to mediocre because it is great in one surface and terrible in another.

Text and legibility. Is text legible at the size the creative actually appears — on a small phone screen, at feed-scroll speed? Text that is too small, too dense, or poorly contrasted is unreadable in practice even if it looks fine in a large preview, and unreadable text is wasted. Check legibility on an actual device at actual size, not on a desktop preview.

The mobile-first reality. Is the creative designed for the small mobile screen where most impressions happen? A creative designed on a large monitor and reviewed there can look great and perform poorly because the mobile reality — small, fast, sound-off, thumb-scrolled — is different. Check every creative on a phone, because that is where it lives, and the gap between how it looks on a monitor and how it performs on a phone is a common, avoidable source of underperformance.

• Aspect ratio and dimensions built per placement — not one format stretched across all surfaces.
• Placement-appropriate design: sound-off, safe zones, thumb-stopping in a fast feed.
• Check per-placement performance — a blended metric hides a creative great in one surface and terrible in another.
• Text legible at actual size and speed on a phone, not just in a large preview.
• Design and review mobile-first — the small, fast, sound-off reality is where most impressions happen.

9. Element Checklist: Colour and Visual

Colour and visual design affect attention and clarity, and they show up mainly in hook rate (visual arrest stops the scroll) and hold (visual clarity keeps attention). Check these as a contributor when the hook or hold is weak, and as a differentiator between otherwise similar creatives.

Visual contrast and arrest. Does the creative stand out visually against the feed? High contrast, unexpected colour, and visual interest help stop the scroll; a creative that blends into the feed's typical palette is easier to scroll past. Check whether the creative has visual pop against the surrounding feed — not garishness, but enough contrast and interest to earn a stop. This contributes directly to the hook.

Colour and brand recognition. Does the creative use colour in a way that builds recognition over time? Consistent, distinctive colour helps people recognise your brand across creatives, which compounds — a viewer who has seen your distinctive palette before recognises the next ad faster. Check whether your creatives are visually coherent enough to build recognition, without being so rigid they all look identical and fatigue faster.

Visual clarity and focus. Does the visual guide the eye to what matters, or is it cluttered? A cluttered creative where the eye does not know where to go loses attention; a clear one with a visual focal point holds it. Check whether the visual has a clear focus and hierarchy, or whether competing elements confuse the eye. Visual clutter is a common, subtle hold-rate drain.

Colour and emotional tone. Does the colour palette match the emotional tone of the message? Colour carries emotional signal — warm, cool, energetic, calm — and a mismatch between the palette and the message's tone creates subtle dissonance. Check whether the colour supports the feeling the creative is going for, because colour that fights the message weakens it.

Accessibility and contrast for legibility. Is there enough contrast for text and key elements to be legible, including for viewers with lower vision or in bright conditions? Poor contrast is both an accessibility failure and a performance one — low-contrast text is hard to read for everyone in real conditions. Check contrast against real viewing conditions, not an ideal screen.

• Visual contrast and arrest against the feed helps stop the scroll — contributes to the hook.
• Consistent distinctive colour builds brand recognition that compounds across creatives.
• Visual clarity and a clear focal point hold attention; clutter is a subtle hold-rate drain.
• Colour palette should match the message's emotional tone, not fight it.
• Enough contrast for legibility in real conditions — an accessibility and performance issue.

10. Element Checklist: Tone

Tone — the voice, attitude and emotional register of the creative — is subtle and powerful, affecting whether the creative resonates with its audience. It shows up diffusely across hook, hold and engagement, and it is often the difference between two creatives with identical elements that perform differently. Check tone when creatives are technically sound but underperform, or when engagement is high but CTR is not.

Audience match. Does the tone fit the audience you are targeting? A tone that resonates with one audience alienates another — playful versus serious, casual versus authoritative, aspirational versus practical. Check whether the tone matches how your specific audience wants to be spoken to, because a mismatched tone creates a subtle repulsion even when everything else is right. Tone is where a creative can be technically correct and still feel wrong to its audience.

Authenticity versus polish. Is the tone appropriately authentic for the platform and audience? On social especially, over-polished, obviously-advertising tone can underperform more authentic, native-feeling creative, because people scroll past what obviously reads as an ad. Check whether the tone feels native to the platform or announces itself as advertising — the balance between polish and authenticity is platform- and audience-specific, and getting it wrong is a common, subtle underperformance.

Emotional register. Does the tone evoke the right feeling for the desired action? Different actions are driven by different emotions — urgency, aspiration, relief, curiosity, trust — and the tone should evoke the one that drives your action. Check whether the emotional register matches the action: a creative going for an urgent purchase needs a different tone from one building long-term trust.

