Changing faces: Can AI personas work in market research?

Artificial intelligence (AI) promises to be a game changer for the world, and market research is highly unlikely to avoid its impact. AI’s uses could be multitudinous, but in recent months, one of the most hyped areas in the insights industry has been the use of AI ‘personas’ to help provide a more personable and relatable way to interrogate consumer data.

AI personas have been rolled out by several companies, helping to bring a human ‘face’ to the consumer data that companies have, often by presenting the views, behaviours and attitudes of particular segments through an AI interface that can then be interrogated by researchers. But do these personas actually portray the views of consumers accurately? And what role could they play in the market research process?

MMR Research recently carried out an experiment to test its persona tool, bringing in the vegan food brand Gosh! as a partner. The aims were twofold: firstly, to get an unbiased view of whether personas could deliver value, and secondly, to see if they were a viable option for small to medium-sized businesses.

For the personas, MMR Research used data from a consumer segmentation study of more than 3,000 people in the UK to feed into a dialogue-based interface via a machine learning model. The persona sought to represent segments from within the study, and therefore respond as the real respondents would have done to new ideas and stimuli. The persona was made in such a way that it would have been possible to run using a brand’s internal data. “We really wanted to understand if personas are viable, useful and do anything for brands in general, and smaller brands like Gosh!,” says Alexandra Kuzmina, innovation associate director at MMR Research.

The experiment compared the responses of 500 real consumers to that of the persona when asked for thoughts on a family-orientated plant-based food product concept. Consumers provided numerous open-ended shorter answers to a series of questions, while the persona provided a few paragraphs of text responding to the concept in general. A thematic analysis was then carried out comparing the answers of the persona and the consumers.

The results found that while personas gave very detailed answers, especially on the importance of sustainability, consumers were able to give a better idea of how important different aspects of the product concept were. For example, good taste and flavour were rated highest by the consumers followed by the quality of ingredients, whereas the persona thought that the product being organic was the most important factor, despite it having been mentioned by only a single person in the consumer survey.

Prompting and probing
How can you get the personas closer to the views of real consumers? The answer was by probing the personas for more information and asking them questions. “We needed to really probe the personas – it is a real skill,” says Kuzmina. “This was not probing around these themes specifically – it was more probing about how we make this concept more appealing.”

Once the persona was questioned and prompted for further feedback, its responses were very close – albeit not identical – to that of the consumers surveyed. “This was such a positive result, and this surprised me personally, because I was sceptical at first,” explains Kuzmina. “It is also very positive as this was not trained on Gosh! primary data, and that shows how useful it could be to brands like them.”

Could another off-the-shelf model do the same? Kuzmina says she replicated the experiment with ChatGPT, including prompts. “I spent a lot of time trying to get something useful,” she adds. “A brand manager at any FMCG company wouldn’t probably have the time to play around with it, even if it was at their fingertips. Using off-the-shelf tools like this would mean they all are chasing very similar innovations.”

Ultimately, the most actionable insights were from the personas, compared with off-the-shelf models, as they gave more detailed responses closer to those of real consumers. “The ability to probe and the skills needed to probe give us that extra detail and extra edge in usefulness for the tool,” Kuzmina adds. “Personas are a really efficient way to get more from primary research – it means that clients don’t need to run primary research every time they want to ask a single question.”

A human face
Gosh! uses consumer research at two stages of product development, but personas provide the ability to give brands an additional access point into their consumer data. Jake Schneider, chief innovation officer at Gosh! Food, says: “Having access to our segments 24/7 via a dialog-based interface means unlocking an additional access point to the consumer point of view.

“This provides smaller businesses, which might be constrained in their research capabilities, with valuable insights. This approach allows us to include consumer-led sense checks in our primary research, supporting internal decisions and offering confidence that we are on the right path. Furthermore, it is cost-effective and provides immediate access to consumer perspectives, allowing us to innovate quicker.”

The experiment highlights the role people can play in working with artificial intelligence. “A persona would not replace primary research – we still need to speak to real people to build it,” argues Kuzmina. “There is a human element that is going into the data, as the personas are grounded in consumer data.

“On the user side, we also need to be more creative in the way we probe – we need to use conversational design, similar to how we design our chatbot questions, to be able to get the most out of it. It is not about personas replacing anything, but it does mean the relationship that we as researchers have with the data, and also that clients have with the data and the expectations they have around deliverables, are going to change.

“I think it is helping clients to extend the shelf life of the research that usually goes on a shelf or is not used beyond the person who commissioned it, and enabling the whole organisation to rely on consumers’ voice and have more touchpoints with the consumer. It is going to change the way we work.”