Launching a new product in the grocery sector involves a high degree of financial risk. If a manufacturer or retailer overestimates consumer demand, they are left with excess inventory that takes up valuable shelf space and increases storage costs. If they underestimate demand, they face stockouts that frustrate shoppers and result in lost revenue.
Traditionally, managers rely on past sales data and consumer surveys to forecast demand. These conventional tools often fall short because historical data cannot always predict future trends, and surveys suffer from the intention-action gap. This gap is the common discrepancy between what consumers state they intend to buy and what they actually purchase.
A recent study published in the Journal of Marketing Research explores a biological approach to this forecasting problem. The research team investigated whether functional magnetic resonance imaging, or fMRI, could improve predictions of store manager adoption and consumer sales for packaged goods.
The Two Hurdles of Product Adoption
For a new product to succeed in a large retail chain, it must clear two distinct hurdles. First, decentralized store managers must choose to stock the item on their shelves. Second, individual shoppers must decide to buy it.
Paulo Albuquerque, a researcher at INSEAD, worked with a cross-institutional team of colleagues to understand how different types of data predict outcomes at each of these stages. The researchers suspected that the data sources best suited for predicting a store manager’s operational choices might not be the same ones that predict a consumer’s purchasing choices.
To test this, the team focused on neuroforecasting. This is an emerging field that uses brain-based data from small groups of individuals to predict broader market-level behavior. Previous research in consumer neuroscience has linked specific brain regions to product desirability, the evaluation of price, and the emotional response to spending money.
Testing Predictions in the Grocery Aisle
The research team collaborated with a large German retail chain operating nearly 7,000 grocery stores. They tracked the performance of 56 consumer packaged goods, which included soft drinks, alcoholic beverages, packaged foods, and prepared meals. They monitored both the percentage of stores that chose to stock each product and the average per-store consumer sales over a two-year period.
The researchers then gathered predictive data from four different categories. The first category consisted of observable market variables like product price and promotional activity. The second category contained survey responses from 1,451 customers who were demographically representative of the retail chain’s broader shopper base.
The third category consisted of fMRI brain scan data. The researchers recruited 44 participants to view images of the 56 products alongside their prices while inside an fMRI scanner. The machine measured blood flow to three specific regions of the brain associated with evaluating products and calculating value.
The final category of data came from an incentivized purchase task completed by those same 44 participants. Instead of asking about hypothetical purchase intentions, the researchers gave participants money and the option to actually buy the products, ensuring the decisions reflected real consumer behavior. The team then fed these four data sources into predictive statistical models to see which variables best forecasted market success in the second year of the products’ life cycles.
Different Data for Different Decisions
The analysis revealed that survey data and fMRI data excelled at predicting entirely different stages of the retail process. When predicting store manager adoption rates, the large-scale consumer survey was the most effective tool. Adding survey data to the baseline market data reduced prediction errors by 11.3 percent.
The fMRI data, on the other hand, performed poorly when forecasting store manager behavior. The researchers argue that this divergence makes sense based on the design of the fMRI task. The brain scans captured individual preference signals from people acting as shoppers, which does not reflect the strategic priorities and operational constraints that guide a retail manager’s stocking decisions.
However, the fMRI data outperformed all other inputs when predicting actual consumer sales per store. Using fMRI data alone improved sales forecast accuracy by 18 percent compared to the baseline model. The brain regions associated with initial product desirability and emotional intensity were particularly strong predictors of how well a product would sell once it reached the shelves.
Forecasting Sales of Innovative Products
The researchers also investigated how a product’s perceived novelty affected the predictive models. They used the large consumer survey to assign an innovativeness score to each of the 56 products. The analysis showed that the predictive advantage of fMRI data was heavily concentrated among the most innovative products.
For standard or familiar grocery items, traditional market data and survey responses provided adequate sales forecasts. For highly novel products, traditional metrics struggled to accurately project consumer demand. In these instances, the brain scans captured affective and anticipatory responses that survey respondents either could not articulate or failed to recognize.
Calculating the Financial Impact
The research team translated these predictive improvements into estimated financial savings. To do this, they calculated the cost of holding excess inventory across the retail network. Based on average product revenues and a standard inventory holding cost rate, they estimated the direct monetary benefits of improved forecast accuracy.
When evaluated strictly on predicting consumer sales per store, the inclusion of fMRI variables generated an estimated annual cost savings of about 115,000 euros for a set of 50 products. When both managerial adoption and consumer sales stages were considered together, the survey variables yielded the highest overall cost savings due to their strong performance in predicting store manager behavior.
The study also produced a notable finding regarding data collection methods. The researchers administered a brief survey to the 44 lab participants right after their brain scanning sessions. This small convenience survey predicted market outcomes almost as accurately as the large survey of 1,451 representative customers.
The authors suggest that their findings open new avenues for neuromarketing vendors. Most commercial neuromarketing firms currently focus on using brain data to optimize advertising content. Expanding their services to include demand forecasting could create a new business opportunity for these companies.
