Knowledge in the neurosciences, theory and methodology, is increasingly applied to improve and deepen our understanding of consumer decision processes, affect and cognition, and behaviours, in a young field known as consumer neuroscience; practical implementation of this knowledge on consumers in marketing management is known as neuromarketing. Relevant specialisitations in the neurosciences include neuropsychology, neuroeconomics, and neurobiology. The research has largely focused on studying those consumer phenomena from a neuroscientific perspective at the individual level. Yet, a new stream of research is emerging in recent years of extending the prediction of decisions or behaviours of individuals, based on neuroscientific knowledge gained at that level, to prediction of aggregate-level metrics, particularly at a real-world market-level (e.g., sales, viewership of video content, funding of loans or investments), namely neuroforecasting.

The latter development is explained and advocated by Alexander Genevsky and Carolyn Yoon in their paper “Neural Basis of Consumer Decision Making and Neuroforecasting” (published as a book chapter, 2022 [1]). The authors describe at first the advances made in consumer neuroscience and make the distinction between neuro predictions at the individual-level (i.e., within-individual) and at the aggregate-level (i.e., “forecast the aggregate behaviour of a separate and independent group”). Subsequently, they review knowledge gained on the neural basis of consumer decision making, covering brain areas or structures activated and implicated in various processes and their outcomes; they further relate to neural predictors of decisions (based in individuals). This is followed by review and discussion of the newer stream of neuroforecasting in “using neural data to forecast markets”. Genevsky and Yoon describe methods, findings and insights reached through this stream of research, and discuss issues raised and implications. Notably, neural measures are often used in addition to other types of data (e.g., self-report judgements and choices, biometric measures, eye tracking), where the contribution of the former can be shown to improve the predictions of decisions by individuals or aggregate outcomes of consumer bahaviour.

Of particular interest in their review, Gevensky and Yoon refer to value-based decision making, wherein decisions are guided by underlying preferences or representations of value. Sources of value include rewards and affect. Brain areas commonly implicated are the orbital frontal cortex (OFC), venrtomedial prefrontal cortex (VMPFC), and ventral striatum (VS) located in the temporal lobe. The latter has a significant role to which we will return below. The OFC is found to be associated with encoding of reward value underlying preferences. The OFC is furthermore engaged in the processing of sensory stimuli (e.g., taste, smell, touch) as well as encoding of more abstract stimuli (e.g., aesthetics, facial attractiveness, social stimuli). However, it is noted that while the OFC is linked to preferences driving decisions, it lacks access to motor output networks that support execution of the choice or action (e.g., stretching one’s hand to pick up a product from a store shelf).

Increased amounts of reward or affective value are associated with higher ratings of liking or pleasantness, willingness to pay, and measures of choice. Neuroimaging studies have demonstrated a link between subjective value and activity in both the VMPFC and OFC; that is, these areas are engaged in subjective valuation which may occur during a choice task or just evaluation of options. Genevsky and Yoon note that studies they cite provide “supporting evidence for a close correspondence in regional brain activity between the anticipation of rewarding events, the consumption of enjoyable goods, and the willingness to pay for them” (p. 565). The authors additionally review neuroscientific research on decision making that substantiated the correlation of activity of the VMPFC and VS with subjective value (‘utility’) supporting choice or decision making.

Evidence of positive effects were seen with respect to both decision subjective value (i.e., when a decision is made) and experienced subjective value (i.e., when an outcome is experienced). Inside the VS resides in particular the nucleus accumbens (NAcc) which is associated with affective value. During pre-decisional processes, significant differences in activation of the NAcc, while products are being presented, could distinguish between (consequently) purchased and non-purchased items, suggesting a predictive contribution. Interestingly, when individuals are asked just to rate the attractiveness or pleasantness of items (e.g., faces, houses in images), greater activation in the VMPFC and VS can single out those items that will be preferred (or chosen) in a later task not announced in advance. In other words, as explained by Genevsky and Yoon, researchers they cite “found that activation in the valuation areas in the absence of choices was indeed predictive of subsequent decision making” (p. 567).

