Winning creative on Meta typically fatigues in the 1.5-3 million impression range because that is where audience penetration crosses the saturation threshold — the ad has reached roughly 25-35 percent of its addressable pool at frequency 3+, which is the point where the auction starts paying rising CPM to show it to the same people a fourth or fifth time. The specific impression number varies by audience size and account structure, but the mechanism is consistent. The signal that precedes the collapse appears four to seven days earlier in rising CPM and declining CTR before the ROAS reflection catches up. Catching the signal early is the difference between a controlled retirement and a reactive scramble.
Winners fatigue at 1.5-3M impressions because audience penetration crosses saturation. Rising CPM precedes ROAS collapse by 4-7 days. Monitor CPM slope on top-spending ads weekly to catch fatigue before it hits performance. Two weeks of rising CPM is the trigger.
Why does the 1.5-3 million impression range specifically mark fatigue?
The 1.5-3 million impression range marks fatigue because that is roughly where a well-performing ad has penetrated 25-35 percent of its addressable audience at frequency 3 or higher. At penetration that deep, every subsequent impression is paying rising CPM to reach people who have already seen the ad multiple times and either converted or chosen not to. The auction system can still deliver, but the marginal cost per converting impression rises because the pool of fresh viewers is depleted. The specific impression threshold scales with audience size — a broad audience of 50 million people absorbs 3-5 million impressions before fatigue, a scoped lookalike of 2 million people fatigues at 500k-1M impressions. The 1.5-3M figure is a reasonable rule of thumb for typical scaled DTC prospecting ad sets.
Why does CPM start rising before ROAS drops?
CPM starts rising before ROAS drops because CPM is a leading indicator of auction pressure. When the ad set's addressable audience depletes, the auction system has to bid up to reach the remaining fresh viewers in the pool, which raises the delivered CPM. CTR often holds for a few days after CPM starts rising because the frequency-3 viewers who do click are still converting at their usual rate — they just cost more to reach. ROAS drops when CPM has been rising long enough that the per-conversion cost crosses the account's break-even threshold. The lag is typically 4-7 days: CPM starts rising on Monday, CTR starts declining by Thursday, CPA breaches target the following week, ROAS reflects it on the second weekly review. Buyers watching ROAS exclusively miss the Monday-Thursday window where intervention is still cheap. The same leading-vs-lagging distinction applies across Meta performance signals and is the core reason weekly reviews need to look at the leading indicators, not just the summary metrics.
How do you spot the fatigue signal 4-7 days before ROAS reacts?
1) Pull daily CPM for every top-spending ad over the last 14 days. 2) Calculate the 7-day moving average of CPM. 3) Flag any ad where the moving average has risen more than 15 percent over the prior 7-day window. 4) Cross-check with CTR trend — declining CTR plus rising CPM is the confirmed signal. 5) Pull frequency and impressions-to-date for flagged ads — ads above 2 million impressions with frequency over 3.5 are the highest-risk candidates. 6) For any ad meeting the flag criteria, start deploying the replacement within 72 hours. The weekly review Monday should scan this signal across all top-spending ads in roughly 15 minutes.
What should the replacement creative actually look like?
The replacement creative should keep the hook structure that was working and shift the specific execution — same angle, same promise, different visual and verbal expression. A testimonial that was winning on hook 'skin cleared in six weeks' should be replaced with another testimonial using a different speaker and different specific details, not with a product-focused ad using the same hook. The reason is that the hook structure was what the auction system identified as a winner, and shifting both the hook and the execution means the replacement is a bet rather than a continuation. In Advize's experience, replacements that preserve the hook structure and shift execution maintain roughly 70-85 percent of the original's performance from day one, while replacements that shift both start fresh and often underperform for 2-3 weeks before finding their rhythm.
What is the common mistake when retiring a worn-out winner?
The common mistake is pausing the fatiguing winner immediately when the signal shows up, which eliminates its remaining revenue production before the replacement is actually working. The right retirement is overlapping: the replacement launches at 30 percent of the fatiguing winner's budget on day one, scales to match on day seven, and the original pauses on day ten when the replacement is demonstrating comparable CPA. Abrupt retirement loses the final 2-3 weeks of profitable (if declining) performance the original would have produced. Overlapping retirement captures that revenue while bridging to the next winner, and typically produces 15-25 percent more cumulative revenue across the handoff than either hard retirement or late replacement.
What does the full creative lifecycle look like end to end?
1) Launch: new creative enters the ad set, learning phase runs for 7-14 days as the system optimises delivery. 2) Ramp: creative crosses learning phase, CPA stabilises, spend scales toward 30-50% of ad set budget if performing. 3) Peak: creative is operating at best CPA, typically between impressions 300k-1.5M, holding stable CPM. 4) Early fatigue signal: CPM starts rising on moving average, impressions in the 1.5-2M range, frequency crossing 3.5. 5) Overlap: replacement launches at 30% budget, original continues at reduced share. 6) Handoff: replacement scales to parity, original drops to 10% as tail revenue. 7) Retirement: original pauses, replacement operates at full share. Total lifecycle from launch to retirement is typically 10-14 weeks for a successful ad on a mid-sized account.
Conclusion
Creative fatigue at 1.5-3 million impressions is predictable rather than mysterious, and catching the preceding signal in rising CPM and declining CTR buys four to seven days of lead time to deploy the replacement. The practical next step for any account where creative is retiring reactively is to add CPM slope monitoring to the weekly review and treat two consecutive weeks of rising CPM on a top-spending ad as the trigger to accelerate the next variant into deployment. Most brands retire winning creatives too early when they panic at the first drop, but the right retirement window opens a week or two after that first signal, not immediately.