Digital Twins Research
June 2026
StatSocial Digital Twins
Two studies on the same Digital Twins method, on opposite ends of the GLP-1 decision. B2B is the employer benefits committee deciding whether weight-loss GLP-1s get covered at all. B2C is the curious consumer deciding whether to start. Every respondent is a real person from PeopleGraph, our patented identity graph, carrying hundreds of observed behavioral signals that shape how they answer. The people are real. The responses are generated: calibrated simulations grounded in each person's observed behavior, with every answer traceable to its source. Read a full report, pull the raw crosstab, or question the audience directly. Real people. Real behavior. Answers you can interrogate.
Every raw answer, broken out across key segments, in a downloadable file.
Each respondent says why, not just what. Quant and qual on every question.
Including 617 traditional PBM gatekeepers, near-impossible to recruit conventionally.
Composite voices from the study, ready to question like a focus group.
The same method runs message tests, creative reads, and fast directional checks in days, not weeks.
Every twin is built from an actual person in PeopleGraph, our identity graph, with hundreds of observed behavioral signals. We start from real people, and every answer traces back to source.
The sample is matched to Census and Pew benchmarks across dozens of dimensions up front, so the data is representative by design, not by heavy after-the-fact weighting.
Across 30+ external benchmarks, Digital Twins land within an average 3.3 points of the measured result. Checked against ground truth, not generated in a vacuum.
Open the data and trace any answer back to the specific respondent and rationale behind it. Nothing is a black box.
What shapes the answers
Age, gender, region, income. Matched to Census benchmarks up front.
Job titles, companies, and seniority for business audiences.
Hundreds of observed interests, from sports to hobbies to causes.
Podcasts, YouTube channels, TV, radio, and the press they engage.
The creators and voices they follow, plus personality traits built from observed behavior.
What you’d use it for
This GLP-1 work is one use. The same Digital Twins method runs fast against almost any audience.
Validate a direction in days, before a full panel study or a media spend.
A/B test claims, concepts, and creative across audiences in days, not weeks.
Read how an audience will react while a story is still moving.
Put questions to the people in the study and follow up, like a live focus group that never closes.
The study
Two studies, run the same way, on opposite ends of one decision: who covers GLP-1s, and who starts them. Open either one below.
StatSocial Digital Twins come from real behavioral profiles in PeopleGraph, our patented identity graph. Each twin stands for an actual person, grounded in observed social activity. Personally identifying information is severed before generation. The attributes are real. The respondent is anonymous. What you read are calibrated, behavior-grounded simulations with a real audit trail: every answer traces back to a specific respondent's rationale. Each chat persona is a composite of a real cohort, not a character a model invented.
From 50 million behaviorally rich profiles, we draw a sample whose makeup matches Census and Pew benchmarks across dozens of demographic and attitudinal dimensions, before any response is generated. The respondents are real. Only the responses are generated.
The payoff shows up in the design effect: the weighted sample carries nearly as much information as the raw sample. Pre-calibration produced respondents whose distribution already tracked the population, so it takes little weighting to get there.
Opt-in panels, recruited from active pools in the hundreds of thousands to low millions, usually need heavier weighting to become representative. Fully synthetic respondents, built with no real-population grounding, cannot calibrate this way. The panel structure is the difference.
Across 30+ external benchmarks, Digital Twins land within an average 3.3 points of the measured result. The respondents are real. The accuracy is checked against ground truth. How we validate →
Digital Twins studies complement traditional research; they do not replace it. They suit fast-turnaround, directional work. For studies that require regulatory or legal evidentiary standards, pair them with a probability-based panel.