• A linkage can be identified between emotional processing and reward processing, wherein the NAcc especially plays a role. Shaw and Bagozzi [2] list a number of primary areas or structures exhibiting neural correlates of emotions. Among them are the medial prefrontal cortex (MPFC); the OFC (implicated in feelings of anger and regret); the NAcc (part of the reward circuit, also engaged in emotional processing related to motivational processes in conjunction with the dopamine neurotransmitter); also named are the amygdala, the insular cortex and anterior cingulate cortex (ACC).
• Specifically with regard to functioning of the reward circuit, they highlight the important distinction between the wanting system, which is more crucial as it accounts for the desire and motivation to obtain rewards, and the liking system, which is focused on responses to pleasures and hedonic sensory stimuli. The NAcc, OFC, ACC, insular cortex, amygdala and other areas are associated with the wanting system. Shaw and Bagozzi explain how these neural areas are related to concepts of consumer decision making such as brand preference and willingness to pay.
• Plassmann and his colleagues [3], who concentrate on implications of consumer neuroscience in regard to branding, refer to the role played specifically by the NAcc (within the VS) in encoding anticipated rewards from (favourable) branded products. Furthermore, they note how activation of the striatum may be related to loyalty (citing previous research by Plassmann): “customers who are loyal to a store as measured by real purchasing behavior … show more activation in the striaturm compared to customers who are less loyal” (p. 24). Additional attention is given to the neural correlations of regions in the prefrontal cortex (PFC) with behavioural measures of value or preferences of brands, and how activity in the ACC can predict individual differences in the influence of brand association on consumer judgements.

The new research area of neuroforecasting aims to reach beyond the ability to predict the choices or patterns of behaviour of individuals, based on their own brain activations, by scaling predictions of behaviour to the aggregate-level of a market or population. Genevsky and Yoon state: “The term neuroforecasting has been used to describe this new direction of neural prediction — where the focus is on forecast of aggregate outcomes” (p. 567, italics in origin, boldface added) — to be distinguished from the focus on prediction of choices by individuals. They suggest that a key to better prediction of aggregate behaviours may lie in neural measures of intervening processes, which are less observable but “responsible for the assessment and evaluation of incoming stimuli”. An important distinction is drawn between studies that focus on the individual as the unit of analysis and inferences or predictions made at an aggregate-level about a stimulus as the unit of analysis.

• Comment: Gevensky and Yoon refer for comparison to inferences made from samples of individuals to predict the behaviour of larger groups. Indeed, with samples as used for surveys, statistical inferences or predictions are made by accumulation of individuals’ responses and generalising them to larger groups of represented populations (e.g., preference shares or ‘market shares’ predicted in simulations based on conjoint models). Subsequently, researchers may seek to compare their predictions to measures or metrics of behaviour in real-world market environments (e.g., actual market shares of brands or products). The authors frame the objective of neuroforecasting as more similar to external validation as in behavioural experiments conducted in real-world environments or related objective independent metrics (e.g., in physical stores, online platforms, geographic market regions). The emphasis is on prediction of outcomes in actual market settings.
• A limitation of neuroforecasting could be linked to differences in the size and type of samples used in surveys vis-à-vis neuroimaging and behavioural lab studies & experiments (e.g., large versus very small sample, its source and sampling method). It needs to be said that the authors attribute the predictive power of the neuroscientific studies for aggregate forecasting to the special properties of the implicit neural data collected and analysed, not to the samples, which would give their predictions greater strength.

Gevensky and Yoon give examples of studies that performed neuroforecasting while focusing on two highly prevalent neuroscientific methodologies, functional magnetic resonance imaging (fMRI) and electroencephalography (EEG). The fMRI studies reviewed spanned different categories as context: (1) songs — uploaded to a music online platform; (2) public health — messages for smoking cessation; (3) microlending — small loans to people in need on an online platform; (4) chocolate — effectiveness of point-of-sale advertisements in a supermarket; (5) crowdfunding campaigns online (investment in entrepreneurs); (6) advertising effectiveness (TV ads of 15 brands from six companies); and (7+8) viewing or sharing media content (video clips, news articles, respectively) online.

