The Science
Grounded in real research.
Our methodology is rooted in the first academic lab of its kind, with peer-reviewed accuracy results to back it up.
Defensibility
What makes our approach different.
Diversity, not averages.
Synthetic personas regress to the mean. Doppels preserve the real variation across people, the quirks and outliers that averaged data quietly erases.
Human-in-the-loop.
Every Doppel is anchored to a real, consenting user who can verify and update their twin. No other company in this category has this.
Continuous improvement.
Accuracy benchmarks are measured live against real human responses on every study, then fed straight back into ongoing model improvements.
The lab
Columbia's Digital Twins Lab.
Led by Professor Olivier Toubia at Columbia Business School, the Digital Twins Lab is the first academic research group dedicated to the design, validation, and applications of AI-based Doppels of real consumers. Doppel Market's methodology is built on the lab's research and evolves through ongoing collaboration with Olivier and his team.
88%
as accurate as humans at predicting their own preferences, with results improving as more data is shared and as the methodology evolves.
Our methodology
How a Doppel is built.
- 01
Curated questions.
We start with a research-backed question set designed to capture the dimensions most predictive of consumer behavior.
- 02
Optional data integration.
Optionally, users securely link their own data sources, deepening the Doppel with real signal beyond the questions.
- 03
AI construction.
Large language models, retrieval, and reinforcement learning translate that input into a Doppel that thinks and responds like the user.
- 04
Continuous validation.
Real users review their Doppel's responses and can correct, refine, or update at any time. Accuracy is measured and improved on an ongoing basis.