- Jun 12, 2019
Marketing online with video messages is increasingly common and can be a powerful way to reach consumers. However, videos produced for purposes such as television advertising and social media messages, represent data that is “unstructured”. This means that it is information that was not produced with the purpose of being analyzed for research and development and, as a result, it’s difficult to directly measure the videos as messages of persuasion. The ever-increasing volume of this type of unstructured data means that potentially important information is available to link video messaging to marketing success if an appropriate method or tool could be used without the need to spend resources cleaning and preparing the videos for analysis. Ivey researcher, Shane Wang, and his colleagues demonstrate such a tool; convolutional neural networks (CNN). CNN is a machine learning algorithm that analyzes visual information without the need for costly and resource consuming data cleaning and data preparation. Testing CNN, the researchers examined the success likelihood of music projects on Kickstarter that featured video content. Success was measured simply as the project reaching its financial goal or not. The first variable, video variation, was measured at the level of the pixel differences between video frames. Video variation was found to have an increasing positive effect on success but also reached a point at which returns became marginal. Video content was the second variable and for this particular category of Kickstarter instruments and humans were used as the measure. Both measures were predictive of project success. For this variable, the inclusion of humans, who were often the music creators, was more indicative of successful outcome on Kickstarter, than was the inclusion of the five musical instruments included for content. Wang and colleagues suggest multiple avenues of further research using this video analysis tool.
Li, X., Shi, M., & Wang, X. (S.). Video mining: Measuring visual information using automatic methods. International Journal of Research in Marketing, available online.