What we measure against
We don't score against a benchmark we designed. Every validated study compares the simulation's prediction to what the audience actually did afterward: the campaign that ran, the product that shipped, the listing that converted.
The Sample
More than 500 studies run between 2024 and 2026, across retail, CPG, food and beverage, pet care, media and entertainment, publishing, business software and fintech, travel and mobility, and financial services. Study communities typically range from 300 to 1,500 AI personas.
What the number means
90% is how often the simulation's predicted ranking or response matched what the audience actually did. It's not a claim that every individual response is correct, it's that the pattern holds. On top of that, 86% correlation with established ground-truth datasets.
Frequently asked questions









