Can Brain Scans Predict What Will Sell?
Launching a new product is an expensive bet. Companies can study past sales, test prices, run promotions and ask consumers what they intend to buy. Yet two basic questions remain hard to answer: Will retailers stock the product? And, if they do, will consumers actually buy it?
Mistakes are costly in both directions. Underestimating demand creates stockouts and lost sales; overestimating it ties up capital in unwanted inventory. And conventional forecasting has an obvious weakness: what consumers say they will buy does not always match what they eventually do.
As a new study in the Journal of Marketing Research puts it, “These limitations highlight the need for alternative approaches to demand forecasting, ones that can capture deeper, often nonconscious drivers of consumer and retailer behavior.”
Marton Varga of Bocconi’s Department of Marketing, together with Paulo Albuquerque and Hilke Plassmann of INSEAD, Anita Tusche of Queen’s University, Nadine R. Gier-Reinartz of Heinrich Heine University and Bernd Weber of the University of Bonn tried to establish whether the human brain might provide some clues. Their study tests neuroforecasting: using brain activity measured through functional magnetic resonance imaging, or fMRI, to predict real-world market outcomes.
From the scanner to the supermarket
The researchers worked with one of Germany’s largest grocery retailers, which operated 6,934 stores at the time of data collection. They examined 56 food and beverage products, while distinguishing between two different stages of commercial success. The first was whether store managers decided to carry a product. The second was how much consumers bought once it was on the shelf.
To forecast those outcomes, the researchers combined ordinary market information such as price and promotions with a survey of 1,451 customers. They then added something much less conventional: brain data from 44 participants who evaluated the products and their prices while undergoing fMRI scanning. The scans focused on brain regions previously associated with desirability, valuation and emotional reactions during purchase decisions.
Brain data improve sales forecasts more than store-adoption forecasts
When the researchers tried to predict second-year sales per store, adding fMRI information to ordinary market data reduced forecasting error by 18%. Adding survey information reduced it by 7.6%. But when predicting whether store managers would adopt a product, the result reversed. Surveys improved forecasts by 11.3%, while fMRI data added only about 2%.
The distinction makes sense. People inside the scanner were responding as consumers: Do I want this product? Is it worth the price? Would I buy it? Store managers face a different problem involving shelf space, turnover and expectations about the wider market.
“The neural responses captured represent individual preference signals rather than managerial adoption considerations.”
The authors underline that difference: “the neural responses captured represent individual preference signals rather than managerial adoption considerations.”
The implication is that more sophisticated data are not automatically better data. Their value depends on whether they capture the kind of decision a company is trying to predict.
The bigger advantage appears with innovative products
For relatively familiar products, surveys remained competitive. But for the 35% most innovative products in the study, fMRI-based models predicted sales significantly better than survey-based models. That is a weaker spot of conventional forecasting, because when a product resembles something consumers already know, people can rely on existing preferences and experience. With an unfamiliar product, they may find it harder to explain—or even consciously recognize—how they will react.
“Neural measures, in particular, may offer additional predictive value for unfamiliar or innovative products, where affective and anticipatory responses not fully captured by self-reports play a greater role.”
As the researchers write, “Neural measures, in particular, may offer additional predictive value for unfamiliar or innovative products, where affective and anticipatory responses not fully captured by self-reports play a greater role.”
For companies, that points to a specific use case: fMRI may be most worth considering not for routine line extensions, but when launching products so novel that historical sales and consumers’ stated intentions provide an unusually weak compass. This may be the strongest argument for neuroforecasting: its potential advantage is not necessarily in predicting the ordinary, but rather in predicting what happens when there is little precedent to rely on.
Is an fMRI scan worth paying for?
Better prediction is of any practical use only if its value exceeds its cost. The authors therefore compared forecasting improvements with the estimated cost of obtaining the data. For a project covering around 50 products, external providers quoted roughly €10,000–€22,000 for a representative survey, versus €29,000–€35,000 for fMRI data.
Their economic calculations suggest that fMRI can generate substantial savings when the objective is forecasting sales. But once both store adoption and consumer sales are considered, surveys can offer greater overall value because they perform particularly well at predicting retailer decisions.
Overall, the study therefore suggests that different signals answer different questions. Surveys may be especially effective for understanding whether the market believes a product will succeed. Neural measures may add more when the challenge is anticipating consumer response to something genuinely new.
