The result page is designed to answer two questions: should this candidate move forward, and what should the creative team change next?
Cortex analyzes each ad in two layers: Gemini extracts structured creative moments and psychology, while the fMRI-trained encoder predicts brain response. The score combines whole-video neural fit, moment-level neural fit, anchor-window fit, and calibrated creative-profile fit against the brand's winners and losers.
Final score
Final score is the headline product score from 0 to 100. It blends brand-specific winner-likeness with general attention:
Final score = 0.65 x Brand winner-likeness + 0.35 x General attention
Scores at or above 50 are labeled winner-like. Scores below 50 are labeled loser-like. Treat this threshold as a decision boundary, not a guarantee of media performance.
Brand winner-likeness
Brand winner-likeness measures how much the candidate resembles the brand's winning references instead of its losing references.
It can include:
- Video score: the whole-video pattern.
- Element score: the average of comparable scored moments in the candidate, such as hook, product shot, proof, transformation, or CTA.
The default brand blend favors element-level evidence. Calibrated brands may use a different blend selected through leave-one-out validation.
General attention
General attention is independent of the brand bank. It measures attention pull from the candidate's predicted response curve. The app summarizes the upper half of the attention curve so one weak patch does not erase a video that repeatedly regains attention.
General attention can lift the final score, but it cannot replace brand fit. A flashy ad with poor brand winner-likeness still deserves scrutiny.
Confidence
Confidence is a 0 to 100 quality signal with labels such as Low, Directional, Solid, and High.
Confidence reflects:
- Number of winner and loser references.
- Balance between the two sides.
- How many candidate moments could be scored.
- How many event-type signatures were available.
Low-confidence results should be used for creative direction, not high-stakes budget decisions.
Recommendations
Structured recommendations summarize what is working, what is weaker, and what to do next. They are grounded in the score payload, Gemini creative analysis, reference patterns, and per-element scores.
Use recommendations as creative instructions. For example, Cortex may identify that the hook is strong but the CTA behaves more like losers, or that desire/motivation is the clearest gap.
Element scores
Element scores show individual moments in the candidate. A high element score means that moment landed like equivalent moments in winners. A low score means that moment drifted toward losers.
Skipped elements are not scored. They may be too short or may not have enough matching winner/loser examples in the signature. Do not overinterpret skipped elements from the brain math alone.
Partial results
If the result is marked partial, numeric scores are complete. The missing piece is usually a secondary artifact such as recommendations. Use the final score, confidence, timeline, and diagnostics normally, but do not assume the missing artifact says anything positive or negative about the candidate.