Advertising effectiveness is a topic of prime interest in the area of advertising; it is especially important to those engaged in the practice of marketing and advertising. The methods for measuring advertising effectiveness have mostly been psychologically-driven (e.g., consumers’ attitudes, memory, intentions) and market-driven (e.g., based on product sales data). The consumer-level models, for instance, usually distinguish between direct responses to the ad and subsequent responses to the target brand, product or service. Models based on sales data often have to derive the portion of sales increase that can be attributed correctly to an advertising campaign apart from other actions and events in the marketplace. More recently, a new dimension of measures is being added, drawing from the fields of neuroscience and physiology. Phenomena and processes (e.g., perception, cognition and affect) that are addressed in the more general context of consumer neuroscience are also relevant to consumer response to advertising.

It is an intriguing topic, but with a multiplicity of different kinds of measures that may be relevant and useful in assessing the effectiveness of advertisements, it becomes more complex to navigate, and deserves more investigation. A team of researchers headed by Vinod Venkatraman chose a specific approach to study the relations between three categories of measures: traditional self-reported (explicit) measures, neurophysiological measures, and market-level advertising elasticity (measuring relative sensitivity). The researchers’ overall goal was to “link traditional and neurophysiological measures to actual market responses to advertising in terms of advertising elasticities.” Their article was published in a Special Issue on Neuroscience and Marketing of the Journal of Marketing Research in 2015 [1].

The researchers propose a theoretical framework for explaining the relevance and contribution of traditional and neurophysiological measures to evaluating consumer response to ads. They associate those measures, as described summarily below, with four mental constructs with respect to advertising: (a) Attention (top-down and bottom-up); (b) Affect (arousal and valence); (c) Memory (encoding, consolidation, and retrieval); and (d) Desirability (‘wanting’).

The traditional measures are self-reported by consumers (explicitly induced, e.g., in surveys and experiments). These measures are assigned to any of the theoretically-associated constructs. For example, liking and relevancy are regarded as measures of attention. Valence of affect is reflected by pleasantness whereas intensity is an expression of arousal. The researchers focus on the retrieval function of memory (recall and recognition). Desirability can be measured by changes in purchase intent. The assignment of measures in not exclusive (e.g., excitability -> attention and affect, liking -> attention and desirability). Eventually, three traditional measures were selected as most representative for further analyses: Liking (ad-related), Purchase Intent (product-related), and Recognition.

The neurophysiological measures can be divided into four main sub-sets: implicit measures (Implicit Association Test [IAT]); measures generated through neuroscientific methods (EEG, fMRI); physiological measures (biometrics), and measures derived from eye tracking. Venkatraman et al. elaborate on the specific measures to be derived from each type of method and how they can be linked to any of the four constructs. Not all possible measures reviewed were necessarily used in their research.

In this domain in particular, it can be seen that measures from a given type of method may be associated with different constructs or distinct aspects of a construct. For example, in eye-tracking, areas of fixation reflect on attention whereas pupil dilation is linked to arousal (affect). Three biometric measures are considered: heart rate (pulse) is linked to attention, breathing (respiration) rate is linked to arousal (but heart and breathing rates are inter-linked, both being associated with affective processes), and properties of skin conductance also inform on arousal (less on valence). Measures derived from fMRI can be associated with each one of the constructs (e.g., the association of desirability with ‘wanting’ receives meaning especially in this context: while ‘wanting’ is connected with ‘liking’, a state of ‘wanting’ is linked to seeking rewards [activated in the striatum and other subcortical areas] and therefore should be distinguished from the less committing, hedonic state of ‘liking’).

The researchers chose to measure and analyse the effectiveness of TV ads. Their research included 37 TV ads for 15 unique brands from 6 companies. They constructed a multi-method experimental design in four phases: Phase 1 concentrated on traditional measures; Phase 2 included the measures of biometrics and eye-tracking; Phase 3 focused on fMRI (functional magnetic resonance imaging), and Phase 4 focused on EEG (electroencephalography). The estimation of advertising elasticities followed; analyses of the elasticities were performed in three stages of modelling vis-à-vis the variables of traditional and neurophysiological measures.

Venkatraman and his colleagues conceptually placed the neurophysiological measures as a ‘second layer’ between the traditional measures, which have been used for several decades already, and advertising elasticities, measures expressing the real-life outcome (effectiveness) of advertising in the market. The aim in this approach was to answer this primary research question: “Which of the [recent neurophysiological] measures explains the most variance in market response [elasticities] beyond the traditional measures that have been used in theory and practice for many years” (the text in brackets is added here for elucidation).

• Advertising elasticity denotes “the percentage change in sales due to a 1% change in the advertising measure being utilized (e.g., expenditures, GRPs)” — this research applied the index of Gross Rating Points (GRPs) as a measure of the scale of advertising executed.

In the first modelling stage, as preliminary analyses, a separate regression model of elasticities was estimated for each set of measures — traditional, IAT, biometrics + eye-tracking, fMRI, and EEG. Results from these analyses showed, as explained by the authors, that “traditional variables were by far the best predictors of ad elasticities” (they produced an improvement of 72% in adjusted-R² beyond company-specific dummies of fixed-effects).