Consistency with the brand and the landing experience. Does the tone match your brand and, crucially, the experience the click leads to? A creative whose tone promises one thing and whose landing page delivers another creates a jarring handoff that hurts conversion — the tone set an expectation the destination broke. Check that the creative's tone is consistent with what happens after the click, because tone dissonance across the handoff is a hidden conversion killer.

The high-engagement, low-CTR tone signal. If engagement is high but CTR is low, tone is a prime suspect: the creative is emotionally resonating (people engage) but not driving action (they do not click), which often means the tone is entertaining or likeable but not motivating the specific ask. Check whether the tone is optimised for being liked versus for driving action — for direct response, a beloved creative that does not convert is a tone that entertained instead of persuaded.

• Tone must match the specific audience — a mismatch creates subtle repulsion even when everything else is right.
• Balance authenticity and polish for the platform — over-polished 'obviously an ad' tone often underperforms native-feeling creative.
• Emotional register should evoke the feeling that drives the desired action.
• Tone must be consistent with the post-click experience — dissonance across the handoff hurts conversion.
• High engagement with low CTR often signals a tone that entertains instead of persuades.

11. Element Checklist: Budget, CPM and Delivery

Budget and delivery are not creative elements in the design sense, but they shape whether a creative gets a fair chance and show up in CPM, frequency and the reliability of every other metric. Check these when metrics are unstable, when CPM is high, or when a creative never seems to get going.

Enough budget to exit learning and gather data. Does the creative have enough budget and time to be judged? A creative starved of budget never gathers enough data for its metrics to be real, and the platform's optimisation never stabilises. Judging a creative on too little spend is judging on noise, and under-budgeting is a common reason a genuinely good creative looks bad — it never got the fair read a stable budget provides. This connects to the [campaign budgeting guide](/guides/campaign-budgeting-guide) on funding tests above the viable threshold.

CPM as a relevance and audience signal. Is CPM telling you about the creative or the audience? A high CPM can mean the algorithm is deprioritising a low-relevance creative (a creative signal — improve the creative) or that the audience is simply competitive and expensive (not a creative signal — a targeting or economics matter). Check whether a high CPM tracks with a low hook rate (suggesting the creative is the problem) or with an inherently expensive audience (suggesting it is not). CPM read alongside hook rate distinguishes a creative-relevance problem from an audience-cost one.

Frequency and fatigue. Is the creative being shown too often to the same people? Rising frequency with declining performance is the signature of creative fatigue — the creative was fine, the audience has now seen it too many times. Check frequency before concluding a creative is bad: a creative whose metrics decayed over time on a rising frequency is fatigued, not flawed, and the fix is fresh creative, not fixing this one. This is the [creative fatigue](/resource/blogs/creative-fatigue-issue-scaling) dynamic.

Fair comparison conditions. Are you comparing creatives on fair terms — similar budgets, similar audiences, similar time periods? A creative that got most of the budget and another that got scraps are not comparable, however the metrics read. Check that a creative you are judging got a fair, comparable read, because unfair delivery conditions produce misleading comparisons that lead you to kill good creative and scale lucky ones.

The delivery-stability check. Is the creative delivering stably, or is it stuck in learning, being throttled, or delivering erratically? Unstable delivery makes every metric unreliable, so before diagnosing the creative, confirm it is actually getting a stable, sufficient delivery to be judged on. A creative that never got a stable delivery has no real metrics to diagnose, and mistaking a delivery problem for a creative problem is a common misdiagnosis.

• Enough budget and time to gather real data — judging on too little spend is judging on noise.
• CPM read alongside hook rate distinguishes a creative-relevance problem from an expensive-audience one.
• Check frequency before blaming the creative — decay on rising frequency is fatigue, not a flaw; the fix is fresh creative.
• Ensure fair comparison conditions — similar budget, audience and period; unfair delivery misleads.
• Confirm stable delivery before diagnosing — a creative stuck in learning has no real metrics to read.

12. Reading Metric Combinations

Individual metrics diagnose stages; combinations of metrics diagnose more precisely, because a pattern across metrics is more informative than any single one. Here are the most useful combinations and what they reveal.

Low hook rate, everything else unreadable. The hook is broken and nothing downstream can be diagnosed until it is fixed. Do not touch the body, CTA or offer — fix the hook, relaunch, re-read. This is the most common pattern and the most commonly over-diagnosed.

Good hook, low hold. The opening works and the middle fails. The hook stopped the scroll and the body lost them — diagnose pacing, payoff, clarity, and length. The hook wrote a cheque the body did not cash.

Good hook, good hold, low CTR. The creative held attention and failed to convert it to action. Diagnose the ask — the CTA's clarity and strength, the offer's pull, the reason to click. The creative did the hard part (earning attention) and fell at the last creative hurdle (the ask).

