About

Built to run
decisions
before they're made.

Most decisions fail not from bad intent, but from untested assumptions. The real world is expensive to test in. We built an alternative.

Simulatte is decision infrastructure — the system organisations run complex, high-stakes decisions through before committing to them in reality. Not a research tool. Not a dashboard. Infrastructure.

The system models synthetic human populations — calibrated on behavioral, attitudinal, and contextual variables — and runs decisions through them at scale. Pricing changes. Product launches. Messaging variants. Brand repositioning. The simulation runs first. The decision follows.

Each engagement improves the calibration of the population it runs on. The system learns from every decision it processes — which means the second simulation for a client is more accurate than the first, and the third more accurate still. This is infrastructure behaviour, not tool behaviour.

We work with a small number of organisations as design partners. The relationship is close, the scope is specific, and the output is a decision architecture — not a report.

Simulations delivered
30+
Decision simulations run across 6 organisations in 5 verticals. Each engagement produced a traceable behavioral prediction before a real-world commitment.
Attributes per persona
133+
Correlated behavioral, attitudinal, and contextual variables per synthetic agent. Generated via progressive conditional filling — not random assignment.
Design partner model
Currently accepting a limited number of design partners. Engagements are scoped, structured, and delivered directly — every client works with the same level of rigour applied to the research program.
Pew USA — Study 1A
88.7%
Distribution accuracy across 15 Pew American Trends Panel questions. 2.3pp below the 91% human self-consistency ceiling.
Pew India — Study 1B
85.3%
The first system to replicate India's political and cultural landscape at population scale. 22 calibration sprints from a 46.2% baseline.
vs. Average frontier LLM
4.9×
Closer to the human accuracy ceiling than the average large language model on India cultural questions. 10 LLMs tested, 5,878 SHA-256 verified API calls.
Independently audited · Pew Research Center ground truth · Stanford Iyengar et al. human ceiling benchmark Full research program →
How we work
01
Behaviour, not opinion.
The system models what populations will do — not what they say they will do. Survey data is an input to calibration, not the output of a study.
02
Simulation, not speculation.
Every decision variant is run through the population before it is run in the market. Outcomes are probable, not guessed. Every result is traceable.
03
Infrastructure, not insight.
We don't produce reports. We produce a decision architecture the organisation can operate repeatedly — and one that improves with every engagement.
Start a conversation.
Reach us at iqbal@simulatte.io  ·  Run the experiment at simulatte.vc