Can brain activity predict chocolate sales? In search of the buy button

Brain icon with a shopping cart
Have researchers really found the holy grail of neuromarketing?

By guest blogger Julia Gottwald

Coming up with the perfect recipe for crisps or the ideal marketing strategy for a soft drink used to depend on explicit measures. In focus groups and surveys, consumers were asked which product tasted best or which commercial was most appealing. But these measures are imperfect: consumers may choose to hide their true opinions or they might not be fully aware of their own preferences. Food and drinks companies need more objective measures. Currently their best hope is functional magnetic resonance imaging (fMRI).

The idea is that somewhere in the brain, a “buy button” is hidden away: a region (or combination of regions) that influence your purchase decision. The promise of neuromarketing is that one day, we will be able to find this region, record its activity when you watch an ad or sample a product, and then predict how well this product will sell. So far, the success has been limited. But in a recent study in NeuroImage, Simone Kühn from the University Clinic Hamburg-Eppendorf and her colleagues claim to have found “multiple ‘buy buttons’ in the brain”.

The researchers showed six different ads for the chocolate bar “Duplo” to 18 healthy women while they underwent fMRI. In these ads, the chocolate bar was presented either next to the face of a woman; a couple; a group of people; in two hands; in two hands and with an additional slogan; or next to a toothbrush (the toothbrush was meant to serve as a control condition).

The researchers measured brain activity during the adverts in eight brain regions thought to be involved in making purchase decisions. Then they entered these measures into a formula that weighed key areas such as the nucleus accumbens and medial orbitofrontal cortex more strongly than others, which were expected to have less influence. Some areas, such as dorsomedial prefrontal cortex, were even weighed negatively, assuming that higher activity in these regions would make a purchase less likely. The team then used this “fMRI sales forecast value” to predict the sales of the chocolate bar. After the scan, the women were also asked which ad they liked most.

Next, the researchers ranked the six ads in different ways: according to the fMRI forecast; according to the women’s explicit preferences; and according to actual sales in a real German supermarket where the chocolate bar was sold in a display showing one of the six different ads, with a different advert used each week. In total, more than 63,000 customers were recorded, of whom 317 bought the chocolate bars. The six weeks were not consecutive and were carefully chosen to be uninfluenced by holidays and promotional periods.

The fMRI-based forecast was a surprisingly accurate predictor of sales, while self-reports performed poorly. The two ads which did best in the supermarket (chocolate bar in front of a group and next to a woman) were also predicted to be most successful by the fMRI sales forecast value. In contrast, the two most-liked ads from the self-reports (hands without text and, somewhat surprisingly, the toothbrush) were actually the least and third to least successful ones.

These seem like impressive results for the field of neuromarketing, but I had a number of reservations. For instance, while the authors tried to justify their selection of brain areas, arguing that these regions had previously been implicated in neuromarketing research, it is not entirely clear how they came up with their specific formula to calculate the “fMRI sales forecast value”. A clearer rationale would have been helpful.

The new results also contradict earlier studies, which found that asking people directly about their preferences was far superior over modern measures such as fMRI. Note though that in this study, people were asked which ad they liked best. But liking does not equal wanting. You might not like an ad very much, but it could still increase your desire for a product. A more useful question for the self-report part of this study would have been: “How much does this ad make you want to buy the chocolate?”

Also, there was a potential effect of the order in which the ads were displayed in the shops. The most successful ad, which was also predicted to be most successful by the fMRI based forecast, was the one that was used first in the German supermarket. The authors argue that most customers would have bought and eaten Duplo before, since it is a very popular sweet in Germany. However, it is easy to see how a new display in your local supermarket could encourage you to buy this product, regardless of the ad attached to it. Future studies should take this effect into account and maybe start with a shop display without any ads in the first week.

But perhaps most concerning is that one of the study authors, Enrique Strelow, is only listed in the paper as a member of the Justus-Liebig University of Gießen, even though he is also Head of Shopper Communication at Ferrero, the company that makes the chocolate bar Duplo. This is a clear conflict of interest, which was not disclosed in the paper. From reading the article, it is clear that the authors had cooperated with Ferrero, since they had early access to the advertisements which would later be placed in the supermarket. However, one of the authors being an employee of this company is a different thing altogether. Since these conflicts of interests can implicitly or explicitly influence study design, analysis, and result presentation, it is common practice to disclose any links to industry. Why this information was not made publicly available for this paper is unclear.

Nonetheless, the study highlights how neuroimaging-based market research in relatively small groups could one day reliably predict product sales on a large scale. Giving companies access to such sensitive data holds the potential for abuse and should therefore be carefully regulated. While there are strict regulations for fMRI based research in academic settings, such rules are lacking for industry applications. A public debate about the merits and dangers of neuromarketing is therefore more and more pressing.

Multiple “buy buttons” in the brain: Forecasting chocolate sales at point-of-sale based on functional brain activation using fMRI


Post written by Julia Gottwald for the BPS Research Digest. Julia is a PhD student in the Department of Psychiatry at the University of Cambridge. Her book, Sex, Lies, & Brain Scans: How fMRI reveals what really goes on in our minds, is out now. Follow her on Twitter: @julia_gottwald.

3 thoughts on “Can brain activity predict chocolate sales? In search of the buy button”

  1. Nice post.

    Another problem is that this study had an effective sample size of just 6 – i.e. there were six ads, and the key comparison was real world performance of ads vs. fMRI response to ads. So the study was effectively plotting a scatterplot with six points, and finding a correlation… which I’m not sure is meaningful.

Comments are closed.