How we test whether our Human Digital Twins reflect the people behind them.
Every Human Digital Twin starts with an in-depth interview with a real, consented participant. We structure 120+ data points, test responses against held-back human answers and publish the limits clearly.
The Right Question to Ask
The useful question is not whether a Human Digital Twin gives the same words every time. It is whether the themes, preferences and reasoning point in the same direction as the person behind it. AYA measures that overlap and shows where the method should—and should not—inform a decision.
How Human Digital Twins Are Built and Tested
Four stages connect every answer back to a real interview and make validation part of the build rather than a claim added afterwards.
Recruit & Interview
It starts with real people. We conduct structured one-to-one interviews with actual consumers — capturing motivations, values, decision triggers, emotional responses, and behavioral patterns that surveys can't reach.
Build the Individual Profile
The interview is structured into 120+ data points across demographics, personality, values, consumer behaviour, communication, technology, interests and decision-making. The detail remains tied to one individual rather than an invented segment average.
Create the Human Digital Twin
The structured profile grounds a Human Digital Twin that can respond to new briefs using the person's recorded context, language and decision trade-offs. The system is instructed to stay within that evidence rather than invent a generic character.
Validate Against the Human
We compare the Human Digital Twin with answers held back from the original interview and, on selected studies, with human responses to the same brief. Differences are evidence too: they define where the result needs caution or further human validation.
120+ Data Points Behind Each Individual.
The profile captures the context that shapes how someone interprets a question and makes a choice. These are the eight customer-facing categories—not an inventory of every internal field.
Each twin profile spans 8 data categories
Demographics
ExamplesPersonality
ExamplesValues & Beliefs
ExamplesConsumer Behaviour
ExamplesCommunication Style
ExamplesDecision-making & Choice Drivers
ExamplesTechnology Usage
ExamplesInterests & Lifestyle
Examples87%
Average Alignment
Based on AYA validation testing
What This Number Actually Means
Across our validation testing, Human Digital Twin responses show 87% average alignment with the people they represent. This measures overlap in themes, preferences and reasoning—not word-for-word identity or a prediction of commercial outcomes.
AYA provides strong directional evidence. It should be paired with appropriate human research for high-stakes, regulated or statistically representative decisions.
Where AYA Works Best, and Where It Doesn't
Directional evidence is most useful when its boundaries are visible. These are the situations where AYA can guide a decision and where human validation matters more.
Sensory & Physical Experiences
Products where texture, taste, scent, or physical handling drive purchase decisions are harder to simulate. AYA excels at evaluating messaging, positioning, and creative — not sensory response.
Specialised B2B & Niche Domains
Highly technical verticals with unique jargon, workflows, or procurement processes may require custom twin calibration. Our standard twin library is optimised for consumer-facing markets.
Cultural Recency & Rapid Shifts
Twins are calibrated against stable behavioral patterns. Rapidly shifting cultural moments, viral trends, or breaking news can create temporary gaps until the next calibration cycle catches up.
AYA is designed to inform decisions, not make them. Pair it with in-market validation when the stakes, regulation or need for statistical representation are highest.
Our Commitments
The principles that guide how we build, validate, and deploy Human Digital Twins.
Designed to Inform, Not Decide
AYA helps teams stress-test assumptions, compare directions and surface blind spots. It does not predict commercial outcomes or replace professional judgement. The final decision remains yours.
Bias Controls & Transparency
We use balanced audience selection, prompt guardrails and validation checks to reduce leading responses and expose disagreement. Material limitations and methodology changes are documented rather than hidden.
Privacy & Consent
Human Digital Twins are built from consented research participants. Customers receive bounded research profiles and generated evidence rather than source interview transcripts or direct identifiers.
Your Data Stays Yours
Other customers cannot access your workspace research. AYA processes it only to operate, secure and support the service, does not sell it to competitors, and uses contracted AI APIs whose foundation models are not trained on customer prompts or outputs by default.