Brand equity is becoming an increasingly important driver of a company’s success. Strong brands sell more products, can offer products at a higher price or can turn customers into loyalists. But how to build a brand? Neurensics has developed a suite of methods to build and increase brand value. At the core of these methods is the thesis that a brand is an emotional preference based on unconscious associations. Evidence for this comes from research on the similarity of the neural networks of emotion and the neural systems that assign value to products and brands (Plassmann et al., 2012). Strong brands evoke activity in those regions of the brain that make you feel good about what you see, hear or touch, and can increase these feelings. They achieve this by associating their brand cues with intrinsic values that people have, ranging from low-level desires (for e.g. food, sex, money, praise, etc.) up to higher-level aspirations (such as independence, esteem, fame, etc.) (Alsem & Kostelijk, 2008).
If we want to increase or change the associations brands evoke, we have to take translational steps. We must understand how brands tie to high-level aspirations and low-level desires, both of which are for a large part implicit. Since traditional market research methods cannot assess the implicit well, and “standard” fMRI only goes up to the secondary emotions that are hard to translate into concepts that marketers and creatives can work with, we need new methodologies to fill the void.
To accomplish this, we developed a 3600 framework aimed at building brands from the brain up. First we map implicit (i.e. unconscious or preconscious) high- level associations with brands, using a rapid implicit association task (RIAT). With the RIAT we identify which associations are key drivers for targeted consumer behaviors, such as loyalty, switch intention, or conversion. Clients use these associations as input for development of marketing material. Next, we use fMRI and deep learning to model how ensembles of low-level brain networks give rise to those higher- level associations. Finally, we apply the models to marketing concepts, storyboards or ads to validate if the developed marketing stimuli are indeed able to activate these associations. With our client, we applied this methodology to the Dutch Health care market to investigate the best strategy to attract potential customers.
In the identification phase, we chose intention to switch of non-customers as the target behavior. In a large group (N=1112) we collected the intention to switch from our customers brand to the competition brands, or vice versa. Next, choices and response times to brand-association pairings were collected using the RIAT (online). These responses were used to quantify how well associations fitted the clients’ brand relative to the competition field. Crucially, we modeled the data to identify associations that were reliably increasing the probability of showing switch intention value using machine learning. Four target associations were isolated and used by the client to develop two commercials in different styles. In order to validate the commercials, we developed network ensemble models by having professional actors mentally evoke these associations in the scanner. We modeled these mental states from low-level emotional networks using a deep-learning framework. The two commercials were tested against these networks in a consumer population to validate if they indeed were able to activate the targeted mental states.
Figure 1 The steps of the Neurensics system 1 brand cycle for our health client. A: identification of ownership and predictive strength of the associations. B: graph of the reliability of the machine learning step. C. Machine learning results that led to the selection of four target associations. D: linear estimation of nonlinear neural network predicting honesty. E: results of two commercials on the four target associations from which commercial 1 was picked. See text for details.
The RIAT is used to obtain data in three dimensions (fig 1A). The x-axis shows association ownership: how strongly the associations predict brand preference relative to the competition. An association on the right hand is “owned” by the client brand and associations on the left-hand side are owned by the competition. The other two dimensions quantify the results of the machine learning analysis. The y-axis shows how well the associations predict the brand value. A positive value means a positive contribution to the brand
value, with higher value indicating a higher intention to switch. Similarly, a negative value indicates a lower intention to switch. The last dimension of the association field is the color. Red associations are System 1 drivers: the fastest responses are more predictive of brand values. Blue associations are System 2 drivers, here slower responses are more predictive of brand value. In the example, caring is owned by the client and moderately predicts switch intention (system 1). Reliable is owned by the competition and strongly predicts switch intention (system 2) making it an interesting target association to attract new customers.
After a validation step to ensure model performance (panel B), the results were visualized (panel C). Affordable, reliable, honest and distinguished were chosen by the brand as target associations on top of the typical brand values of the client (empathetic and stable). This resulted in a briefing for the creative company that made two commercials in two directions. Neurensics constructed brain networks predicting these four associations with fMRI and deep-learning (panel D). The two commercials were tested against these networks in a representative consumer sample. The results showed that the commercials together were able to activate three out of four associations. Commercial 1 was picked run as main commercial for the campaign.
Neurensics has developed a 3600 framework to measure brand associations based on the proposition that they should evoke emotional preferences based on unconscious associations. Associations are measured using rapid implicit response tasks, which enable to model what associations contribute positively or negatively to brand value. Targeted campaigns can then be developed to strategically strengthen the right associations enabling a brand to move to a more unique position in the competition field. Before broadcasting, it can be verified if campaign materials are indeed able to activate the right associations.
Although neuromarketing research has seen significant innovations in the last years, we still need to address two major issues. Most metrics used are very low-level or have no predictive strength. Consider doing an IAT-EEG study on Coca-Cola. You find low alpha power and a strong association with the concept sugar. Would we need to boost the pairing of Coca-Cola with sugar? Should we increase alpha activity and how would we achieve this? Making translational models from associations to emotions and benchmark the predictive quality with AI are two crucial steps to move neuromarketing away from “blobology” and take it to the next level.
This article was originally published in the Neuromarketing Yearbook. Order your copy today!