Synthetic Data Standard launches certification for synthetic research data

Synthetic Data Standard on Nov. 4 launched a field guide that classifies six kinds of synthetic data used in market research and consumer insights. The independent group also is building SDX-1, a third-party certification meant to judge fidelity, transparency and bias handling as synthetic research moves into wider use.

Why it matters: - Synthetic data is increasingly being used in consumer and market insights, but buyers have lacked a shared vocabulary to compare methods and evaluate quality. - SDX-1 is designed to give providers and buyers a third-party standard for fidelity, transparency and bias handling. - The certification could make it easier for insights, analytics and CX teams to decide when synthetic respondents are fit for purpose.

What happened: - Synthetic Data Standard launched Nov. 4 in New York with the release of a field guide for synthetic research data. - The guide defines six types of synthetic data used in market research and consumer insights. - The organization is developing SDX-1, a certification that will assess whether a provider's methods meet a defined bar. - Providers can register interest in the founding cohort at more information.

The details: - The guide, titled The Six Types of Synthetic Data: A Field Guide for Buyers of Market Research, sorts synthetic research solutions by two questions: who the system speaks for and what data drives its answers. - The six categories are Digital Twin, Derived Persona, Grounded Persona, Derived Panel, Grounded Panel and Sample Boost. - A Derived Persona models a typical customer from general sources such as social media. - A Digital Twin is built from one real person's consented data and can be checked against that person. - The guide includes a definition and worked example for each type. - It also explains which questions synthetic respondents answer reliably and which they do not. - The guide gives buyers six questions to ask providers, including which type is being offered and who ran the accuracy test. - The framework draws on a mapping of 34 synthetic data solutions from 30 providers, academic research on digital twins and generative agents, and guidance from Esomar and the Insights Association. - A 14-member advisory council from industry and academia shaped the guidelines and guardrails behind the standard. - Council members will not certify, endorse or evaluate individual providers and will not take part in certification decisions or provider publications. - SDX-1 guidelines are expected later this year. - Synthetic Data Standard says SDX-1 will be a third-party audit of whether a provider's methods meet a defined bar for fidelity, transparency and bias handling.

Between the lines: - The launch reflects a market moving from experimentation to routine use, while the terminology around synthetic data remains inconsistent. - By classifying methods first and promising certification second, Synthetic Data Standard is trying to make the category easier to buy, govern and compare. - Christina Habib, a strategic advisor and council member, said the language around synthetic data is changing faster than the practice and that the group's publications will evolve as the field matures.

What's next: - Synthetic Data Standard expects to publish SDX-1 guidelines later this year. - The first providers in the founding cohort will be audited after those guidelines are released. - The group says future publications will be revised as the practice matures.

The bottom line: - Synthetic Data Standard is trying to impose order on a fast-growing but confusing category, with a taxonomy now and certification to follow.