Each research offers a particular angle on neuroforecasting with its compatible tasks, measures and analytics. The research about popularity of songs (Berns & Moore 2012) is presumed to be the earliest endeavour in this direction, though it started as a neuroscience study at the individual level while the opportunity came a few years later to compare neural and preference responses to sales outcomes for the same songs. This research set course for following studies. The research on advertising effectiveness (Venkatraman et al. 2015), however, is different in its primary objective and scope. Firstly, since it compares and contrasts the predictive power of results from multiple methodologies, including traditional self-report measures, biometrics, eye tracking, EEG and fMRI. The aggregate-level target metric is market-based advertising elasticity (the percentage change in sales in response to 1% change in GRPs,, i.e., gross ratings points used as a measure of scale of advertising). Hence, it can be considered research in neuroforecasting. But more importantly, it has demonstrated the greater and significant improvement to predictions that can be produced by fMRI beyond traditional measures (eye tracking and EEG seem to parallel or overlap with traditional measures).

These fMRI studies unveil a general and typical pattern: (a) Activation signals in brain areas relevant to preferences, rewards and decision making are well correlated with stated preferences or choices made by the individuals-participants; (b) Evaluations, preference ratings or choice responses of individuals tend to correlate weakly in turn with the metrics of aggregate market outcomes (i.e., being less informative); (c) Yet, the respective signals of activation are associated (correlated) with the aggregate market outcomes (e.g., increased sales, success of loan requests or funding campaigns). Additionally, and more concretely, the neural activity (e.g., in NAcc, MPFC) was predictive of aggregate outcomes, in accordance with their prediction of decisions of individuals (as reviewed above). Some studies specifically show a contribution to predictive power by the neural activity data beyond subjective responses (i.e., self-reported).

A curious implication surfaces from the fMRI studies, perhaps even a conundrum: Subjective, self-reported evaluations, preferences, choices or intentions of consumers tend to be unsatisfactory predictors of their actual market behaviours (related to the known ‘say-do gap’). However, the neural information seems to be robust enough to predict the aggregate-level market outcomes of consumers’ actual behaviours. That so appears, although the neural evidence may be regarded as ‘basic’ or ‘raw’ information of mental states underlying the subjective responses of consumers. The driving values of rewards or affect ‘recognised’ implicitly by our brains are more predictive than what we may think and say about our preferences or expected choices.

EEG studies employ types of measurement and analyses that are reliant on prompt timing of signals (rather than localising them as in fMRI — see note below on differences in resolution). The EEG studies analyse, for example, event-related potentials (ERPs) in response to stimuli presented, de-compose frequency bands, and utilise correlated neural signals across subjects. Studies applying EEG techniques, as cited by Genevsky and Yoon, reveal mainly: (1) the ability of shared signals of neural activity to predict real-world outcomes (e.g., neural responses among study participants while viewing TV series’ episodes or films correspond with their real-world success as indicated by public engagement or box office sales, respectively); (2) an improvement in prediction accuracy that can be achieved when adding EEG measures beyond subjective measures (e.g., EEG metrics improve prediction of aggregate-level response to products on top of survey-based self-reported preferences and intentions to purchase when viewing ads). Studies using sophisticated analyses (e.g., spectral analysis of frequency bands, machine learning), suggest other predictive capabilities.

• Note: The fMRI and EEG methodologies differ in important ways. Whereas fMRI has a clear advantage of higher spatial resolution (detecting specific locations of neural activity signals), EEG offers greater temporal resolution (capturing signals nearer the time of events). However, neuroimaging studies employing magnetic scanners are characterised by high costs and complexity (operational, analytic) — which likely also restrict the size and type of samples in use — and studies destined for marketing purposes further entail ethical issues. An advantage of EEG is in achieving greater flexibility to perform studies in sites outside the lab. (A similar capability is also available in eye tracking, but in both techniques additional corrections have to be made to account for measurement inaccuracies and sources of noise from the environment.)

