A different approach to consumer simulation — one grounded in what people actually buy, not what they say they would.
Everyone wants to understand actual consumer behavior — will they switch brands? What flavors do they truly prefer? What drives their real buying decisions?
Most players in this space build profiles based on "personas." They create digital and AI versions of consumers using surveys or census data, which speaks to the general demographics of a population. However, what those people will actually do when it comes to a buying decision is merely inferred.
The say-do gap is real: what people say they'll buy and what they actually buy diverge sharply at the point of purchase. Humans are notoriously hard to predict. Survey respondents project intentions, not behavior. Census data describes populations, not individuals. And the gap between stated preference and revealed preference is where billions of dollars in product decisions go wrong.
Twin Persona takes a fundamentally different approach. Instead of building twins from what people say, we build them from what people do — specifically, what they spend their hard-earned dollars on.
Even a slice of real purchase data is enough to anchor a persona in actual behavior. This is revealed preference, not stated preference. Every transaction reflects a real decision made with real money — the brand chosen, the one passed over, the price point accepted, the substitution made when the first choice was out of stock.
The result is a behavioral digital twin grounded in how a consumer actually shops — not how they say they would.
Every dollar spent is a preference expressed. Every brand chosen over another is a decision made — not hypothesized.
The Twin Persona principleFour steps from real purchase data to actionable consumer insight.
Not all digital twins are built the same way. The data source determines the ceiling of what a simulation can tell you.
| Approach | Data Source | Strength | Limitation |
|---|---|---|---|
| Interview-grounded | 2-hour structured interviews | Deep individual psychology | Expensive, hard to scale, still stated preference |
| Pure synthetic | Census, public data, social media | Massive scale, instant | No individual grounding, purely inferred |
| Survey / workflow | Existing surveys, CRM, NPS data | Integrates what you already have | Ceiling limited by source data quality |
| Purchase-grounded | Real purchase transactions | Revealed preference, actual behavior | Requires consent-based data infrastructure |
We respond within 48 hours. No sales sequences. A direct conversation.