
04 Feb ROLE OF AI: SYNTHETIC DATA FOR HUMAN INSIGHTS
In a couple of years, will there be any industries left where artificial intelligence (AI) does not play a role? The chances seem slim. In the market research sector AI is set to significantly disrupt workflows in the coming years, says Stefan Boom, SVP Corporate Sales at Dutch data specialist Dynata. Synthetic data, in particular, is expected to play a key role.
Stefan has been in the consumer research business for fourteen years, and in that time, he has seen plenty of buzzwords come and go. “From panels that were supposedly doomed to disappear to big data and voice—name it, and I’ve seen it (laughs),” he says. “The only relevant question in all these developments has always remained the same: will you still be around in five years? With AI, that question is resurfacing again today. That AI will undoubtedly play a major role in our industry is beyond question. But AI will primarily complement human data and human actors rather than replace them entirely.”
One of the most promising new technologies Stefan sees is the rise of synthetic data. “Everyone in our industry knows that data quality is often a problem. The number of bots and click farms is uncountable, and consumer engagement is steadily declining. Supplementing real human data with AI-generated synthetic data could provide a solution.”
Mirroring Real People
Synthetic data is already being used in industries such as healthcare and finance. Stefan explains: “Essentially, it means supplementing real consumer data with high-quality, relevant data generated by a computer. If person X has certain characteristics and says, does and feels certain things, he then will likely also say, do and feel other specific things. So, we try to generate additional data for a person that mirrors real data. This way, we can create a more complete and qualitative picture of the consumer. Moreover, it immediately resolves privacy concerns.”
While synthetic data may sound like a great invention, we should be cautious, Stefan warns. “The biggest pitfall is that it’s difficult to predict how accurate synthetic data will be and how to prove its accuracy. And that is crucial. Of course, you could also ask how reliable human data is. It depends on many factors, from biases and mood to engagement and the wording of prior questions – all of which can influence research outcomes.”
Ozzy Osbourne and Prince Charles
Much also depends on the proportion of synthetic data used versus real data, Stefan explains. “The higher the share of synthetic data, the greater the risk of a mismatch. You need to be careful with it. A great example I once read about was the similarity between Prince Charles and Ozzy Osbourne. They are the same age, both from the UK, and share similar socio-demographic traits. But will they have the same opinions? That seems unlikely to me (laughs).”
Additionally, all the general concerns about AI still apply to synthetic research data. “For example, AI sometimes tends to ‘hallucinate’,” Stefan points out. “If you ask an AI system the same question twenty times, you might get twenty different answers, which doesn’t help research accuracy. Moreover, AI completely lacks emotion, even though emotions often play a decisive role in human decision-making. How consumers truly feel about a brand can be incredibly nuanced – even down to scents. AI, on the other hand, is black or white, zero or one. Finally, AI is always based on past data, which doesn’t always make future predictions reliable. Relying too much on pure rationality is not always an advantage here.”
Small-Scale Adoption
Currently, synthetic data is still being used on a very small scale in market research, Stefan says. “You sometimes see it in persona applications and some data providers run pilot projects where, for example, 10% of all data is AI-generated. But I am not aware of any significant research studies where it has been used on a large scale yet. It’s still very new. Nevertheless, we already see the benefits: research becomes faster and cheaper, and it helps make hard-to-reach target groups more significant. Think of high-net-worth individuals, rural residents or highly specialized B2B profiles.”
The fact that synthetic data will be more of an evolution than a revolution does not mean it will never play a significant role, says the Dynata executive. “Oh, in five or six years, we will already be living in a different world. But it will happen gradually. I don’t think AI will have completely taken over the industry in just three years. However, this phase is inevitable – just like mobile phones, the internet, and e-commerce once were. For our industry and its employees, this does not necessarily have to be negative, quite the opposite. I don’t believe jobs will disappear, but they will be redefined. Boring tasks like quality control, creating PowerPoint presentations or designing surveys from scratch will be automated. But gathering human insights and translating data into compelling recommendations will remain the domain of people.”