When a winning ad stops working, check things in the order of what they cost to check: tracking and attribution first, then delivery, frequency and hook rate, then the creative itself. Only the last step costs money, and it is the step most teams start with. Nine times in ten the honest answer is that the ad was always going to die on this schedule, and the useful question is not what broke but whether you have replacement stock ready.
Everything below is free to check except the final item. Work top to bottom, and stop at the first check that explains the numbers.
Start with the ad's age, because it is free and usually decisive
Before you open anything else, look at the launch date. Our 2026 ad fatigue benchmark, a synthesis of published platform and agency figures rather than our own experiment, puts CTR decline at 15 to 20 percent over a creative's first two weeks, then accelerating. Week three is a cliff: 45 to 70 percent below the launch baseline. By week five a creative averages 38 percent below its peak, and most creative is effectively dead within three weeks.
Decay speed varies by vertical. Days to a 40 percent CTR decline run at 9 for food and beverage, 12 to 14 for fashion, about 18 for beauty and DTC, about 21 for electronics and about 28 for B2B SaaS.
So if you sell food and your hero ad has been live for 11 days, you have most of your answer already. The rest of the checklist is confirming it rather than hunting for a villain.
The triage order, with the cost of each check
| # | Check | Where | Time | Money | Rules out |
|---|---|---|---|---|---|
| 1 | Launch date against your vertical's decay curve | ads manager | 2 min | none | ordinary fatigue |
| 2 | Change history on the ad, ad set and campaign | change log | 5 min | none | self-inflicted resets |
| 3 | Conversion events against backend orders, same window | platform vs your own database | 15 min | none | a tracking break |
| 4 | Attribution window and comparison period | reporting settings | 5 min | none | a reporting artefact |
| 5 | Frequency and whether reach has stopped growing | delivery breakdown | 10 min | none | audience exhaustion |
| 6 | CPM trend against the previous 30 days | trend chart | 10 min | none | auction and seasonality |
| 7 | Hook rate now versus the first week | 3-second views over impressions | 10 min | none | creative fatigue |
| 8 | Post-click conversion rate, page speed, stock | analytics and site | 30 min | none | an offer or site problem |
| 9 | Ship new variations | production | days | production plus media | nothing, it is the fix |
Checks one to eight cost an afternoon and no budget. Check nine costs real money, which is exactly why it belongs last and why so many accounts pay for it first.
Why tracking and attribution go before creative
A tracking break and creative fatigue look identical on the surface: reported conversions fall while spend holds. They are separated by one comparison. Pull platform conversions and your own order or signup count for the same date range and see whether they moved together. If your backend held flat while the platform's count fell, the ad is still working and your measurement is not.
The specific things worth ruling out in that fifteen minutes: a consent banner or CMP update that changed what fires, a server-side event stream that failed quietly, a duplicate deduplication key change, a developer deploy that renamed a button or removed an event, an attribution window someone switched from 7-day click to 1-day click, and a comparison period that includes a promotion week. Any of these produce a graph that looks exactly like a dead ad.
Check the change history before you touch anything else, too. Editing a live ad's creative, or making a material change to the ad set, can push it back into a learning phase and depress performance for several days on its own. We have watched a team spend a fortnight rebuilding a "broken" campaign that had been edited on a Friday afternoon.
Frequency and hook rate, the second pass
Frequency is the mechanism behind most genuine decline. Weekly frequency is (weekly spend / CPM) x 1000 / audience size, and our frequency glossary entry covers the reach deduplication caveats. The benchmark puts the onset of decline at a weekly frequency of 2.5 on Meta prospecting, with a ceiling around 3.0 on TikTok, and records a 45 percent CTR drop after a fourth exposure to the same creative from Meta internal research.
Then read hook rate against the ad's own first week, not against a benchmark. Benchmark medians are there for context: 28 percent hook rate on Meta feed and Reels, 33 percent on TikTok, 22 percent on YouTube in-stream. The comparison that matters is this ad now versus this ad then.
