Stopper Framework

Report · 2026-08-15

Our 9-layer visual analysis model and the scientific foundation behind it.

What the framework is

The Stopper Framework is the analytical model behind Stopper AI and Stopper Analyze. It breaks the question "will this visual stop someone?" into nine measurable layers, scores each one, and combines them into a single 0–100 Stopper Score.

The nine layers

  • Eye Tracking & Attention Simulation: where fixations are likely to land and in what order

  • EEG Attention Simulation: an engagement proxy derived from the attention representation; a simulation, not a measurement

  • Saliency: how early and how strongly the main element stands out

  • Composition & Hierarchy: balance, alignment with the thirds or the centre, and whether the eye is guided

  • Heatmap: how concentrated the predicted attention is and whether it lands on the message

  • Typography: headline size ratio, number of text levels, text area and headline contrast

  • Color & Contrast: text legibility against its background, figure–ground contrast and palette restraint

  • Negative Space: breathing room around the subject and clutter

  • Analyze Layer: the aggregation that turns the eight scores into the Stopper Score and the written diagnosis

How the score is built

Each layer is scored by design principles rather than by a single threshold: a plateau where a value is healthy, falling scores on either side. The layers carry different weights (attention, composition and contrast weigh most; the EEG simulation weighs least), and the weights can shift by campaign objective and channel. The report shows every layer, its weight and its share of the total, so the score is never a black box.

What is in a report

The visual with nine rendered overlays, forty measured variables with plain-language explanations, the score breakdown and written recommendations: what works, what weakens attention and what to change first.

Limits

The framework predicts. It is calibrated against human studies in Stopper Lab and keeps improving as validation data grows, but a prediction is a starting point for a decision, not a substitute for testing with people.