The source of predictive power in neuroforecasting, wherein individual-level neural responses are scaled-up to aggregate-level behaviour, may be attributed to just a subset of brain regions activated during a decision-making process (in reference to a theory suggested by Knutson and Gevensky 2018). Genvensky and Yoon suggest that specifically the basic affective responses, originated in subcortical circuits, “may represent a more universal, or generalizable, measure of the response to a stimulus” (p. 573), that is, having the stronger impact in ‘shaping’ the overt response. They refer to a framework of affect-integration-motivation: a decision stimulus initially elicits an affective reaction, which is then integrated by (or with) high-order cognitive processes (e.g., preferences, concerns), consequently followed by a motivational state (approach or avoidance). The response outcome is a manifested behaviour of the individual, ending his or her decision process.

Eventually, however, the authors assign the greater weight in directing the response to the affective component; the neural response that “represents the generalizable response” is particularly associated with the NAcc. It is proposed that neural activity (especially affective) may be a more representative, “generalizable index of preferences across individuals”, which may thereby be better related to aggregate behaviour (e.g., compared with traditional, survey-based self-report measures). So, we may ask, are beliefs, perceived values, preferences and intentions, products of cognitive processing, not much more than interference or distraction on the way to the final decision? When self-stated, are they truly inferior to signals of neural activity? The answer to the first question should be negative; the authors also do not neglect more rational considerations. The answer to the second question is probably that there are enough deficiencies in self-stated preferences, for instance, that neural information can offer an improvement to predictions of real-world behaviour. The propositions made by Genevsky and Yoon invite a more comprehensive discussion of the interactions between affect and cognitions, such as when and how affective reactions and cognitive processes are integrated, and which has a stronger impact, or are they interlinked.

The line of thought on forecasting brought by Gevensky and Yoon, supported by research insights, is unquestionably original, intriguing, conceivable, and deserves serious consideration. Yet, and at the same time, it raises difficulties and leaves some open questions for debate. First, the approach raises again the question whether prediction should be founded on a sound and clear explanation or can predictions be free of understanding the underlying processes. Second, information on neural activity may better be utillised for refining predictions beyond traditional measures and also in combination with other complementary methods (e.g., eye tracking, biometrics). Third, the approach may need to account for the balance in influence between affect and cognitions on decisions to gain greater acceptance.

Ron Ventura, Ph.D. (Marketing)

References:

[1] Neural Basis of Consumer Decision Making and Neuroforecasting; Alexander Gevensky and Carolyn Yoon (2022); in APA Handbook of Consumer Psychology [Chapter 27, pp. 563-577], L.R. Kahle (Editor-in-Chief), The American Psychological Association (available for reading online, viewed July 2026)

[2] The Neuropsychology of Consumer Behavior and Marketing; Steven D. Shaw and Richard P. Bagozzi, 2018; Consumer Psychology Review, 1, pp. 22-40

[3] Branding the Brain: A Critical Review and Outlook; Hilke Plassmann, Thoman Zoega Ramsoy, & Milica Milosavljevic, 2012; Journal of Consumer Psychology, 22, pp. 18-36

Additional Readings:

[a] Brain and Brands: Developing Mutually Informative Research in Neuroscience and Marketing; Tyler K. Perrachione and John R. Perrachione, 2008; Journal of Consumer Behaviour, 7, pp. 303-318

[b] Heart and Mind in Conflict: The Interplay of Affect and Cognition in Consumer Decision Making; Baba Shiv and Alexander Fedorikhin, 1999; Journal of Consumer Research, 26 (December), pp. 278-292

[c] A Gateway to Consumers’ Minds: Achievements, Caveats, and Prospects of Electroencephalography-based Prediction in Neuromarketing; Adam Hakim and Dino J. Levy, 2018; WIREs Cognitive Science (Wiley), pp. 1-21 (doi: 10.1002/wcs.1485)