Synthetic People Glossary
A native reference for platform metrics, validation methods, scenario types, and behavioural science concepts used across Synthetic People.
Showing 50 of 50 terms
Factors that prevent users from starting to use a product or service. In synthetic research, these are identified by asking synthetic personas what holds them back: hesitation, cost concerns, trust issues, or friction in the onboarding experience.
A statistical test focused on the extremes, or tails, of a distribution. It checks whether synthetic and human responses behave similarly at the high and low ends of the response scale.
The percentage of survey questions where synthetic and human respondents arrive at the same dominant conclusion, even when exact numeric distributions differ.
The average statistical similarity between human and synthetic response distributions, measured at a question and distribution level across all test runs.
Real-world behavioural data, including purchases, transactions, and intent patterns, analysed and embedded into synthetic personas. Behaviour signals reflect what people actually do, not only what they say they will do.
A research scenario focused on measuring consumer awareness, recall, and associations with a brand. Synthetic personas simulate how a target audience perceives a brand without requiring brand exposure studies with real participants.
A statistical test that checks whether the overall distribution of responses across answer options is statistically similar between human and synthetic data.
A research scenario that evaluates whether a product, idea, or proposition works before launch. If multiple concepts are being compared, it still falls under Concept Validation rather than Option Preference.
A statistical measure of how consistently response patterns move together across options in both datasets. A high correlation means synthetic personas and human respondents react to options in a similar relative order.
A statistical test that evaluates how closely the overall cumulative distributions match across the full range of responses. It provides a global view of distribution equality rather than focusing on specific points.
The second tier in the alignment scoring framework, with similarity between 70 and 79 percent. The overall response pattern holds, with only minor localized differences across options.
A research scenario focused on understanding how users choose between competing factors, such as price versus features or speed versus quality.
A metric measuring how often synthetic and human respondents reach the same research conclusion, regardless of whether the exact response numbers match.
A research scenario focused on how users find or hear about a product or service. If the primary focus shifts to brand awareness, it should be reclassified under Brand Perception.
A high-level statistical comparison of the overall shape of human versus synthetic response distributions, capturing differences in spread, balance, and consistency across answer options.
A tier in the alignment scoring framework indicating similarity below 60 percent. The response pattern shows a noticeable shift that may influence how results are interpreted or applied.
A composite measure summarising the average difference between predicted synthetic responses and actual human responses across all answer options. Lower error indicates higher prediction reliability.
A research scenario that evaluates how users navigate or interact within a product, journey, or interface. If the focus is on measuring outcomes rather than navigation, it should be reclassified under User Behavior.
A research scenario focused on whether and why users would start using a product for the first time. If the goal is measuring continued or repeat use, it falls under Retention & Usage.
Synthetic People's proprietary AI model trained on terabytes of consumer research data, including surveys, focus groups, interviews, point-of-sale data, and social listening.
Synthetic People's commitment to transparency in how insights are generated. Every output includes data provenance, audit logs, and reasoning traces so users can see the source, logic, and evidence behind each synthetic response.
A behavioural science concept describing the measurable difference between what people say they will do and what they actually do in practice. Synthetic People's approach is designed to reduce this gap by simulating behaviour rather than only capturing opinion.
A statistical measure that evaluates how much the human and synthetic response distributions overlap. It provides a balanced, symmetric measure of similarity.
Kullback-Leibler Divergence measures the information lost when the synthetic response distribution is used to approximate the human one. A lower score indicates a more faithful representation of human patterns.
A curated library of high-quality, topic-relevant sources such as industry reports, research papers, and think tank publications. These sources ground synthetic persona responses in accurate contextual knowledge.
A statistical test that measures the maximum difference between the cumulative response curves of human and synthetic data, identifying the largest single point of divergence between the two distributions.
A non-parametric statistical test that compares the central tendency of human and synthetic responses, checking whether one consistently ranks higher or lower than the other across response options.
A research scenario focused on testing how well messaging, positioning, or communication resonates with a target audience. If the message is embedded in a broader product concept, it should be reclassified under Concept Validation.
Subconscious decision-making indicators that go beyond stated responses, capturing how choices are formed at a psychological level. These signals are aligned to the audience and scenario of the research study.
A research scenario where the goal is to identify which option a user prefers among a set of alternatives. If the comparison is about viability rather than pure preference, it should be reclassified under Concept Validation.
A research scenario that evaluates whether users achieve their intended goals or benefits when using a product. If the focus is on the usability process rather than the outcome, it should be reclassified under Experience Flow.
The number of distinct synthetic personas modelled to represent the audience segments defined in a human survey, matched to the same demographics, context, and decision environment as the real research study.
A metric measuring how reliably synthetic outputs anticipate the dominant human choice across all questions within a given study.
A tier in the alignment scoring framework for similarity between 60 and 69 percent. The relative order of answer options changes, while the broader directional pattern remains comparable.
A research scenario focused on determining willingness to pay or the optimal price point for a product. If price is one of several competing decision factors, it should be reclassified under Decision Tradeoffs.
A research scenario measuring the likelihood of users taking a future action, such as buying, signing up, or using a product. If the intent is tied to evaluating a specific concept, it should be reclassified under Concept Validation.
A distinctive platform feature that enables deeper follow-up conversations with synthetic personas. Users can probe beyond first-level answers, challenge responses, and explore why a persona holds a particular view.
A metric assessing how consistently patterns between answer options hold across both human and synthetic datasets, not just whether individual answers match.
A research scenario focused on understanding repeat usage, engagement, or churn among existing users. If the product is pre-launch and users have not yet adopted it, it should be reclassified under First-Time Adoption.
The number of distinct types of research problems tested in the Test Studio, including pricing, behaviour, adoption, and intent studies.
The eight dimensions used to classify survey questions in the Test Studio: Attribute, Behaviour, Knowledge, Preference, Evaluation, Emotion, Reasoning, and Intent.
The highest tier in the alignment scoring framework, with similarity above 80 percent. Synthetic and human responses follow nearly identical patterns across options.
A category of AI-driven consumer intelligence that generates insights through synthetic personas and data modelling, simulating real behaviours and preferences without recruiting human participants.
An AI-generated representation of a consumer segment, built from demographic profiles, psychographic attributes, behavioural signals, neuroscience indicators, and knowledge sources.
A statistical test that checks whether the average response differs significantly between human and synthetic outputs. A non-significant result indicates that synthetic respondents are answering at a similar average level as humans.
Synthetic People's public validation environment: a transparent benchmarking system that reruns real-world surveys on the platform and measures how closely outcomes align with human results.
A research scenario focused on understanding what users do or how they act in a given situation or context. If the focus is specifically on navigating through a product interface, it should be reclassified under Experience Flow.
A research scenario that measures how satisfied users are with an experience or product. If satisfaction is closely tied to repeat or habitual behaviour, it should be considered under Retention & Usage.
A summary indicator in the Test Studio showing how many of the 12 statistical validation checks a study has passed and what level of confidence the results carry.
Also known as Earth's Mover Distance, this statistical measure shows how much the synthetic response distribution would need to shift to match the human distribution.