Kellogg marketing professor confines AI synth personas to the prep work before customer interviews

The savings synth personas promise come out of fieldwork. Gathering data in the field, running focus groups and executing test campaigns take time and money, according to a Kellogg Insight article by Riley Mann ^[16][1]. The companies building the tools pitch them as a faster, cheaper alternative ^[4]. A team that buys them has to decide which of those costs it keeps. Gina Fong, a consumer anthropologist and clinical assistant professor of marketing at the Kellogg School, draws the line at the point where a researcher meets a customer ^[5].

"They're a handy research tool when you're preparing and exploring," Fong said, "but they are much less effective at capturing the nuance, emotion, empathy, and unpredictability that real people offer." ^[6] The article frames her objection as one of scope. The tools work, but they are being used for jobs they were never designed to do ^[15].

Personas are not new. Marketers have grouped customers with similar behaviours and needs into profiles for decades, and synth personas build those profiles with AI from large amounts of customer data ^[14][2]. What is new is that the composite now answers back. A researcher can ask a model to role-play new moms or gamer teens without recruiting any ^[3].

Some vendors claim their personas mimic human responses with around 85 percent accuracy, the article reports ^[4]. Taken at face value, that leaves about 15 percent of responses that do not match ^[1]. The article does not say how the figure was measured or who checked it. A finance lead could fairly argue that 85 percent is good enough for an early read, and Fong's advice allows for that stage. I think the share matters less than where the misses fall. Fong's account puts the loss in the individual quirks and contradictions that real customers show ^[12]. "When you're using synth personas, you don't get any of the messiness that being human entails," she said. "Everything is rounded; nothing is sharp; nothing is contrasted." ^[12]

Her DIY example shows the order of work. During the Covid home-improvement boom she worked with a company that wanted to enter the DIY market. She says personas would have helped her learn how enthusiasts behave before she spent time in their homes ^[9]. "Synth personas are an especially useful way to get up to speed in an unfamiliar category and get ideas flowing," she said ^[10]. The home visit still happens. "I would really want to go to someone's house," Fong said. "I'd want to see their workshop, their tool shed, their tool belt, what kind of projects they've done in their house, and which ones they're most proud of. That's going to be different for every person who does DIY home improvement." ^[11]

The choice this quarter is whether personas become a prep stage or a substitute. Bought for prep, they help researchers find blind spots in their questions before the interviews start ^[8]. Fong's support stops there: "it would be a mistake to expect those personas to replace the people you hope to understand," she said ^[7]. Bought as a substitute, they leave next year's product and campaign decisions resting on a composite built from aggregated customer data ^[2]. The article recommends using both, with personas coming before the in-person research ^[13].