Predicted Brand Lift estimates how much an ad is likely to move brand metrics (awareness, consideration, action intent) before a study is run. You upload a creative and receive, for each metric, an expected lift alongside the measured ads that informed it.
It is designed for the decisions that happen before media is spent: which creative to run, which to cut, and which to test first.
What you get for each KPI
| Output | What it tells you |
|---|---|
| Expected lift | Our best estimate of the point lift the creative will achieve |
| Similar ads | The measured ads behind the estimate, with their actual results |
Predicted Brand Lift is a k-nearest-neighbours (kNN) machine learning model, a well-established method that makes predictions by learning from the closest real examples.
Find the measured ads most similar to yours, and let their real results inform the estimate.
Predictions are never invented by a black box. Every estimate is anchored to what comparable ads actually achieved in real, measured brand lift studies. This is what makes the output explainable: we can show you which ads shaped a prediction and what each of them delivered.
When a creative is uploaded, it is analysed automatically to understand:
Similarity is judged in two independent ways: by what the ad does (its creative attributes) and by how it looks (its visual profile). An ad only counts as comparable if it is a close match on both. This keeps out ads that resemble yours on the surface but are different in substance.