Improving Banner Ad Strategies Through Predictive Modeling

Abstract

In this study, we apply unique predictive modeling and data mining methods to identify visual and temporal factors that have significant impacts on both the effectiveness and pricing of Internet banner ads. An analysis of 18,956 online advertising records aims to identify the optimal banner advertising strategies for achieving different business metrics, including effective cost per activity (eCPA). Specifically, we find that banner ads with high visual complexity and attractive offers tend to draw users to participate in online activities, while voluntary banner ads with low visual complexity tend to draw user clicks. Further, banner ads with lower visual complexity tend to have lower costs. The size and shape of banner ads also play a key role as larger banner ads are more effective. Finally, we find that the third quarter of a year is the most important period for online advertising campaigns. Advertisers can use the findings from this study to create an effective and cost-efficient banner advertising strategy. Specifically, firms should use larger banner ads with features and offers, advertise at the end of the year, and use caution with banner ad animations as they can significantly increase costs.