User-Level Click Probability Decays as (N+1)^-0.43: Why Meta's Ad Auction Penalizes Creative Fatigue

Every time a user sees the same Meta ad for a second time, the algorithmic probability that they will click drops by 26 percent, measured against the baseline performance of their first impression. There is no grace period for familiarity and no wear-in phase where an audience warms up to a message. The platform's auction mathematics penalize repetition immediately, forcing the advertiser to pay more for the exact same placement.[6]

This immediate degradation contradicts a foundational assumption held by many digital marketers. While marketing agencies sell the concept of brand building through repeated impressions, Meta's real-time predicted-value system operates on the exact opposite principle, treating every subsequent view as less valuable than the one before it.[1]

The mechanism driving this penalty is the estimated action rate, a core variable in the equation Meta uses to rank ads. Every time a user opens Facebook or Instagram, an auction runs in roughly 200 milliseconds to decide which ad fills the slot. The winner is not simply the highest bidder, but the ad that generates the highest total value score.[4]

That total value is calculated by multiplying the advertiser's bid by the estimated action rate, and then adding a user quality score. If the algorithm predicts a user is highly likely to click, the estimated action rate acts as a multiplier, allowing a relevant ad to win the auction against a much higher cash bid from a competitor.[3]

However, this multiplier works in both directions. When the estimated action rate falls, the advertiser's bid must increase proportionally just to maintain the same total value score and win the impression. This is why campaigns with stale creative see their cost per thousand impressions climb steadily, even when market competition remains flat.[2]

The Mathematics of Creative Fatigue

The decay in click probability is not a slow, linear drift, but a steep power-law curve. Analytical modeling of exposure benchmarks reveals that user-level click probability decays according to the function (N+1)^-0.43, where N represents the number of prior times the user has seen the exact same creative.[6]

Under this mathematical model, the first impression represents peak efficiency, carrying a multiplier of 1.0. By the second impression, the multiplier drops to 0.74. The algorithm has already discounted the ad's predicted performance by 26 percent before the user has even consciously registered the repetition.[6]

The degradation compounds relentlessly as frequency climbs. By the fourth exposure, the conversion likelihood has dropped by roughly 45 percent, a figure corroborated by Logical Position's analysis of Meta campaign data. The auction system now requires the advertiser to bid nearly twice as much to achieve the same auction rank they held on day one.[2]

By the fifth exposure, Nielsen's research on digital ad attention decay shows that click-through rates typically fall by a full 50 percent from their initial peak. The derived decay curve perfectly predicts this halving, demonstrating that the algorithmic penalty is a structural feature of the platform rather than a behavioral quirk of the audience.[1]

This power-law decay explains why account managers often see their return on ad spend oscillate wildly from week to week. A brand-new ad starts with no history, allowing it to temporarily capture cheap impressions while the system explores its potential. Once the frequency pushes past 1.5, the mathematical penalty activates, and the subsidized preview ends.[4]

The Andromeda Engine and Budget Concentration

The penalty for creative fatigue is severely amplified by Meta's current delivery architecture, specifically the Andromeda retrieval engine. Andromeda is designed to reward concentrated signal, funneling campaign budget toward the single ad that shows the strongest early performance. It is a winner-take-all dynamic operating inside the ad set.[4]

When one ad accumulates a slight early advantage in its estimated action rate, the auction awards it more impressions. Those impressions generate more data, which increases the system's confidence in the ad, prompting it to allocate even more budget. Multiple advertisers reported this concentration behavior spiking sharply after a mid-2025 platform outage reset learned delivery signals.[3]

While budget concentration maximizes short-term efficiency, it drastically accelerates the decay curve. The winning ad exhausts its core audience at an unnatural speed, driving its frequency up and its click probability down. Because the system starved the backup creatives of impressions, they have no momentum to take over when the winner inevitably collapses.[3]

This creates an expensive, frustrating loop for media buyers. The winning creative runs hard, frequency climbs on the warmest audience segments, and the cost per acquisition deteriorates. Meta's system detects the falling estimated action rate and pulls back delivery, leaving the advertiser with a stalled campaign and no primed assets ready to rotate in.[3]

