Test your decisions in synthetic societies.
Our public benchmark compares MaxEnt with generalized raking, the usual method for weighting survey samples. In one documented configuration (40 attributes, 5,626 individuals, constraints up to third order), mean statistical reconstruction error is 73% lower. See the metric and its scope →
THE TEAM BEHIND THE TECHNOLOGY
Adoption. Opposition. Influence. Polarisation. Cascades.
Up to 10 million simulated individuals.
Simulated trajectories to compare your options, based on the scenario’s assumptions.
Behind every collective movement, talk to any individual and understand why they react the way they do.
“I am not against the principle. But without an alternative near my home, I will pay twice.”
For AI agents
Imagine All The People is not only accessible through our platform.
Through MCP, your AI agents and agentic systems can create synthetic populations, question them and consult our expert agents.
They can test a creative idea, explore a hypothesis, examine simulated reactions or use the responses to continue their own work.
Explore Imagine All The People for AI agentsAn LLM can generate a persona.A panel can measure people.Imagine All The People builds the population your decision will encounter.
| Dimension | Human panel | AI personas / LLM | Synthetic panel | Customer Twin | Imagine All The People |
|---|---|---|---|---|---|
| Starting point | Recruited real people | Profiles, prompts or personas | Synthetic audience or population | Known customers and proprietary data | A population defined by the problem |
| Statistical structure | Sampling and calibration | Not natively controlled | Depends on methodology | Proprietary behavioral data | Explicit probabilistic constraints |
| What is represented | A surveyed sample | Plausible individuals | A synthetic audience | Observed customers | A coherent target population |
| Scenarios | Often requires a new survey | Possible but variable generation | Depends on the solution | Depends on architecture | The same individuals can be exposed to multiple scenarios |
| Collective dynamics | Observed in the real world | Not native | Usually limited | Mostly individual | Interactions, influence, groups and evolution |
| Primary use | Measure reality | Explore and generate hypotheses | Rapidly question an audience | Personalize customer relationships | Simulate the impact of a decision |
INFORMED DECISIONS. COMPARED SCENARIOS.
Creative industriesFrench-language radioOBSERVED4% → 7.5%audience share at six months
ConsumerMen's skincareSIMULATED2.3×more purchase intent with the new positioning
FraudFraud detectionSIMULATED−52 %of estimated potential losses
Urban mobilityPedestrianisationSIMULATED34 % → 76 %acceptance with a resident exemption
BrandHeritage biscuitsSIMULATED+41 ptsof modern perception without losing the historic base
Internal changeA&E protocolOBSERVED78%staff support measured at twelve monthsResult reported by the commissioning organization.
RetailReturn feesOBSERVED38% → 28%return rate, six months after rollout
InfrastructureOffshore windProject acceptanceCompare simulated reactions to the scenarios studied.Keep control of your data, your models and your simulation environment.
France
Runs on our infrastructure hosted in France.
For organizations prioritizing fast deployment and operational simplicity.
Dedicated environment
An environment dedicated to your organization.
Architecture and isolation levels adapted to your security, data and governance requirements.
Your country · Your infrastructure · Offline possible
Deployable in your country, on your own infrastructure, down to an environment with no Internet connection.
For decisions and data that must remain within a controlled perimeter.
State your decision.
See the dynamics it sets in motion.