Good CTR, low conversion rate. The creative did its whole job and the problem is downstream. Diagnose the landing page, the offer, the click-to-landing message match, or an audience-message mismatch — but do not keep editing the creative, which already succeeded in earning the click. This is the single most important combination to read correctly, because it stops you wasting effort on a creative that is not the problem.

High engagement, low CTR. The creative resonates emotionally but does not drive the action. Diagnose tone (entertaining versus persuading) and the clarity and strength of the ask. This pattern is a warning for direct response and sometimes fine for brand — a beloved creative that does not convert has optimised for being liked over driving action.

High hook rate, low hold, and the wrong audience clicking. Sometimes a hook works too well in the wrong way — a sensational hook stops the scroll of people who are not your audience, who then do not hold, click or convert. Diagnose whether the hook is stopping the right people; a hook that stops everyone but holds no one may be attracting the wrong audience, which is a hook-relevance problem masquerading as a hold problem.

The principle across combinations: the pattern localises the problem more precisely than any single metric, and reading the pattern — especially the good-CTR-low-CVR pattern that exonerates the creative — is what separates precise diagnosis from flailing. Learn the common patterns and their diagnoses, and most creative problems become quickly legible.

• Low hook rate → hook broken, rest unreadable; fix the hook only.
• Good hook, low hold → the middle fails; diagnose pacing, payoff, clarity, length.
• Good hold, low CTR → the ask fails; diagnose the CTA and offer pull.
• Good CTR, low CVR → the problem is downstream of the creative; stop editing the creative.
• High engagement, low CTR → tone entertains instead of persuades; a direct-response warning.

13. When It Is Not the Creative At All

The most important diagnostic skill is knowing when the creative is not the problem, because endlessly editing a creative that already did its job is wasted effort and a common, expensive mistake. Several patterns point away from the creative.

Good CTR, poor conversion — look downstream. As established, a creative that earned the click has done its job; a poor conversion rate points to the landing page (slow, confusing, mismatched), the offer (not compelling, not what the creative promised), or an audience-message mismatch (the creative attracted people the offer does not suit). Diagnose these, not the creative, when the click-through was strong.

The audience problem masquerading as a creative problem. A great creative shown to the wrong audience performs poorly — not because the creative is bad, but because it is well-matched to a different audience than the one seeing it. If a creative that should work is underperforming, check whether it is reaching the audience it was built for. Fixing the audience can make an unchanged creative suddenly perform, which reveals the creative was never the problem.

The offer problem. Sometimes the creative and the landing page are fine and the offer itself is not compelling enough — the thing being offered, at the price and terms offered, does not motivate the action. No creative fixes a weak offer; a brilliant creative for an offer nobody wants earns clicks and no conversions. If conversion is poor across many good creatives, suspect the offer, not the creatives.

The product or market-fit problem. At the deepest level, if nothing converts across good creatives, good landing pages and reasonable offers, the problem may be that the market does not want the product at the terms available — which no amount of creative optimisation fixes. This is the hardest diagnosis to accept and the most important, because pouring creative effort into a product-market-fit problem is the most expensive form of avoiding the real issue.

The diagnostic discipline: exhaust the creative diagnosis, and when the metrics say the creative did its job (good hook, hold and CTR) but the outcome is still poor, look outward — to the landing page, the offer, the audience, and ultimately the product. The checklist's most valuable output is sometimes 'the creative is fine; the problem is elsewhere', which redirects effort to where it will actually help. A team that keeps editing creative on a downstream problem is optimising the one thing that is already working.

• Good CTR, poor conversion → look at the landing page, offer, or audience-message match, not the creative.
• A great creative shown to the wrong audience underperforms — fix the audience, not the creative.
• A weak offer converts poorly across all good creatives — no creative fixes an offer nobody wants.
• If nothing converts across good creatives, offers and pages, suspect product-market fit — the hardest, most important diagnosis.
• The checklist's most valuable output is sometimes 'the creative is fine; the problem is elsewhere'.

14. Statistical Honesty: Enough Data Before You Judge

Every diagnosis in this guide is worthless on thin data, because all these metrics are dominated by noise on small samples. This section is short and it is the one that most protects you from confident wrong conclusions.

Small samples lie. A hook rate on a few hundred impressions, a CTR on a handful of clicks, a conversion rate on a few conversions — these are noise, and they will show large, meaningless swings that look like signal. A creative can look like a winner or a loser on early data and be completely average once enough accumulates. Reading the diagnostic funnel on thin data produces a confident, wrong diagnosis, which is worse than no diagnosis because you act on it.

