Juan Isaza is Chief Strategy Officer at DDB Latina & Global Head of Social Media and Brand Strategy at Catorce/DDB.
In consumer research, the term “synthetic panels” is gaining traction. While research companies view them skeptically, many marketers are attracted by their speed and cost-effectiveness. How much can robot-based research truly reveal about real human insights? As a strategist who lives to understand the human emotions of consumers when creating brand campaigns, here are my thoughts on the issue.
Understanding AI And Synthetic Panels
Synthetic panels are virtual people created with generative artificial intelligence, fed with information about purchase preferences, consumer profiles or reports on category dynamics. Just like ChatGPT can write a letter for us from a simple prompt, these panels can answer surveys as if they were real people, give us their comments on a proposed slogan for a brand, or provide verbatims on the reasons for purchasing a specific category.
The first question concerns their effectiveness. They may provide ease, speed and low cost, but how much can they truly identify consumer insights? To answer this, we must remember that one of the essential elements of insight is its originality and its ability to give us an unconventional view of things. Andy Davidson from Flamingo Consultancy provided a valuable definition of an insight years ago: “An insight is a disturbance in discourse.” The problem, then, is that generative AI knows how to write discourses perfectly without generating disturbances.
The “magic” of generative AI lies in its ability to predict the next word to make us feel like we are conversing with a human. In that sense, seeking insights with generative AI tools seems contradictory. However, we must also be realistic. Much traditional research often ends without truly providing great insights and without generating any disturbance in the discourse. Recently, a friend who works at a major research company told me that many focus groups, particularly those conducted by low-quality agencies, are filled with professional participants who attend a toothpaste focus group in the morning and an instant soup group in the afternoon. These participants, like the algorithm of generative AI, know how to respond in a way that best fits the expected answer. The difference, as always, lies in the researcher who can challenge the answer beyond the obvious.
The Evolution Of Insight Generation
The idea here is not to judge the quality of research because we would surely conclude that well-done research has always been useful and has led us to valuable insights.
• The discovery of how hunger drives people to behave in ways that make them lose connection with their group is a very original view that came from qualitative research and was the origin of a famous campaign for Snickers: “You’re not you when you’re hungry.”
• “Dirt is good” is an unconventional and unpredictable view from Unilever’s detergents of what stains express about children’s physical activity, which resulted from listening to many mothers until one dared to question the demonization of dirt in children’s clothing.
• A typical traveler would never say staying with strangers is the best way to see the world, but Airbnb (subscription required) found the power in their non-obvious vision.
Insight generation has, indeed, evolved greatly. I still believe that the best insights come from qualitative research that understands, empathizes, questions and cocreates among humans. But it is also true that today, brands can get inspiration from many sources: consumption data; transactional data; media consumption platforms; first-, second- and third-party data; social media listening; and/or customer service surveys. In that sense, generative AI could indeed integrate all the data better than any human. It can help us generate hypotheses, identify blind spots and enter consumer conversations with more tools to test and better penetrate their motivations.
Asking The Right Questions
While new companies offering synthetic consumer panels have begun to emerge, more traditional research companies are more cautious. Ipsos, in a global paper, insisted on handling synthetic data “with great care.” They currently warn against using synthetic data “as a substitute for real-world testing without proper validation.” They believe that AI tools have to be a complement and support but not a replacement. Kantar published an experiment comparing synthetic data with a real human panel. Their conclusion is that “right now, synthetic sample currently has biases, lacks variation and nuance in both qualitative and quantitative analysis. On its own, as it stands, it’s just not good enough to use as a supplement for human sample.”
However, this does not mean synthetic panels will not knock on our doors with their seductive message of low cost and speed. There is still much to understand about the ability of synthetic consumers to replace human consumers and questions that we collectively need to answer. For me, the three that are top of mind are:
1. Can a synthetic interviewee say the absurd is a very valuable element in generating disturbances in discourse? Understanding the significant difference between an AI hallucination and the provocation of a rebellious human leads us to question the essence of a category or the relationship with a brand.
2. What is the role of synthetic panel interviewers so they are not seduced by form and fail to see the essence? We already know the power of AI to sound convincing, just like that constant group session participant who always knows what to say to avoid generating any noise and avoid the researcher’s questioning.
3. How can we harness AI’s power to help the researcher and strategist find, with the help of big data, those contradictions between what people say and what they do thanks to behavioral data or the declared reasons for purchase and those revealed by their purchase process through their actions? As we all know, in those contradictions or “consumer lies,” there is always a fresh source for insights.
There is much to explore, and we should keep an eye on this topic. The evolution of synthetic panels will show us how much value they will add in identifying insights that influence human behavior and the potential of machines as true representatives of flesh-and-blood consumers. We’ll see.
Forbes Business Council is the foremost growth and networking organization for business owners and leaders. Do I qualify?
Read the full article here
