The standard way operators think about creative fatigue: "CPA is climbing, time to rotate creative." By the time CPA has moved enough to be readable, you have usually lost two weeks of spend and a chunk of performance. CPA is a lagging indicator. The signals that actually fire first are frequency, thumbstop ratio, and the shape of the engagement curve.
This is field notes on the three leading indicators, the thresholds that trigger action, and the weekly rhythm that catches fatigue before it costs money.
Why CPA fails as a fatigue signal
CPA is downstream of everything. It incorporates audience saturation, creative fatigue, platform volatility, seasonality, competition, and a dozen other variables that shift week over week. By the time CPA has moved 20 percent, three or four of those variables have probably shifted at once and you cannot tell what is causing the move. Creative fatigue is one of the signals in the mix, but it arrives late and mixed with noise.
The leading indicators are specifically metrics about how the creative itself is performing in the feed, before the downstream conversion step. They move first. They are readable week over week. They let you rotate creative while your CPA is still stable.
The three indicators
Signal 1: Frequency
Frequency is the number of times the average person in your reach has seen the ad in the reporting window. Meta reports it per ad set by default. For a prospecting ad at a well-scaled DTC brand, a healthy weekly frequency sits between 1.5 and 2.5. Above 3.0, you are hitting the same viewers repeatedly.
High frequency alone is not fatigue. A narrow audience with a tight budget can run at frequency 3+ for weeks without fatigue if the creative is strong. What matters is the direction of change. Frequency climbing week over week while reach is flat means you are exhausting the audience. Rotate creative or expand the audience before CPA responds.
Threshold that triggers action: Frequency above 3.0 with reach flat or declining for two consecutive weeks. Write a new creative or expand the audience.
Signal 2: Thumbstop ratio
Thumbstop is the percentage of impressions that result in a 3-second video view. Think of it as the hook's effectiveness. A healthy UGC ad for a well-performing product runs 35-45 percent thumbstop. A polished brand campaign runs 25-35 percent. Ad type matters.
What matters more than the absolute number is the trend. An ad that launched at 40 percent thumbstop and is now at 28 percent thumbstop has lost a third of its hook effectiveness. The audience has seen this hook enough that it no longer stops the scroll. CPA will follow, usually within a week or two.
Threshold that triggers action: Thumbstop drops more than 25 percent from launch, or falls below 30 percent for a UGC ad / 22 percent for a brand ad. Rotate the hook.
Signal 3: Engagement curve shape
Saves, shares, and comments are qualitative signal. They tell you whether the creative is doing more than getting watched. A creative that accumulates saves and shares in its first two weeks is compounding audience reach beyond paid spend. A creative that stops accumulating saves and shares has lost its organic lift even if thumbstop holds.
The shape matters. Week one: high saves and shares. Week two: similar or slightly lower. Week three: 20-30 percent drop in saves and shares. That is fatigue, even if frequency and thumbstop are stable. The creative is still being watched but it is no longer landing.
Threshold that triggers action: Saves, shares, or comments drop more than 20 percent week over week for two consecutive weeks. Creative is fatigued even if the other metrics look fine.
The weekly review
Set aside 30 minutes every Monday for this review. It runs across all your consolidated ad sets from the learning phase consolidation work.
For each creative active in the past 14 days, record three numbers: current frequency, thumbstop vs launch thumbstop, and week-over-week change in saves/shares. If any of the three triggers fire, the creative goes into rotation planning. Rotation does not mean killing the creative immediately. It means queueing a replacement and moving the tired creative out over 5-7 days while the replacement ramps.
This is boring operational work. It is the difference between rotating creative before CPA moves and rotating creative after.
Why the 20 percent threshold on engagement
Week-over-week variance in saves and shares is noisy. A single viral comment can spike numbers artificially. A single dud post from your creator partner can pull numbers down. The 20 percent threshold is chosen to be large enough to cut through normal noise while small enough to catch fatigue early. Two consecutive weeks of 20+ percent drops is the signal. One week could be noise.
This follows the same logic as avoiding week-over-week reactions on CPA in stuck-learning accounts. Random variance gets noisy. You want thresholds that clear the noise floor.
The rotation readiness question
Catching fatigue early only helps if you have replacement creative ready. A brand that spots fatigue on Monday and starts shooting new creative on Tuesday has already lost two weeks by the time the replacement is live. The operational fix is to run your UGC testing pipeline always-on, so there is a stable of tested winners ready to promote into main ad sets.
Fatigue detection and creative pipeline are the same problem from two angles. Detection without pipeline produces frustration. Pipeline without detection produces spend at fatigued creatives. You need both.
“The operator who reads frequency, thumbstop, and engagement weekly catches fatigue before it costs them. The operator who only reads CPA is always one creative cycle behind.
”
What not to do
Do not rotate based on single-week spikes. One week of 30 percent drop in saves could be a holiday weekend or a platform anomaly. Wait for two consecutive weeks unless you are seeing the drop across multiple creatives at once (which suggests a platform issue, not a creative issue).
Do not rotate everything at once. If you kill three creatives in the same week, the ad set drops back into learning phase. Stagger rotations so no more than one creative per ad set is exiting at a time.
Do not confuse fatigue with wrong-audience. A creative that never had good thumbstop is not fatiguing. It just never worked. The rotation playbook is for creatives that worked and are now degrading. New failures are a testing problem, not a rotation problem.
How long does a typical UGC creative last before fatigue hits?
At $100K+ monthly ad spend on Meta, I see UGC creatives typically last 3-5 weeks at full budget before the three signals fire. At smaller spend levels (under $50K/month), the same creative can last 6-8 weeks because audience exposure accumulates slower.
Does frequency above 3.0 always mean fatigue?
No. For retargeting ad sets with narrow audiences, frequency naturally runs higher. The trigger is frequency above 3.0 combined with flat or declining reach. If reach is growing and frequency is 3+, the audience is still expanding and the creative is still working.
Should I use Meta's own creative fatigue prediction signal?
Meta's "Creative Fatigue" signal in Ads Manager is a crude binary. It fires late and often misses gradual fatigue. Use it as one input among the three. Do not trust it as the only signal.
Related reading
This piece is part of the paid social for DTC operators hub. For the pipeline side of creative rotation, the UGC testing budget piece covers how to keep a stable of tested winners ready. If your account is in fragmented learning-phase noise, the fatigue signals will be hard to read, which the learning phase piece addresses first. For measurement hygiene, the DTC Stack Audit covers the upstream data layer.