How it works

From real data to decisions, in five steps.

We don’t generate isolated characters. We rebuild an entire population, calibrated on real data, then let you question it like a panel you own.

  1. 01

    Real data

    National census, panel and retail data, local cultural insight from ThinkONE, and your own studies.

  2. 02

    Modelling

    Maximum entropy, probabilistic models and machine learning reproduce distributions and their correlations.

    The core
  3. 03

    Synthetic population

    Thousands of coherent, representative individuals, with latent attitudes and values inferred as well as demographics.

  4. 04

    Simulation

    Ask questions, show stimuli, run interviews and test “what-if” scenarios on any segment.

  5. 05

    Insights

    Quantified, segmented results with confidence levels, plus verbatims that explain the “why”.

The core of the system is step 2, where scientific rigour separates a true synthetic human from an improvised persona.

The real challenge

A coherent population, not a pile of profiles.

Recreating a population is a multivariate problem. Each variable follows its own distribution, and every variable depends on the others.

  • Generate individuals one by one

    Absurd combinations emerge, such as a 14-year-old retiree or an executive with no income. The profile makes no sense, and neither does the whole.

  • Generate the entire population

    Real distributions and correlations are respected together. Each individual is plausible and the population is representative.

Our engine combines maximum entropy, probabilistic models and machine learning to solve exactly this problem.

Latent variables

Demographics describe people. Latent variables explain them.

Observed variables alone can’t produce coherent, reasoned answers. We infer what can’t be observed directly, so a descriptive profile becomes a respondent who can explain their choices.

What we observe

  • Age
  • Gender
  • Region
  • Income
  • Social class
  • Education

These describe people, but they don’t explain decisions.

What we infer

  • Values
  • Preferences
  • Trust
  • Attitude to risk
  • Lifestyle

They can’t be observed directly, but they drive decisions.

Why it matters

Same observed profile, different answers. Two 35-year-old urban women with the same income can react in opposite ways to the same offer. Latent variables capture that difference.

Hear the consumer at every stage, not just at the end

Traditionally, research arrives when it’s too late to change much. With a synthetic population, you can kill weak ideas early, iterate faster and launch winning concepts with more confidence.

Before
  1. Ideation
  2. Concept
  3. Creation
  4. Optimisation
  5. Launch
With HumanONE
  1. Ideation
  2. Concept
  3. Creation
  4. Optimisation
  5. Launch

Over time

Synthetic panels that evolve with you.

If you need it: your personas can persist, remember past interactions and respond wave after wave, just like a real panel, but one you fully own.

Persistent personas

Stable, reusable profiles that can be re-interviewed indefinitely and remember previous waves.

A proprietary panel

You own it. No churn, and no need to recruit participants again.

Tracking & trends

Some key KPIs and insights can be followed wave by wave.

See your market answer, live.

A 30-minute walkthrough of a synthetic population working on your category, in your market.

Book a demo