What Is Neuromarketing?

Ale Smidts, a professor at Erasmus University Rotterdam, defined it in his inaugural lecture on 25 October 2002: the aim of neuromarketing is to better understand the customer and their response to marketing stimuli by measuring the processes in the brain directly. Secondary sources credit him with the term; in the lecture itself he does not claim it.

The difference from classic market research lies in the method. A survey asks; neuromarketing records signals while someone watches an ad or reads a page. The signal does not depend on what the person chooses to say, but it does not explain itself either: someone has to interpret it.

Each method measures one thing and leaves others out: - EEG (electroencephalography): electrodes on the scalp record changes in the brain's electrical field. Its temporal resolution is in milliseconds and its spatial resolution around one centimetre (Ariely and Berns, 2010): it detects when activity changes, but locates poorly where it comes from. - fMRI (functional magnetic resonance imaging): it measures the BOLD signal, the blood oxygenation level, which is taken as a proxy for neural activation. It locates activity to within 1 to 10 millimetres, but each measurement takes 1 to 10 seconds. - Eye-tracking: a camera follows eye movements and records what catches visual attention and how long each point is looked at. It shows where someone looks, not what they think about it. The gaze-based heatmap works on the same principle. - Skin conductance (EDA or GSR): sensors measure the electrical conductance of the skin, which the autonomic nervous system controls. ESOMAR links it to the arousal an ad produces. A rise shows intensity, not which emotion lies behind it. - Facial coding: software codes movements of the facial muscles according to the Facial Action Coding System (FACS). According to the review by Barrett and colleagues (2019), a scowl often communicates something other than an emotional state, and how people express an emotion varies across cultures, situations and individuals.

Barrett and colleagues summed it up in 2019: there were no objective measures, either singly or as a pattern, that reliably and uniquely identified one emotion category from another in a replicable way.

The research literature sets three limits before results are used.

The first is sample size. According to the ESOMAR guide, providers recommend samples of 15 to 30 people, against 100 or more in traditional research. Ariely and Berns note that any task in which the response is expected to vary across individuals needs a sample of at least 30. Button and colleagues (2013) estimated the median statistical power of neuroscience studies at between 8 % and 31 %: with low power, real effects are harder to detect and a statistically significant result is less likely to reflect a true effect.

The second is reverse inference. A brain region lighting up does not prove a particular feeling, because a region takes part in several processes. Russell Poldrack showed in 2006 that such inferences are not deductively valid, although they can provide some information. Ariely and Berns warn against anyone who claims to know what a person thinks from brain activation alone.

The third is that a reaction is not a purchase. Some studies link the two. Knutson and colleagues (2007) scanned 26 adults with fMRI, and activity in each of three regions predicted their immediate purchase decisions in the lab beyond what they reported themselves. Venkatraman and colleagues (2015) compared six methods with real sales after TV ads, and fMRI explained the most variance beyond the traditional measures. These are results of individual studies; they do not allow the conclusion that an ad will sell more because it activated a brain region.

Neuromarketing can complement A/B testing and surveys in stages that leave no trace in clicks, such as interest and desire in the AIDA model. It cannot replace them.

The European Data Protection Supervisor (EDPS) published a report on neurodata on 3 June 2024, defining them as information gathered from the brain or nervous system. According to the EDPS, they allow individuals to be uniquely identified, are therefore personal data, and often constitute special categories, for example as biometric data or data concerning health.

The GDPR is more precise than that sentence. Biometric data are a special category under Article 9 only when processed for the purpose of uniquely identifying a person. Health data are special when they reveal information about a person's health status. A study that falls under Article 9 needs one of the exceptions in Article 9(2); the first is explicit consent for specified purposes (point (a)). A study outside Article 9 still needs a legal basis under Article 6.

The EU AI Act adds two rules. Since 2 February 2025 it has prohibited AI systems that infer emotions in the workplace and in education institutions, except for medical or safety reasons (Article 5(1)(f)). Since 2 August 2026 deployers of an emotion recognition system must inform the people exposed to it (Article 50(3)). Software that infers emotions from faces can fall under that definition.

• Ask what each metric measures and how many participants it is based on.
• Pair the signals with an observable task, such as a choice.
• Treat results as hypotheses and test them against sales data.
• Document the legal basis and consent before the first session.
• Keep what was measured apart from what was interpreted.

• Presenting a brain activation as proof of an emotion.
• Reading a measured reaction as a purchase decision.
• Drawing general conclusions from a study with 20 people.
• Quoting effectiveness percentages without a study or date.
• Assuming that body signals are not personal data.