Audience preference prediction for commercials using fNIRS

• ¹ School of Biomedical Engineering, Science and Health Systems, Drexel University, United States
• ² Lebow College of Business, Drexel University, United States

Advertisers rely on self-reported measures for assessing and improving the advertisements. Methods such as surveys, focus groups, interviews are extensively used specifically for commercials targeting high profile events like the Super Bowl. Consistent with the Neuroergonomics approach, brain activities measured in response to viewing of natural stimuli could be used to assess audience preferences as a more direct measure (Ariely & Berns, 2010; Dmochowski et al., 2014; Plassmann, Venkatraman, Huettel, & Yoon, 2015; Venkatraman et al., 2014). Neuroimaging tools, such as functional magnetic resonance imaging (fMRI) and Electroencephalogram (EEG), have been utilized in initial neuromarketing studies, however various limiting factors such as high operational cost, restrictions on the participant during data collection, and the speed and ease of sensor setup limit further use of these in large scale deployment as well as in actual field conditions. Functional near infrared spectroscopy (fNIRS) is the youngest and still an emerging neuroimaging technique that utilizes near infrared light to measure oxygenation changes in outer cortex. Latest generation of optical brain imaging utilizes wearable and wireless sensor pads to enable measurement of brain activity in non-tethered and ambulatory settings (Ayaz et al., 2013). In this study, we recorded brain activity from a group of individuals viewing popular, previously-broadcast television advertisements played during the past year’s Super Bowl. The Super Bowl is the largest sporting event in US with more than 100 million viewers each year with well documented preferences from large audience. We utilized fNIRS to monitor anterior prefrontal cortex of 14 volunteers (mean age 27.6 years, 64% male). Participants viewed 30 super bowl commercials (30 second videos) of which fifteen were ranked highest (high-rated) and the other fifteen ranked lowest (low-rated) based on USA Today’s Ad Meter scores compiled from thousands of online viewers’ self-reported measures. In our protocol, each advertisement was followed by a set of self-reported measures to further capture the likeability and participant’s familiarity with the advertisement, intentions to purchase and recommend the products/services of the advertised brand and the level of excitement/feelings towards the advertisement. Preliminary results indicate that brain activation predicts the self-reported rating by broad audience. There was a consistent high cortical oxygenated hemoglobin concentration changes during the viewing episode of low-rated advertisement videos. Data acquisition for the study is still ongoing and further analyses will compare if ratings of the larger audience are related to the accuracy or brain activity than those of the individuals from whom the brain data is obtained. Figure 1. Average oxygenated-hemoglobin changes in anterior prefrontal cortex while watching the advertisements indicate consistent pattern across high and low rated videos. Figure 2. F-statistics map of significant difference implicate medial prefrontal cortex in both left and right hemisphere (FDR corrected q=0.005)