Food companies are shifting away from one-size-fits-all flavor profiles and using AI to match taste preferences to specific demographic and geographic groups. Rather than guessing what consumers want, they are building data-driven models that link flavor preferences to age, gender, culture, and location.
The precision here is advancing faster than before. According to Bernard Lahousse, co-founder of AI company Foodpairing, which has partnered with PepsiCo and Nestlé, the change is "from a marketing story to a measurable engineering problem." The company collects taste data, consumer feedback, and sales performance, then matches them up to discover which flavor profiles resonate with which groups.
Age and taste narrowing
Age is one of the clearest signals. Younger consumers aged roughly between 21 and 39 have narrower taste preferences than older people, whose preferences are more diverse. This means companies need a wider portfolio of flavors to reach the same percentage of older consumers compared to younger ones.
Gender matters less than exposure. Cargill found that men prefer dark, roasted, and bitter chocolate profiles, but the difference is only about 2 percent higher than women. On its own, this gap is too small to drive flavor decisions. What actually drives preference is familiarity with certain flavor profiles, not inherent demographic traits.
Geography and cultural memory
Geography shapes taste in ways companies cannot ignore. Some preferences are innate, like a newborn's attraction to sweetness. But most are learned. UK and French consumers reject American chocolate because it contains butyric acid, which they are unfamiliar with. American consumers, by contrast, grew up eating it.
The trick is balancing global trends with local preferences. In Sweden and the US, local and global preferences align. In the UK, there is stronger divergence. Within the US itself, national preference patterns do not show up in certain regions. Companies must look at both demographics and geography together, not separately.
The realistic scale of personalization
The future is not one flavor per person. Foodpairing's Lahousse says the realistic near-term endpoint is "intelligent segment-level precision, moving from 4 to 6 product variants to 15 to 20, each backed by a validated consumer model rather than category intuition." Full individual personalization would destroy economies of scale.
Anne Berends, R&D director for Cargill's consumer product experience team in EMEA, stresses the goal: "We don't want to guess. We want to be sure from the start." AI allows companies to tailor formulations to specific situations or moods. An ice-cream bought during the day might have a different formulation from one eaten at home in the evening.
The balance is clear. Personalisation must be weighed against operational efficiency. Companies can now segment more precisely, but practical manufacturing means stopping short of full individual customization.