There is a hard limit on how much evidence a commercial team can gather before a decision. Real shoppers can only be surveyed so often, panels can only be so large, and concept tests can only cover a handful of the variants a team is actually considering. The result is that most range, listing, and pricing scenarios are never tested against shoppers at all. They are argued about, and then one of them ships.
Synthetic shoppers close that gap. They are modelled shoppers, built from real behavioural intelligence rather than assumption: the observed behaviour that Watch Humans captures directly from shoppers, structured by 28 behavioural science principles that describe how purchase decisions are actually made. A synthetic shopper is not a guess about a person. It is a simulation grounded in how real people in that category have been seen to decide.
What they make possible is scale. Where a concept test evaluates three variants with a few hundred people over several weeks, the Synthetic layer simulates how a category’s shoppers respond to every version of the decision under consideration: each candidate range, each price move, each new listing, each piece of creative. Scenarios that would never justify a research budget get tested anyway, which is precisely where the surprises tend to be found.
The system also compounds. Every simulation’s results feed back into the Product Graph, so what the model learns from one category’s decisions sharpens its predictions in the next. The evidence base grows with use rather than resetting with each project.
None of this replaces real shoppers, and it is not meant to. Synthetic shoppers are downstream of real ones, and they are only as credible as the behavioural data underneath them, which is why the grounding matters more than the simulation. What they replace is the untested decision: the scenario that shipped because there was no practical way to put it in front of anyone first.
