Simulayer

Simulate human behaviour and movement.

Companies spend enormous amounts of time trying to understand how people will react. Simulayer lets you try your change on a small world of people first.

Ask a question

Illustration, not a result.

Understanding your customers takes time.

You run surveys. Conduct interviews. Build focus groups. Commission market research.

These methods are valuable, but they are slow, expensive and difficult to repeat across hundreds of possible scenarios.

Simulayer adds simulation.

We build simulated populations from published survey data and official statistics.

The usual way

  1. Survey
  2. Interview
  3. Focus group
  4. Market research

Weeks → months

With Simulayer

Run it again for every scenario.

One change.
Thousands of reactions.

A new price, a new product, a new policy, a disruption. Try each one before it happens.

The questions behind the decision

  • Who changes their mind?
  • Who buys?
  • Who leaves?
  • Who adapts?
  • Where do people go next?

The behaviour model

What will people do? Ask how a population responds to a price, a rule or a new service, and see who changes and who stays.

The towns on this page are illustrations, not results.

If a clinic opens
If prices rise

We built our own model. We locked questions away before training, then tested it on them. These results hold for the kinds of question it was trained on.

82%Right direction, more often than any model testedWhen a group differs from its country, it says which way.
10.7%More accurate than assuming a group answers like its countryA standard statistical method manages 4.0%.
27BParameters, on one server we controlNot a frontier API.

Right direction, against others

  1. Our model82%
  2. Claude Opus 575%
  3. GPT-674%
  4. Statistical method74%
  5. Claude Opus 5.571%
  6. GPT-5.666%
  7. Kimi K364%

Same test and same prompt for every model, September 2026, each at its best setting. On accuracy it is level with GPT-6 and Claude Opus 5.5. On a newer test in October 2026, Opus 5.5 was ahead.

How it works

  1. Learn how groups differIt studies how groups of people respond.
  2. Build a populationSimulated people shaped to match the real one.
  3. Change one thingA price, a rule or a new service.
  4. Read who changesYou give the national figure. It splits it by group.

How to read it

  • It predicts how groups differ, not the national figure.
  • It learns your customers from your own data.
See the full test

The movement model

Where will people go? We simulate a whole population going about its day, from home to work, to the shops and back.

Say a bus stop closes. Do people walk to the next stop, or take the car instead? Which streets fill up, and who is left with the longest trip? You can see it before anything changes in the real world.

Open the movement lab

Useful wherever people move

Try your change on a small world first.

Ask a question