AI-driven insights: hype or promising future?

Our Insights Director, Eva Vandenberge recently attended the annual ESOMAR Congress and in this interview, she shares her insights.

This annual event by the global association for market research and insights offers researchers and consultants from all over the world the chance to connect and discover the latest trends and innovations in their industry. More than 1,000 people from 78 countries attended this year’s congress in Athens and boobook’s Insights Director, Eva Vandenberge, was one of them. With over 135 speakers, choosing which sessions to attend was a daunting task, but judging by the insights and reflections she shares with you today, she mastered it gracefully.

It was not the first time you attended the congress. What keeps you coming back?

“Nowhere else do you get such a comprehensive image of the current state of affairs in market research and a glimpse into its future. The presentations from the industry's top minds are inspiring, and the informal conversations with researchers from all over the world are fascinating. While listening to their insights, success stories and challenges I get so many new ideas to help our clients even better with their business questions. The industry has been rapidly evolving since the arrival of AI. It is important to keep track of the developments and discover whether new technologies and tools can help us gain better and more reliable customer insights.”

Before we dive deeper into the AI topic: were there any activities you particularly enjoyed?

“I appreciated that ESOMAR organised the YES Awards: a global competition that allows young research professionals to share their - often refreshing - ideas with the public. From all the pitches submitted, a jury selected ten who got to share their 60-second pitch at the congress. Via live voting, the audience selected three finalists who could give their full presentation. The winning presentations were on cultural bias, regression analysis using AI and data collection via WhatsApp in emerging countries. As a young researcher, it is quite an honour to get the chance to share your findings with such an audience.”

“There were no Belgian candidates this year, unfortunately, but I am also a board member of the Belgian Research Federation CUBE, and at the CUBExEsomar event that will take place on 24 October, several young researchers will share their take on the future of research and insights. Hopefully, this will lead to some interesting pitches by young Belgian research talent next year.”  

‍

Did you discover ideas that you would like to implement yourself?

“Absolutely. One of the YES award winners talked about cultural response bias, something we also struggle with. Certain cultures have difficulty sharing negative feedback and judge everything so positively that we barely detect differences between groups and brands in our analyses, while in reality, of course, there are. In one of our studies, ran in India, we corrected for this during the data analyses, but we are now inspired to tackle this issue by asking different questions. Questions about behaviour instead of attitudes, for example, or questions with neutral rather than numerical scales, so there is no longer a better or worse option. I had a couple of interesting discussions about the topic and will certainly put the new insights to the test.”

‍

AI was undoubtedly an important topic in Athens. Is it already reshaping the industry?

“AI was a hot topic indeed. Many AI developments and solutions were showcased. And several speakers discussed the potential, but also possible pitfalls of synthetic data, meaning data that has been artificially created as opposed to collected from humans. This could be used in analysis in the same way human data is used. Filling in missing data is nothing new; statisticians have been maximising samples via imputations for a long time. It might be as simple as filling in the occasional empty response with ‘don't know’ or mean scores based on the rest of the sample. Other times more complex models are used to predict those missing values, based on respondents' other answers, and what the rest of the sample says. Using AI, we can now do much more than fill in the occasional missing value.”

What is already possible with synthetic data today?

“As described, synthetic data might be used to complete missing answers, but you can also generate additional cases. If, for example, your study lacks young men from a specific region, you can generate more of them to boost your sample. You could even generate synthetic respondents; virtual participants that provide answers to a survey just like human participants would, and whose answers you can analyse as you normally would. It sounds futuristic, but companies are already experimenting with it, although there is still a lack of trust. Based on what we have seen and tested ourselves, we believe this distrust is justified and we would not recommend making business decisions based on synthetic data at this point.”

It doesn’t look like human respondents will be obsolete any time soon?

“The future will tell. The presented cases show that synthetic respondents can generate reliable results if closed behavioural questions are used. But virtual respondents can’t tell you how they feel, and they can’t answer open-ended questions well. We are curious to see if and how this will evolve.”

‍

What are some of your own experiences with using AI?

“At boobook, we have done several test on how AI would handle a segmentation study, for instance. When comparing the segments AI came up with the ones we determined, some overlapped but others didn’t. As long as results remain unreliable, real data, captured from real people, remains a must. But, of course, we keep track of the new developments and will continue running tests.”

‍

What would you like to put to the test in the short term?

“There is quite some enthusiasm about using synthetic cases to map hard-to-reach groups, such as B2B audiences. Synthetic boosters in under-sampled groups would increase the reliability of the study. We have our doubts because it seems unlikely that this data would indeed be reliable. How can it be, when your research pool is small to begin with? But the proof of the pudding is in the eating, so we plan to run a test and find out for ourselves if it could be useful.”

“What was also discussed at the congress, and I very much agree with, is that you should mainly use AI when a human being can’t add value. If using AI to script and translate questionnaires means resources are freed to generate better insights, we can only encourage it. But in terms of data collection and insight generation, we still need actual respondents and human researchers to bring brands closer to their consumers.”

‍

Which speaker left a lasting impression?

“It is not easy to choose just one, but I will not soon forget the closing keynote by Vivienne Ming, an incredibly intelligent woman who has travelled an unlikely path and achieved so much. She has founded several start-ups and solving seemingly unsolvable problems is her life's goal. She believes AI can be truly transformational and has developed several AI tools, but she is most passionate about maximising our human potential. To do so we need to create open cultures and safe spaces where people dare to experiment. Most of our initial ideas are wrong she claimed, so it is only through failing that we will stumble upon the great ones that can potentially change the world. It left me inspired and proud to be part of the boobook family, where we are all encouraged to share our ideas and get the chance to pursue them.”

“Especially now that there are so many new developments, companies must give their employees room to try new things. If you want to grow you need to accept that you will also fail sometimes. Sharing experiences and ideas is more important than ever, so I would leave you with a warm invitation to connect with us if you want to explore the possibilities of AI together!”

‍