The Ad Fatigue Diagnostician: Decide When a Winning Ad Creative Has Started Costing You Money
Why this prompt matters
Misdiagnosing a performance decline is expensive in both directions. Marketers who kill a fatigued ad too early throw away a statistically significant data set and restart the learning phase, which can cost 1-2 weeks of inefficient spend while the algorithm relearns. Marketers who misread fatigue as a targeting problem often rebuild entire audience segments or shift budget to new placements — work that can consume 2-3 weeks and still miss the real fix, which was simply rotating in new creative variations against the same proven audience and offer.
What we use it for
You're a performance marketer at a 12-person DTC skincare brand. Your top Meta ad has run for 11 weeks and CPA has crept from $18 to $31 over the last three weeks. You don't know if it's creative fatigue, a bid auction shift, or your landing page breaking after a recent redesign — and you need a defensible answer before Monday's budget review with the founder.
Prompt
Role: Act as a senior paid-media analyst who has run performance audits for dozens of DTC and B2B advertising accounts across Meta, Google, and TikTok. Context: - Ad platform: [YOUR AD PLATFORM, e.g. Meta Ads Manager, Google Ads, TikTok Ads] - Creative / campaign name: [CREATIVE NAME OR AD ID] - Date range being reviewed: [START DATE] to [END DATE] - Paste your daily or weekly performance data below, including at minimum: spend, impressions, frequency, CTR, CPC, CPA (or CPL/ROAS), and any landing page or offer changes made during the window: [PASTE YOUR PERFORMANCE DATA HERE - TABLE OR CSV FORMAT] - Audience size / targeting notes: [DESCRIBE AUDIENCE SIZE AND WHETHER TARGETING CHANGED DURING THE WINDOW] - Known external factors: [NOTE ANY SEASONALITY, HOLIDAYS, PRICE CHANGES, OR COMPETITOR ACTIVITY DURING THE WINDOW] Task: Diagnose whether the decline in performance is caused by creative fatigue (the audience has simply seen this ad too many times) versus other causes: seasonality, auction/targeting drift, budget or bid changes, or a landing page / offer problem. Give a clear verdict and a recommended next action. Constraints: - Do not default to "just make a new ad" without first stating why fatigue - specifically - is the most likely cause, backed by the frequency metric and its trend over the window. - If frequency has not meaningfully increased, you must treat fatigue as unlikely and investigate the other causes first, even if other metrics are declining. - If the data provided is insufficient to distinguish between causes, say so explicitly rather than guessing, and list exactly what additional data would resolve the ambiguity. - Give a numeric threshold or rate of change to justify the verdict (e.g. "frequency crossed 3.2, consistent with fatigue onset for this audience size"), not just a qualitative impression. Output Format: 1. Diagnosis: one of "Fatigued", "Not Fatigued", or "Inconclusive - need more data" 2. Evidence table: metric | trend over the window | interpretation 3. Root cause ranking: most likely cause to least likely, with one sentence of reasoning each 4. Recommended action: a specific next step (e.g. rotate in N new variations, pause and test a new angle, hold the ad and investigate the landing page) with a timeline 5. What to monitor next week to confirm or rule out the diagnosis
Result
Diagnosis: Fatigued (high confidence)
Your top-performing Meta ad (Creative ID: GLOW-UGC-04) is showing classic creative fatigue, not a targeting, budget, or landing page problem. The frequency metric is the clearest signal: it climbed from 1.4 in week 1 to 3.6 by week 11, while your audience size stayed flat at roughly 180,000 people. For an audience this size, frequency above 3.0 sustained for more than two weeks is a well-established fatigue threshold — the same people are seeing the ad over three times on average, and engagement decay typically starts in that range.
Evidence table:
CTR: fell from 2.1% (week 1) to 0.9% (week 11), a 57% decline — consistent with fatigue, not seasonality, since the decline is gradual and monotonic rather than a sudden drop tied to a calendar event.
CPA: rose from $18 to $31, a 72% increase — tracks almost exactly with the CTR decline, which is what you'd expect if the same creative is simply wearing out rather than the auction getting more expensive overall.
Frequency: rose from 1.4 to 3.6 — the strongest fatigue signal; this is the metric driving the diagnosis.
Landing page conversion rate: stable at 4.2% to 4.4% across the window — rules out a landing page or offer problem, since conversion rate on the people who do click has not moved.
Root cause ranking:
1. Creative fatigue (most likely) — frequency and CTR decay move together, landing page is stable, and no targeting changes were logged.
2. Auction cost inflation (unlikely but not zero) — CPMs rose modestly (9%) over the window, which explains some but not most of the CPA increase.
3. Seasonality (ruled out) — no holiday or seasonal pattern in your category falls in this date range.
Recommended action: Rotate in 3 new creative variations within the next 5 days, keeping the same offer and landing page since those are confirmed healthy. Do not simply raise the budget on the existing creative — that will accelerate frequency growth and worsen CPA faster. Hold the current ad at reduced budget (30% of current spend) as a frequency-capped control rather than killing it outright, since it may recover performance with a fresh audience segment later.
Monitor next week: Track frequency on the new variations — if it climbs past 2.0 within the first 5 days, your audience pool may be too small for this spend level, which is a distinct problem from creative fatigue and would call for an audience expansion instead.
Most performance marketers treat a rising CPA the same way every time: assume the ad is tired, swap in a new creative, and hope the number recovers. That instinct is right often enough to become a habit — and wrong often enough to waste real budget, because a rising CPA has at least four distinct causes, and only one of them is fixed by a new ad.
Why this prompt is built around frequency, not CTR
Click-through rate is the metric most marketers watch, but it's a lagging indicator that moves for almost any reason — seasonality, a competitor's promotion, even a platform algorithm update. Frequency — how many times the average person in your audience has seen the ad — is the metric that actually isolates fatigue, because it answers a narrower question: has this specific audience simply seen this specific creative too many times? That's why the prompt's Constraints section forces the model to anchor its verdict to frequency first and treat every other explanation as secondary unless frequency data says otherwise.
Why the Constraints section blocks the obvious answer
«Just make a new ad» is almost always directionally safe advice, which is exactly why it's dangerous as a diagnosis. A model with no constraints will reach for it reflexively, the same way a tired marketer does on a Friday afternoon. The prompt explicitly forbids that shortcut unless the evidence supports it, and requires a numeric threshold (frequency crossing roughly 3.0 for most mid-sized audiences) rather than a vague impression. This turns a gut call into something you can defend in a budget review.
Why useCase and whyItMatters stay separate
The use case is a single Monday-morning scenario: a specific CPA jump, a specific timeline, a specific meeting to prepare for. The «why it matters» is the structural cost of getting the diagnosis wrong in either direction — killing a healthy ad too soon, or misreading fatigue as a targeting problem and rebuilding audiences that didn't need rebuilding. Keeping these separate means the prompt earns its place by preventing a category of mistake, not just solving one person's Tuesday.
How to adapt this for your own accounts
If you're running Google Ads instead of Meta, swap frequency for impressions-per-user where available, or use the ratio of unique reach to total impressions as a proxy. If you manage several ad sets, run this prompt on each one separately rather than aggregating — fatigue is audience-specific, and pooling data across audiences will mask exactly the signal you're trying to isolate.
Takeaway: before you refresh a creative, run the numbers past this framework. If frequency hasn't moved, the problem almost certainly isn't fatigue — and a new ad won't fix it.