Creative fatigue is a decline that happens after a creative has already earned meaningful delivery. Creative failure is an ad that never established acceptable performance in the first place. Teams that confuse the two end up refreshing weak ads that needed a different idea, or killing promising ads before downstream results were visible.
A 2026 dataset covering 368 Meta ads and 48,450,354 impressions makes the distinction unusually concrete: 80% of creatives never reached 100,000 lifetime impressions. Most ads that get called fatigued never had the exposure required to wear out.
The dataset, and what it cannot tell you
Interconnections published the following scope:
| Field | Published value |
|---|---|
| Ads | 368 |
| Managed DTC brands | 15 |
| Impressions | 48,450,354 |
| Spend | $685,791 |
| Daily ad records | 10,533 ad-days |
| Collection window | April 8 - August 11, 2026 |
This is first-party data from one organisation's reporting warehouse, covering DTC accounts it manages. It is not a survey and not a platform-wide Meta sample, so it cannot establish a universal fatigue threshold for every advertiser. Read it as a well-documented observation, not a law. The full methodology is at Interconnections: The 2026 Meta Creative Fatigue Benchmark.
Most ads never reached a plausible fatigue stage
| Reported statistic | Result |
|---|---|
| Creatives below 100,000 lifetime impressions | 80% |
| Creatives below 250,000 lifetime impressions | 89% |
| Median active lifespan | 18 days |
| Median lifetime impressions | 10,665 |
| 25th-percentile lifespan | 6 days |
| 25th-percentile impressions | 524 |
An ad that stopped at 524 impressions may have a delivery, allocation, eligibility, or early-performance problem. Calling it fatigued assigns the wrong cause, because it never had the opportunity to wear out.
So the first diagnostic question is not "has this fatigued?" It is: did this creative receive enough exposure to establish a stable baseline of its own?
What happened to the ads that did reach scale
The publisher restricted its fatigue-curve analysis to 57 creatives across 13 brands with at least 150,000 lifetime impressions, comparing each creative against its own first 100,000 impressions.
| Impression stage | CTR vs. that ad's baseline | Other published observation |
|---|---|---|
| 100K - 250K | -5.5% | Early decline |
| 250K - 500K | -5.3% | Broadly flat vs. prior bucket |
| 500K - 1M | -8.2% | Not a smooth continuous slide |
| 500K - 1M | - | CPA ran 19.6% above baseline |
| Across the range | - | CPM stayed within 3.1 points of baseline |
ROAS sat 7.3% below baseline in the 250K-500K bucket and then appeared to recover at higher exposure. The publisher warns that this recovery is probably survivorship: only ads someone kept funding could enter the later buckets at all.
Three things follow. CTR showed an early step-down rather than endless decay. CPA moved considerably more than CPM, which means the cost of reaching people was not the problem. And later-life averages describe a selected population, not the same group followed without intervention.
Early CTR did not predict which ads earned delivery
For 86 creatives with at least 3,000 impressions in their first three days:
| Early CTR group | Median days 1-3 CTR | Median lifetime impressions | Median lifetime spend | Creatives |
|---|---|---|---|---|
| Bottom third | 1.04% | 324,522 | $2,750 | 28 |
| Middle third | 2.16% | 97,944 | $1,209 | 29 |
| Top third | 3.80% | 167,847 | $4,340 | 29 |
The top-third CTR group did not produce the highest median lifetime impressions. The bottom third did. With 28 to 29 ads per group this cannot establish a stable reverse relationship, and it should not be read as one. What it does show is that the common heuristic "kill the lowest early-CTR group" did not behave as expected in this sample.
Plausible explanations include high-CTR ads attracting cheap but low-converting attention, lower-CTR ads producing better conversion quality, allocation decisions responding to metrics other than CTR, or brand differences inside each group. The data does not identify which one dominated.
A four-way diagnosis, so a new image stops being the answer to everything
| Diagnosis | Evidence pattern | Response |
|---|---|---|
| Creative failure | Enough impressions, but weak attention and weak downstream efficiency from the start | Replace the proposition, proof, or execution |
| Creative fatigue | A previously efficient ad deteriorates after meaningful cumulative exposure, while comparable conditions stay stable | Introduce a fresh concept and protect the control |
| Auction pressure | CPM rises across several creatives or audiences simultaneously | Inspect competition, seasonality, audience, placements, bid context |
| Funnel or measurement failure | Click behaviour is stable but CVR, value, or backend reconciliation worsens | Inspect landing page, offer, checkout, tracking, attribution |
The threshold, published
Most fatigue guidance stops at "watch for it" and never gives a number you can act on. Here is the one the AdRiseLab creative-fatigue checker uses, in full, so you can reproduce it in a spreadsheet without the tool:
CTR decline component = (launch CTR − current CTR) ÷ launch CTR × 100, multiplied by 2, capped at 60 points. Days-active component = (days active − 5) × 3, capped at 20 points. Frequency component = (frequency − 2) × 5, capped at 20 points. Fatigue score = the sum of the three, capped at 100. 70 or above is High Fatigue. 40 to 69 is Medium.
Worked example: an ad whose CTR fell from 2.0% to 1.4% is a 30% decline, contributing 60 points and hitting the cap. At 11 days active it adds (11 − 5) × 3 = 18. At frequency 3.2 it adds (3.2 − 2) × 5 = 6. Total 84, which is High Fatigue.
Be clear about what this is. It is a transparent heuristic scoring one ad at one moment from numbers you type in, not a figure Meta publishes and not a measurement of your account. Its value is that the weights are visible and arguable, which is more than a threshold you cannot inspect.
Make the rule account-relative
The honest replacement rule is not a number copied from another advertiser. It is a sequence:
1. The creative has crossed a minimum evidence threshold. 2. It previously met the business efficiency target. 3. CPA or contribution efficiency has deteriorated across more than one comparable window. 4. The deterioration exceeds normal account volatility. 5. CPM, placement mix, offer, landing page, and tracking changes do not explain it better. 6. A replacement concept is ready for a controlled transition.
To compute the relative change: current-window metric ÷ reference-window metric − 1. A reference CPA of $40 against a current comparable-window CPA of $49 is +22.5%. Run the same calculation for CTR, CVR, CPM, ROAS, and contribution-adjusted CAC, then compare against other active creatives, the account total, the same weekdays, and the placement and country mix.
One day is not a durable change. Use multiple comparable windows and report the uncertainty rather than hiding it.
Why frequency alone will not carry the decision
Frequency is average impressions per person reached. Two ads at frequency 3 can end differently because audience composition differs, repetition is distributed differently across people, placements and viewing conditions differ, the offer or market moved, or one simply converts profitably despite repetition.
The Interconnections study did not have the reach data to measure creative-level frequency directly, so it can neither validate nor invalidate any single frequency threshold. What it does show is that most creatives never accumulated enough impressions for high-frequency wear-out to be the obvious first explanation.
Related Reading
For the underlying mechanism, read what creative fatigue is and how Meta delivery responds. To decide how much a replacement test should cost before you run it, see how many ad creatives a Meta account needs. And why Meta ads stop working after 10 days covers the decay pattern this dataset both supports and complicates.