Your customer intelligence,
made queryable.

02Built around your universe

Make the twin
native to your market.

Shape the population with the evidence closest to your business, then deepen it with Native research, benchmarks, and the signals surrounding your category.

Each layer adds a different kind of context, making the twin increasingly native to the customers, market, and decisions your organization works with.

1. 01Your selected intelligenceBring forward the research, CRM, surveys, transactions, segmentation, media, sales data, and other proprietary evidence most relevant to the audience or decision you want to model.
2. 02Native research & benchmarksValidated surveys, experiments, benchmark studies, and other human evidence used to strengthen grounding, establish context, and support validation.
3. 03Market contextReviews, retailers, forums, social discourse, demographic structure, and other external signals that add behavioral, sentiment, and category context.

Bring forward the evidence that matters most to the audience, market, and decision you want to understand.

03Modeling

From millions of signals to a population you can question.

Native normalizes the inputs, reconciles differences across sources, and resolves the patterns that distinguish one consumer from another.

Rather than flattening the evidence into a single average, the model preserves the variation that matters: attitudes, behaviors, budgets, needs, preferences, constraints, tradeoffs, and contradictions.

Those patterns become a population of individually modeled consumers that can respond to new questions while remaining grounded in the evidence that shaped them.

04How a twin takes shape

Assemble. Construct. Prove.

Building a twin is not one modeling step. It is a controlled process for defining the audience, constructing the population, and proving that it behaves credibly enough for the work it is meant to support.

1. 01 / AssembleDefine the decision and the population.Establish who the twin needs to represent, then bring together the Native, customer, and market evidence relevant to that audience and question.
2. 02 / ConstructBuild the population.Match records, cohorts, distributions, and behavioral patterns. Resolve material gaps with controlled modeling while preserving the variation that makes the audience distinct.
3. 03 / Prove & deployValidate before putting it to work.Benchmark the population against human evidence, calibrate and version it, deploy it in HumanOS, then refresh and revalidate as the underlying evidence changes.

06Fidelity controls

Choose how the twin
works with the evidence.

Every setting stays grounded in the same underlying evidence. Fidelity controls how the twin uses that evidence for the task at hand, from highly literal and directly traceable responses to broader synthesis across related signals.

1. High FidelityPrioritize direct support.
   Best when the response needs to stay especially close to observed evidence, known distributions, and highly traceable source material.
2. BalancedConnect related evidence.
   Combines direct support with measured synthesis across patterns in the data, giving the twin room to answer naturally while remaining grounded.
3. High CreativityExplore the edges of the evidence.
   Allows broader synthesis across related or thinner signals when the goal is exploration, ideation, or understanding possibilities that are not represented verbatim in the source material.

07Trace the answer

Follow the response
back to the evidence.

See what informed the answer and trace the supporting evidence back to its source.