The standard response of launching a dozen new ads into the same ad set actually makes the problem worse. "Meta's delivery system rewards concentrated signal: fewer ads with more data each, so that estimated action rates get confident and auctions get cheaper," notes Aurelius Media's 2026 analysis of the Andromeda engine.[4]

Diagnosing the Decay Signal

Most marketing teams measure ad fatigue incorrectly by relying on account-wide average frequency, a vanity metric that masks underlying decay. They assume that an ad is safe until its overall frequency hits 3.0, treating the metric as a blunt threshold. In reality, fatigue hits specific, high-intent audience segments long before the average frequency reflects the saturation.[1]

The true diagnostic signal for the penalty is a decoupling of click intent and purchase intent. If the click-through rate drops by more than 20 percent week-over-week on a top-performing creative, but the conversion rate among those who do click remains stable, the ad has lost its hook. The audience still wants the offer, but they are ignoring the vehicle.[1]

Advertisers can confirm this diagnosis by filtering their ad-level reports to compare cohorts. If users at a frequency of four or higher show a click-through rate 30 percent lower than users seeing the ad for the first time, the algorithmic penalty is actively suppressing delivery. The creative is mathematically exhausted.[1]

Another clear indicator is a sudden, unexplained rise in cost per thousand impressions without a corresponding increase in market competition. When the estimated action rate decays, the auction compensates for the weak predicted response by demanding a higher effective bid. A 15 percent cost increase on a static click-through rate isolates the auction penalty as the culprit.[1]

This dynamic is particularly punishing during high-demand periods like the fourth quarter. In 2025, Meta's average price per ad rose 9 percent year over year, while impressions grew by 12 percent. Running fatigued creative into a market where baseline costs are already elevated guarantees that a flat budget will buy progressively less volume.[5]

Structural Fixes and Creative Rotation

Escaping the decay curve requires abandoning the idea of a stable, evergreen Meta campaign. The platform is a creative testing treadmill, and stepping off it is a decision to pay retail prices for wholesale results. The immediate fix is structural isolation: placing one creative concept per ad set with its own strict budget cap.[3]

Isolation prevents the Andromeda engine from concentrating spend and starving backup assets. By forcing the system to distribute impressions across multiple distinct concepts, advertisers can dilute the per-unit frequency and extend the overall lifecycle of their creative library. It is a mechanical defense against a mathematical penalty.[4]

Rotation must also shift from a calendar schedule to a signal-driven cadence. The cheapest time to replace an ad is just before the decay curve steepens. When the click-through rate falls 20 percent from its peak, the creative must be swapped out, regardless of whether it has been running for three days or three weeks.[1]

Video creatives typically resist this fatigue slightly longer than static images, often lasting seven to ten days before the degradation becomes terminal. User-generated content and testimonial formats provide the best longevity, as their native aesthetic blends into the organic feed and delays the user's immediate recognition of an advertisement.[5]

The New Burden of Production

Advertisers must build a deep creative library with at least four to six active variations ready to deploy at any moment. Scaling a Meta account without creative depth is impossible, because the algorithm will always concentrate spend and accelerate fatigue. The only sustainable strategy is to feed the machine fresh assets faster than it can decay them.[3]

This reality shifts the primary burden of digital marketing from media buying to creative production. The ad set targeting still matters at the margins, but most high-spending accounts now run broad audiences and rely entirely on the creative to do the targeting work. The hook, the format, and the copy dictate who stops scrolling.[4]

When a brand understands that the auction actively punishes familiarity, they stop fighting the platform and start working within its constraints. They stop tinkering with audience exclusions and start pouring effort into creative volume, knowing that every new asset resets the decay clock back to zero.[4]

The decay rate is not a flaw in Meta's architecture; it is the intended function of a system designed to prioritize user experience over advertiser convenience. As long as the auction rewards relevance, the mathematical penalty for repetition will remain absolute. The brands that win are simply the ones that rotate before the algorithm forces them to.[6]