The second modelling stage, that constitutes the main body of analyses, addresses the primary research question as quoted above. A regression model was estimated for each subset of implicit and neurophysiological measures, after controlling for the effects of traditional measures (and company-specific dummy variables). This series of regression analyses revealed that, when controlling for traditional measures, only the fMRI measures were significant predictors of ad elasticities (the strongest [significant] effect was found in relation to activation in the ventral striatum). The authors add: “Consistent with this result, fMRI measures were the only variables to produce a positive percentage increase in adjusted-R²” (p. 447). The advantage of fMRI can be attributed to the strength of this methodology in high spatial resolution that is critical to identifying activated brain areas; as lucidly stated by Shaw and Bagozzi (2018): “Although fMRI has much higher costs than EEG, in terms of scanner maintenance, operations, and participant incentives, and lower temporal resolution (on the order of seconds), the spatial resolution allows one to discern between cognitive pathways” (p. 29, [2]).

The third modelling stage was executed as a complementary set of analyses, aimed to check if any of the neurophysiological measures, other than fMRI, had an explanatory power equivalent to the self-reported, traditional measures. This time, effects of each set of neurophysiological measures were estimated simultaneously with the traditional measures (i.e., without controlling for the latter). The analyses indicated that “eye-tracking and EEG measures were moderate predictors of ad elasticities”, similar to the self-reported measures, suggesting that the eye-tracking and EEG measures could explain the same variance in ad elasticities as the traditional measures (i.e., they seem to overlap).

This research was not free of trouble. The researchers had to overcome a number of methodological obstacles and data complications. Due to the differences between companies, products or services, and the market data provided, elasticities were estimated in a separate model for each of the companies included (consumer products, financial and travel services). The dependent variable also was defined differently between companies (log of sales or market share as log of sales ratio, one company provided its own elasticity estimates). Other difficulties arose in executing the various measurement methods, each method with its own unique requirements, and different types of data used and measures produced.

Additionally, it can be noticed in the results reported that the results on effects of the various traditional measures were mixed and partial across models including different categories of implicit and neurophysiological models. With regard to the prospective new measures tested, the greatest contribution, the only statistically significant one, to adjusted-R² was produced by fMRI [59%]. Comparatively, the contribution from eye-tracking and EEG were negative low (-7% and -3%, respectively), but non-significant. The more conspicuous result, yet, was obtained with regard to biometrics: a decrease by 29% in adjusted-R² after the traditional measures.

The authors point to the important correlations they found between the traditional measures and selected neurophysiological measures of different types (e.g., eye-tracking & biometrics, fMRI, EEG), and especially how they relate together to the higher-level constructs (attention, affect, memory, and desirability). These findings clarify the relations of measures to pertinent advertising constructs, and support their internal validity, demonstrating consistency across samples and measures.

Venkatraman and his colleagues recognise that the information to be gained when aggregating a neurophysiological measure over 30 seconds of the whole TV ad clip is limited; in biometrics, eye-tracking, and EEG, for instance, more can be learned by taking measures at different points during the ad screening. The authors also confer that these methods are more expensive and less accessible. Further research is advised to study the incremental value of each method relative to its cost and accessibility. It may be added that using neuroscientific methods, foremost fMRI, may raise substantive ethical concerns when application is intended for commercial purposes (e.g., implications of subjecting participants to a medical procedure, location of the magnetic resonance machines in hospitals, and sharing of the costs between business and non-profit research institutions or governmental agencies); hence, the use of fMRI is still acceptable primarily in scientific research [3].

The contributions of Venkatraman and his colleagues are more salient in setting theoretical foundations of advertising constructs (at the mental consumer-level) with a methodological approach to measuring them by different types of self-reported (explicit), implicit, and neurophysiological measures. The contributions shown from neurophysiological measures in explaining market-level ad effectiveness (via elasticities), however, is somewhat disappointing. It is less convincing mainly in view of the effort needed in resolving various data complexities, measurement procedures, and multiple models that need to be estimated separately. Undertaking the research enterprise was nevertheless important for the lessons that could be learned from it. First, gaining better understanding of the capabilities and limitations in employing neurophysiological methods and measures; Second, the traditional measures remain useful and relevant and should continue to deserve their place in the consumer / advertising researcher’s toolkit; Third, if practitioners do decide to include neurophysiological measures in their advertising research, the method of fMRI seems to be the best candidate to employ.

Ron Ventura (Ph.D.), Marketing

Notes:

[1] Predicting Advertising Success Beyond Traditional Measures: New Insights From Neurophysiological Methods and Market Response Modeling; Vinod Venkatraman, Angelika Dimoka, Paul A. Pavlou, Khoi Vo, William Hampton, Bryan Bollinger, Hal E. Hershfield, Ishihara Masakazu, & Russell S. Winer, 2015; Journal of Marketing Research, 52 (August Special Issue Neuroscience & Marketing), pp. 436-452.

[2] The Neuropsychology of Consumer Behavior and Marketing; Steven D. Shaw and Richard P. Bagozzi, 2018; Consumer Psychology Review, 1, pp. 22-40 (first published online December 2017).

[3] Brains 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 (available from ResearchGate.net, retrieved June 2021).