Each metric needs its own threshold. The metrics accumulate data at different rates — impressions fast, clicks slower, conversions slowest — so a creative can have a readable hook rate while its conversion rate is still noise. Judge each metric only when it individually has enough data, which means the upper-funnel metrics become readable before the lower-funnel ones. Do not diagnose a conversion rate on three conversions just because the hook rate on ten thousand impressions is solid.

The patience the funnel requires. Because the lower-funnel metrics are slowest to become real, the full diagnosis takes time to accumulate — you can diagnose the hook early and must wait for the conversion signal. This patience is uncomfortable and it is the difference between diagnosis and guessing. The temptation to call a creative's conversion performance on early clicks is the temptation to guess, and it is where creative testing programmes waste their spend.

The connection to testing rigour. This is the same discipline as the minimum-data threshold in automated creative testing — never judge a creative before its metrics are real. Whether you are diagnosing manually or building automated rules, the data threshold is the guardrail that keeps you from acting on noise. Build it into your habit: before every diagnosis, the question is 'is this metric based on enough data to be real', and if not, the answer is 'wait', not 'guess'.

The honest bottom line: the diagnostic funnel is powerful and it is only as good as the data under it. Give each metric enough data, judge each only when it is individually real, and be patient with the slow lower-funnel signal. Rush it, and you will confidently misdiagnose, kill good creative, scale lucky ones, and learn the wrong lessons — the exact opposite of what the diagnostic method is for.

• Small samples lie — thin-data metrics show large meaningless swings that look like signal.
• Each metric needs its own threshold; upper-funnel metrics become readable before lower-funnel ones.
• Diagnose the hook early; wait for the slow conversion signal — patience is the difference between diagnosis and guessing.
• Same discipline as automated testing's minimum-data threshold — never judge before the metric is real.
• Rushing it produces confident misdiagnosis: killed winners, scaled flukes, wrong lessons.

15. Common Mistakes, and What to Do Instead

Judging on the outcome metric alone. Good or bad tells you nothing transferable. Instead, read the diagnostic funnel to localise the failing element and learn something you can reuse.

Reading metrics out of order. A contaminated downstream metric misdirects you. Instead, read top-down and diagnose the first break, ignoring the contaminated metrics below it.

Fixing everything at once. You change things that were not broken and muddy the next diagnosis. Instead, fix only the first break, relaunch, and re-read.

Blaming the creative for a conversion problem. A good CTR means the creative did its job. Instead, when CTR is strong and conversion is poor, look downstream to the landing page, offer and audience.

Ignoring per-placement performance. A blended metric hides a creative great in one placement and terrible in another. Instead, check placement breakdowns.

Confusing fatigue with a flaw. Decay on rising frequency is fatigue, not a bad creative. Instead, check frequency before concluding, and refresh rather than fix.

Reviewing on a monitor. The creative lives on a small, fast, sound-off phone. Instead, review every creative on an actual phone at actual size.

Judging on thin data. Small samples lie confidently. Instead, judge each metric only when it individually has enough data to be real.

Comparing on unfair terms. A creative that got scraps of budget is not comparable to one that got most of it. Instead, compare on similar budgets, audiences and periods.

Never suspecting the audience, offer or product. You edit creative endlessly on a problem that is not the creative. Instead, when good creatives all fail to convert, look outward.

16. Putting It Together

Knowing which creative is working and why is not about staring at a scorecard of metrics — it is about reading those metrics as a funnel, where each one isolates a stage of the creative's job and the first one that breaks names the element to fix. CPM flags delivery and relevance, hook rate the opening, hold rate the middle, CTR the ask, and conversion rate the handoff beyond the creative.

The method is: confirm enough data, read top-down, diagnose the first break, fix only that element, relaunch and re-read. The contamination principle — an upstream break makes everything downstream look broken — is why top-down, first-break reading is the whole discipline, and why the novice who sees 'everything is broken' and rebuilds the creative is almost always fixing symptoms of a single upstream cause.

The per-element checklists turn a diagnosis into an action: a broken hook sends you to the first frame, relevance and curiosity; a broken hold to pacing, payoff and clarity; a broken ask to the CTA and offer pull; and a good-CTR-poor-conversion pattern away from the creative entirely, to the landing page, offer, audience and ultimately the product. Knowing when it is not the creative is as valuable as knowing which element is.

Do this consistently and creative testing stops being a slot machine and becomes a learning system — every test teaches you a transferable lesson about what works, which compounds into better creative. That is the entire point. If you would rather have creative diagnosis, testing and scaling built and run for you, that is where our [ROAS optimisation](/solutions/roas-optimization) work sits, alongside the [automated creative testing guide](/guides/automate-meta-creative-testing-n8n-guide) for doing the measurement at scale and the [creative strategy](/resource/blogs/meta-ads-creative-strategy) work for making better creative in the first place.