Simulate choices
when people shop
with AI.

Surveys and A/B tests measure people looking at products. Visibility tools count mentions. Neither tells you what a shopper will choose when an assistant shortlists for them.

So we run both versions of your market and compare. One change, everything else fixed.

First, what the assistants say. Same request, both prices.

It wins the recommendation. A recommendation is not a purchase, so now the shopper. This one usually buys Headphones B.

Any one change.

You get estimated customer choices by assistant and shopper group, confidence in the difference, and the recommendations behind each estimate.

We run experiments with real people, with and without AI recommendations, and calibrate the choice model on them. We publish an outline of each study before collecting data.

Soon, in public. We will forecast whether AI assistants recommend the products that sell, compared with Amazon best-seller rankings.

We announce the category and the forecasts when we publish. In development. We publish what our studies find, including null results.