What is it?

Synthetic modelling is a broad and evolving term that covers the use of artificially generated versions of consumers, panels or datasets, for example, to simulate aspects of responses and behaviours from real people, using AI, statistical modelling or machine learning. The technology uses models trained or informed by real-world data to generate data or responses.

Synthetic modelling and synthetic panels are related; synthetic modelling is generally the overarching methodology, covering approaches including generating artificial datasets, simulated responses and broader market or scenario changes. Synthetic panels are often defined as one application of the methodology, generating results simulating a large sample of virtual respondents.

One use of synthetic modelling is simulating aspects of a market, process or scenario. For example, if a supermarket wanted to understand how a packaging redesign could affect its sales, synthetic modelling could allow it to use existing data to build a model replicating aspects of the market in which the product would be launched, and therefore test ideas before they were tried out in the real world. This would give an indication of how the changes could affect areas such as customer behaviour, sales and other wider factors. In contrast, a synthetic panel would be presented with versions of the packaging design, with synthetic respondents responding to the stimulus provided.

Why does this matter?

Synthetic modelling can help increase capacity for testing campaigns and products, as well as sometimes dramatically reducing the time data collection and analysis can take. The technology could arguably broaden access to research, allowing a wider range of people across organisations to begin projects and gather insight.

What does it mean for research?

There is an argument that synthetic modelling could supplement, or in some cases replace parts of, traditional quantitative surveys, with the claim of sidestepping some data quality issues with fraud, bots and low-quality responses, as well as generally falling levels of response rates. Its biggest impact, however, could be felt in democratising insight across organisations. With more access to easy-to-generate insight, there is the potential for the technology to both open doors to market research across organisations.

There are some concerns, however. There are risks around reduced ability to capture new, changing or underrepresented behaviour, as well as risks in receiving generic outputs. There would be a need to benchmark synthetic modelling outputs against human research. Other areas of concern include transparency about where synthetic data has been used, replication or amplification of historical bias, and dependence on the quality and provenance of source data.

Debrah Harding, managing director at the Market Research Society (MRS), said: “Synthetic approaches can help practitioners work faster, test hypotheses and make better use of existing information. However, the ability to generate more data, more quickly, does not automatically produce more insight.

“It’s important to remember that synthetic models are tools, not substitutes for independent human evidence, and their value depends on the quality of the data they’re trained on and the expertise of practitioners using them.”

Harding said that practitioners considering synthetic approaches should assess the intended application and the potential consequences of getting the answer wrong. In some situations, synthetic modelling can be extremely useful, such as helping businesses simulate how consumers might respond to different product propositions before deciding which to test through primary research. However, in more critical or sensitive scenarios, direct engagement with people remains essential, with Harding adding that “practitioners therefore have a crucial role in determining what approach to take and then deciding what evidence can be trusted and how it should be used”.

She said: “As AI and related technologies become more widespread, the profession’s value increasingly lies in combining technological capability with fundamental research skills of critical thinking, methodological expertise and ethical judgement. Practitioners must be transparent about where synthetic methods have been used, explain their limitations, and demonstrate how outputs have been tested and validated. Trust in our sector depends on that openness.

“The MRS Code of Conduct and MRS guidance on AI and related technologies provide the framework for navigating these developments. Together, they establish that appropriateness, transparency, accountability and professional expertise must remain at the heart of research and insight if we are to maintain confidence in the evidence that informs important business and policy decisions every day.”