Benchmarks

The benchmark universe

In short

The benchmark is a large library of real-world images and videos, analysed with the same models as your creatives and labelled by media type, creative type, sector and, for videos, platform. Comparison groups are only offered when they hold enough creatives for a stable average.

What it is

The benchmark is a large library of real-world creatives: hundreds of thousands of images and thousands of videos, grouped by creative type, sector and platform. It is the reference that every Foresight score is positioned against, and the source of the comparison values shown next to your scores.

What is in it

• Images and videos. The library holds both. They are kept apart: every score is compared with creatives of the same media type (see how scores are positioned and Image vs video benchmarks).
• Advertising and more. The image library covers advertising and other visual formats, such as packaging and other designed visuals. The overall image population is therefore broader than ads alone. When you want to compare against ads specifically, pick the Advertising creative type.
• Labels. Each creative is labelled by media type, creative type, sector and, for videos, platform (Instagram, Facebook, X or TikTok), where these are known.
• Video moments. For videos, the library also keeps individual frames, so that each moment of your video's timeline can be positioned against moments of the reference videos.

How it is built

Every creative in the library has been analysed by the same emotion and attention models that analyse your own creatives. That is what makes your score and the reference comparable: both are scored with the same models and formulas.

Reference creatives are analysed on their own, full frame, without a platform interface around them.

Which groups are offered

Comparison groups are offered for images analysed without a platform frame. A group is only offered when it contains enough creatives to give a stable average. That is why some sector and creative-type combinations can appear unavailable in the setup. The one exception is Packaging, which is always offered and carries a caution note in the app.

Videos and images shown in a platform frame have no comparison group: their scores are positioned against the library, but no comparison value is shown (see Benchmarks).

How to read it

Think of the library as "what typical creatives look like" to Foresight's models. A score of 50 means your creative sits in the middle of it; higher means it beats more of it. The comparison value then zooms in on the slice you chose, such as advertising images in one sector.

Common pitfalls

• Assuming the overall image population is made only of ads. It is broader, which is why choosing a creative type matters when you want an ad-to-ad comparison.
• Over-reading small differences against a group average. Group averages differ between benchmark generations.