The split between hook rate and post-click conversion is the most diagnostic pair of numbers in the account.
| What fell | What held | Most likely cause | First action |
|---|---|---|---|
| Reported conversions across every campaign at once | clicks, CPM, hook rate | tracking or attribution | compare platform events with backend orders |
| Hook rate, with frequency rising | CPM, reach | creative fatigue | replace the creative, do not rebid |
| Hook rate, with reach flat and frequency high | everything else | audience exhaustion | broaden the audience, then refresh creative |
| CPM, up sharply | hook rate, CTR, conversion rate | auction competition or seasonality | accept it or move budget, the ad is fine |
| Post-click conversion rate | hook rate, CTR | offer, landing page, stock-out | check the site before the ad |
| Everything, within days of an edit | nothing | learning phase reset | revert or wait three to four days |
Two of those six rows are not creative problems at all, and one of them tells you to leave the ad alone. That is the value of running the checklist in this order rather than reaching for the editor.
The answer is usually that it was always going to happen
Most of these investigations end in the same place. Nothing broke. The ad reached the point on the decay curve where it stops paying, and it did so roughly on schedule. Our benchmark's week-three cliff at 45 to 70 percent below launch baseline is the evidence, and the vertical figures tell you when to expect it in your category.
That reframes the problem usefully. If decline is scheduled, the failure was not the ad dying but the absence of a replacement, and the fix is inventory rather than diagnosis. The benchmark's operational conclusion is that throughput, not talent, is the bottleneck: the number of unique creative concepts a team ships per month predicts campaign longevity better than the quality of any single ad. It puts 8 to 20 live variations as what a typical active campaign needs in rotation, and finds that brands shipping 15 to 50 variants a month see 3 to 5 times longer campaign lifespan than quarterly refreshers.
The definition and mechanics of the decline itself are in our creative fatigue glossary entry if you need to make the case to someone who thinks this is a bidding problem.
What to do in the next 48 hours
Run checks one to eight today. If they come back clean and the hook rate has fallen against its own launch week, you have confirmed fatigue and the job becomes supply. Pull whatever footage you already have, get four to six new variations live against the same audience, and let the ad you were trying to save wind down rather than resuscitating it with a budget increase. Then build the standing process so the next one is a scheduled swap instead of an incident, which is what our creative testing workflow guide sets out.
Genyad is our product and it addresses only the supply half of this. You upload footage you already own once, it transcribes and tags every clip, and each variation is a fresh script, shot selection, voiceover, caption set and export built from that library rather than a re-cut of the same timeline. One variation is one credit and the free plan gives you five variations with no card, which is enough to see whether your library can carry a replacement set.
What it will not do is warn you in advance. There are no predicted performance scores, so it cannot tell you which ad is about to fall off the cliff, and there is no direct publishing into Meta or TikTok, so somebody still uploads and names the replacements. It also has no AI avatars or synthetic presenters, no static banner formats, no product-URL import and no product-feed or CSV template rendering. If your diagnosis is that the offer is wrong rather than the creative, none of this helps and you should fix the offer.
Frequently asked questions
My ad's performance dropped overnight. Is that fatigue?
An overnight cliff is more often a tracking, attribution or edit problem than fatigue, because fatigue accumulates over days as frequency builds. Check the change history and compare platform conversions with your own backend numbers for the same window before you touch the creative. Fatigue usually shows as a hook rate sliding across one to three weeks, not a single-day collapse.
How do I tell creative fatigue from an audience problem?
Look at whether reach is still growing while frequency climbs. If frequency is rising because the audience has stopped expanding, broadening the audience buys you time even with the same creative; if reach is healthy and hook rate has fallen against the ad's own first week, the creative is the problem. The benchmark puts decline onset at a weekly frequency of 2.5 on Meta prospecting and a ceiling near 3.0 on TikTok.
Can I revive a winning ad by increasing its budget?
Almost never, and it usually makes the numbers worse, because more spend against the same audience raises frequency faster and the benchmark records a 45 percent CTR drop after a fourth exposure to the same creative. A budget increase also risks a fresh learning phase. Spend the money on replacement variations instead.
How long should a video ad last before it needs replacing?
Expect a 40 percent CTR decline in about 9 days for food and beverage, 12 to 14 days for fashion, 18 days for beauty and DTC, 21 days for electronics and 28 days for B2B SaaS, per our benchmark synthesis. Week three is where most creative hits a cliff of 45 to 70 percent below launch baseline. Plan the replacement before launch rather than after the decline shows up in reporting.