What is it?

A digital twin is a virtual model of an object, system or process.

The twins are typically connected to their real-world counterparts through a two-way flow of data, either continuously or periodically, helping to mimic counterparts accurately.

Not to be confused with: AI personas

Digital twins are designed to be real-time replicas of a system based on first-party data, while AI personas are representations of a customer segment built from generalised data.

A digital twin can simulate or forecast individual behaviour, if built with predictive models, while a persona can help researchers to understand what a person from a segment might think. Generally, personas are static, focused on segments and qualitative research, while digital twins are more dynamic.

Why does this matter?

Digital twins allow users to simulate scenarios or monitor impacts and performance in a virtual environment, with the digital twin designed to mimic the behaviour of its real-world counterpart as closely as possible.

The objective is to allow users to test actions and decisions on the digital twin before they are rolled out in the real world. With increased availability of real-time data, plus advances in AI, the technology is becoming much more competent.

What does it mean for market research?

Digital twins can support predictive modelling, concept or message testing, and decision-making, offering insights on how, for example, a marketing campaign works in the simulation before launch.

Digital twins can either model people or systems, such as markets or categories, using data from surveys, web interactions, purchase decisions and social media signals.

Debrah Harding, managing director at MRS, said: “Digital twins are a valuable tool to explore, for example, ‘what-if’ scenarios and to simulate likely behaviours where live data is limited, prohibitively expensive or ethically sensitive. The speed and scale which this technology offers, and its ability to build understanding of, for example, hard-to-reach groups of the population or vulnerable participants is a major opportunity for the sector.

“However, as with all synthetic data practices, digital twins should always be used as a complement to, rather than a replacement for, real-world, human insight. The value of its outputs relies on quality primary data feeding the model, as well as methodological rigour, transparency and governance. 

“This means being clear and open about what the technology is doing – and, importantly, what it isn’t doing – with explainable models, plainly defined assumptions and professional judgement to understand the full context of any results. It’s this careful governance and meticulous oversight which will maintain trust in the insights digital twins generate.”

Ray Poynter, chair of Esomar’s Professional Standards Committee, said: “A digital twin in market research is a data-driven model of a real person, designed to simulate how they might respond to new questions or scenarios (in the context of information we hold about them). The concept offers an intuitive appeal: if a model is grounded in real data, we can see how it could draw useful inferences, and we can validate the process by comparing the model’s responses with those of the real person.

“For any organisation with an online community, digital twins are a natural fit as a starting point with advanced synthetic data. The important caveat is that they are best suited, for now, to lower-risk use cases with smaller budgets and a need for fast turnaround.”

• The first instalment of this series looked at agentic AI.