How it works

How images are analysed

In short

Each image gets one attention prediction and one emotion prediction, describing the first impression of a static view. Images add the gaze plot, attention build-up, and areas of interest, and are compared with image benchmarks by creative type and sector. Image scores are not interchangeable with video scores.

An image is analysed as a single moment of viewing: what a typical viewer is likely to look at and feel in the first seconds of seeing it.

What it produces

• One attention prediction: the heatmap and fogmap, plus two image-only views:
  • the gaze plot, whose numbered stops mark the creative's main hot spots of predicted attention, joined in an indicative order;
  • the attention build-up (Convergence), an animation of how attention is likely to spread over the first seconds, from the centre toward the most attention-grabbing areas.
• One emotion prediction: the eight emotion values, Emotional Engagement and Valence.
• The full set of KPIs: E-Score, Emotional Engagement, Valence, Attention, Clarity, Focus, Spread and Mental Demand.
• Areas of interest (AOI): draw rectangles around the elements you care about (a face, the logo, a headline, the product) and give each a type and a name. Foresight does not detect elements automatically. For each area you get its share of predicted attention and an estimated viewing time; areas of type Face, Logo or Text are compared with typical values for that element type and size. See Elements and areas of interest.
• AI insights.

What feeds it

The attention model reads the image's pixels (inside the platform screen if you chose a platform). The emotion model reads your original image, and also uses a short description of the creative's content as context. Very large images are resized to a standard working size first, so every image is processed the same way.

Benchmarks for images

You can compare an image with an image benchmark group: a creative type (Advertising or Packaging) and, optionally, an industry sector. If you pick a platform, the image is framed in that platform's screen and has no comparison group: the creative type and sector choice is not available, and no comparison value is shown. If you choose no comparison group, scores are still positioned against the reference set, but no comparison value is shown. See Sectors and creative types.

Image vs video

Images and videos share the same KPIs, the same 0–100 scale and the same colour bands, but they are not interchangeable:

• Different references. Every score is positioned against creatives of the same media type (see how scores are positioned).
• Different benchmarks. Unframed images can be compared with a creative type and sector group; videos show no comparison value.
• Different kind of measurement. An image score describes one view; a video score summarises a whole timeline.

So the same number does not mean the same thing for an image and for a video. Foresight keeps them apart: Compare and reports take one media type at a time. Videos show no comparison value, so judge an image against a video only roughly, through the colour bands, not through their raw scores. See Image vs video benchmarks.

How to read it

• Image scores describe the first impression of a static view, not a timed exposure.
• Start with the heatmap and gaze plot to see whether the eye finds your key elements, then use areas of interest to put a number on it.
• Within one image, area shares compare directly: if the product gets a larger share than the logo, that difference is meaningful.

Common pitfalls

• An animated GIF is analysed as a video, not as an image.
• Don't rank faces, logos and text by their gap to the reference: each element type has its own yardstick.
• Don't put an image's E-Score in the same ranking as a video's E-Score.
• On a platform-framed image, area shares are shares of the whole screen, including the interface.