Millions are spent each year concocting the latest tastes and flavors to tantalize consumers, with companies experimenting with various AI-based technologies to help come up with interesting combinations. New research from the Monell Center suggests a big data analysis of reviews left by Amazon consumers may be a handy short-cut towards a better understanding of the changing tastes of the market.
In total they crawled through around 400,000 food reviews left by consumers, with the general consensus being that many of the foods on the market today are excessively sweet.
“This is the first study of this scale to study food choice beyond the artificial constraints of the laboratory,” the researchers say. “Sweet was the most frequently mentioned taste quality and the reviewers definitively told us that human food is over-sweetened.”
Not so sweet tooth
In total, the team analyzed 393,568 unique reviews of 67,553 food products over a 10-year period. The analysis used statistical modeling to identify specific words related to everything from the taste to the odor of food, as well as aspects such as the cost and the quality of customer service surrounding the product.
“Reading and synthesizing almost 400,000 reviews would essentially be impossible for a human team, but recent developments in machine learning gave us the ability to understand both which words are present and also their underlying semantic meaning,” the authors continue.
The data revealed that around 1% of all product reviews featured the words “too sweet”, with this materializing regardless of the food type being reviewed. Indeed, over-sweetness was referenced 25 times more often than under-sweetness. Sweet taste was found in around 11% of all reviews, which is three times as many as mentions of a bitter taste, suggesting that sweetness is the key characteristic people look for.
Marmite foods
The team also examined foods that drew contrasting responses from consumers to try and understand just what it was that made people either love or hate that particular foodstuff. It seemed that two main factors were responsible. The first of these was product reformulation, and the second a differing perspective on the taste of the product. In this latter characteristic, sweetness was often a conflicting point, with people rating the sweetness of the product differently.
“Genetic differences in taste or olfactory receptor sensitivity may help account for the extreme reactions that some products get,” the researchers say. “Looking at the responses to polarizing foods could be a way to increase understanding of the biology of personal differences in food choice.”
Ultimately the team hope that their work is a good illustration of the potential of a big-data approach to consumer market research and that not only food companies but any organization with an interest in nutrition will take inspiration from the approach and gain a better understanding of consumer tastes.
