The divergence heatmap is a visual diagnostic artifact. It shows where the candidate's predicted response differs from the average winner response when that artifact can be produced.
What the heatmap means
Read the heatmap as a supporting explanation, not the main score. It helps answer:
- Where did the candidate diverge from winner-like response?
- Is the divergence broad or localized?
- Does the heatmap agree with the playback timeline and element scores?
The final score, brand winner-likeness, confidence, and recommendations are more important for the go/no-go decision.
When a heatmap is unavailable
Some scorings may not return a heatmap. That can happen when the scoring backend returns network-level signals without a surface artifact, when rendering fails, or when the artifact is not relevant for the current prediction shape.
A missing heatmap does not mean the scoring failed. The result page will tell you when no heatmap artifact was returned. Use playback, brain activity curves, element scores, network scores, and the narrative as the primary diagnostic surfaces.
How to use it
Use heatmaps alongside the timeline:
- Start with final score and confidence.
- Review recommendations and weak elements.
- Open the playback-linked activity view.
- Compare the moments called out by the narrative with the heatmap.
- Decide which edit to test next.
Do not use the heatmap to make medical or clinical claims. Cortex translates predicted response patterns into product diagnostics for paid-social creative.
Good interpretation
Useful:
- "The result is low-confidence, and the heatmap is also noisy. Add more clean references before using this as a hard decision."
- "The CTA element scored low, and the heatmap suggests the candidate diverges late in the video. Test a faster, clearer CTA."
Not useful:
- "One colored area proves this ad will fail."
- "The heatmap is missing, so the scoring is invalid."
- "The heatmap is more important than final score or confidence."