Neuromarketing on brand extension evaluation: from neural data analyses to deep learning-based applications

Abstract

On average, each company invests about 12% of its annual budget in marketing. This includes a consumer survey to pinpoint what consumers want. Even though companies invest a considerable amount of money every year in consumer surveys, consumers often do not know exactly what they want and respond to, so traditional methods such as surveys and interviews often do not produce satisfactory results compared to the time/costs invested by companies. To solve these problems, a new convergence field was proposed called “neuromarketing”, the use of methods or results of neuroscientific researches to the marketing activities. The current dissertation is a neuromarketing study to investigate the brand extension strategy, which is one of the important branding strategies, consisting of two human experimental studies and following a deep learning (DL)-based application study. Chapter 1 aims to introduce the background of two keywords: neuromarketing and brand extension. Firstly, neuromarketing is a convergence field of cognitive neuroscience, computer science, and marketing, finding marketing insights by investigating consumer’s neurophysiological responses toward marketing strategy or developing a method quantifying cognitive/affective states of consumers using neuroscientific apparatus. A lot of neuromarketing studies have been conducted and published since 2002 when the neuromarketing term was first used. Furthermore, google hits and related firms are also being increased. Secondly, a brand extension refers to the use of a well-established brand name for a new offering. While the success of brand extension allows companies to reduce marketing costs and entry barriers to entry into new markets, failure adversely affects the parent brand's image. Hence, many researchers have conducted studies to find exogenous factors in successful brand extension, but they could not provide evidence for the endogenous evaluative process. Neuromarketing researchers have been using electroencephalographic (EEG) to investigate consumer’s evaluation of brand extension. However, those have some limitations yet. First, they only focused on goods-to-goods brand extension. Service is accounting for the main part of the contemporary industry, and it has distinguished properties from goods. Nevertheless, the evaluation process of service offering is relatively unexplored. Therefore, the aim of Chapter 2 is to explore the evaluative process for brand extension in the service offering. In the experiment, participants reported the favor about brand extension pairs, consisting of parent brand name (S1) and extension service name (S2) during EEG measurement. I divided stimuli into three groups (i.e., high-, low-, and mid-fit) based on the individual’s affirmative response. Successively, I selected only stimuli that were considered as high or low-fit for most participants and named them as a high or low population-fit group. Based on this stimuli grouping method, I compared event-related potentials (ERPs) using paired t-tests. The results showed significant differences between population-fit groups at N2 and P300 amplitudes, but subsequent decoding analysis revealed that N2 was an inappropriate indicator for use in interpretation of the cognitive process for a brand extension due to high intersubject variability. Therefore, I propose that the evaluation process of service-to-service brand extensions is similar to the evaluation process of product-to-product brand extensions, comparing current and previous findings. As a second limit, previous studies have used only EEG, an inappropriate device for measuring in-depth brain activity. Therefore, the purpose of Chapter 3 was to conduct functional magnetic resonance imaging (fMRI) study to investigate the involvement of in-depth brain areas in consumers’ evaluation of the brand extension evaluation process. During the fMRI scan, participants looked at the name of the beverage brand and that of extension goods selected in the beverage or household appliance category. They responded to their acceptability for a given brand extension. Behavioral results revealed a noticeable pattern in which participants reacted more obvious to negatively felt stimuli. A brain connectivity analysis supported this by showing denser activation throughout the whole brain during positive responses. The general linear model (GLM), including contrast and parameter modulation models, presented that insula is a significant area for brand extension evaluation and its role is related to the emotional process. The fMRI analyses also showed lateralized activations in the insula as well as sensorimotor areas. Therefore, we propose a hierarchical clustering-based connectivity analysis, which shows that neural activity can be separated into sensorimotor and evaluative clusters. Based on the results of insular activity and behavioral response, I suggest that the emotional switch precedes the cognitive evaluative process as shown in Chapter 2. Finally, previous studies could not deviate from a basic research level, not showing a feasible application. Thus, Chapter 4 aims to develop a deep learning model to estimate how consumers evaluate brand extension from their brain activity. To this end, the results of Chapter 3 were used to approach this problem. As a meaningful result, I could develop an fMRI-based deep learning model. I extracted 2D feature patches from preprocessed fMRI images in Chapter 3 and augmented them by rotating and shearing to improve a low number of samples. A convolutional neural network (CNN) architecture was used to classify fMRI images in response to the visual presentation of each brand extension sample into one of the two classes: acceptable vs. non-acceptable. The CNN model was evaluated with 10-fold cross-validation and showed almost 90% of accuracy. Furthermore, I suggest a visualization method to find activated brain areas during the task, by showing class-discriminative regions using weakly supervised learning. The results indicate that one can predict how a consumer evaluates a new brand extension proposal only by their brain activity. As concluding remarks, Chapter 5 summarized the main findings of the current dissertation and provide marketers implications for brand extension.