How does the Creative Fatigue Predictor work?

Olivia analyzes the uploaded ad creative for visual fatigue signals (template detection, color saturation, contrast, composition), copy signals (hook strength, CTA effectiveness, tired phrases), and creative DNA (uniqueness, scroll-stop power, emotional resonance). These combine into a fatigue score on a 0-100 scale with a heuristic lifespan estimate. Severity classification on each detected signal is computed deterministically; the LLM does not pick severity.

What does the fatigue score mean?

Score buckets: 0-39 Fresh; 40-59 Warming; 60-79 Fatiguing; 80-89 Critical; 90-100 Dead. The score is the vision-LLM's expert assessment based on signals visible in the uploaded image; it is not a measured CTR-decay prediction. The lifespan-days estimate is a heuristic anchor, not a forecast.

What inputs sharpen the analysis?

An image is required. Optional inputs that sharpen the confidence rating from low to high: ad-age (days since launch), spend, impressions, clicks, frequency. Headline and CTA text fields can be supplied if those aren't visible enough in the image; otherwise Olivia reads them from the image directly.

Why do scores vary between runs of the same creative?

The vision model has non-zero temperature, so scores include some run-to-run variance. Structural detection (template style, has-face, has-product, dominant colors, tired-phrase flags) is more stable. Treat the score as directional, not exact; the value is in the specific refresh suggestions, not the number itself.

Does the tool predict actual ad lifespan?

No. The tool surfaces fatigue signals visible in the creative itself. Actual ad lifespan depends on audience size, platform delivery, frequency caps, creative variation, and competitive context the tool cannot see. Account-data-driven fatigue tracking lives in the connected Cresva product.

What should I do after the analysis?

The result includes specific refresh suggestions with the current state Olivia observed and the suggested change, sorted by difficulty (quick-fix / moderate / extensive). Start with quick-fix suggestions, then plan major refreshes if the score is in the Fatiguing or Critical bucket.

Now in pilot: Agent Commerce Protocol, the open standard for AI commerce.Pilot live: ACP for AI commerce.Explore ACP

Predict creative fatigue before it lands.

Score fatigue risk on an ad creative and surface specific refresh suggestions from visual, copy, and DNA signals.

OliviaCreative

Ad Creative

Drop image or video here

or click to browse

Industry

Platform

Fatigue velocity: moderate

Days Running (optional)

Ad Headline (optional)

CTA Text (optional)