How much is enough? Determine the optimal frequency of internet display advertising (IDA)

"Half the money I spend on advertising is wasted; the trouble is I don't know which half."

The words of John Wanamaker (1838-1922), a pioneer American marketer and advertiser, still hold true almost 100 years since they were spoken in the context of the ubiquitous banner advertising also termed as internet display advertising (IDA). Whilst IDA has become a major component of the internet advertising industry with an estimated size of US$ 56.5 billion (ZenithOptimedia, 2014) and a compounded annual growth rate of 21.5% (ZenithOptimedia, 2014), our understanding of how IDA works is fairly nascent. Despite thousands of advertisers utilising this platform to target billions of audiences worldwide, there is still a lack of clarity of how IDA really works. Specifically there is a gap in understanding the level of IDA impressions that are required to drive a specific goal. The blistering pace at which internet advertising has grown and evolved has contributed in a large part to a research gap that has stubbornly persisted. On one hand internet's constant evolution has spawned user journeys that are a complex tangle of advertising impressions, search clicks and website visits. On the other hand internet's exponential growth has led to constantly expanding data sets both in terms of volume and number of variables. This has led to an ongoing need to constantly update the findings and methods in order to understand how IDA works. Developments in clickstream technology that track complex user behaviour on the internet have provided an opportunity to further research in the area of determining optimal frequency level in IDA to drive marketing goals.

This research seeks to assist researchers and practitioners in their quest to improve their understanding of IDA and internet advertising as a whole by uncovering a deeper understanding of the impact IDA impression frequency plays in driving marketing goals, especially online purchases. Further, the study analyses the moderating influence of consumers' characteristics and spacing of IDA impressions after controlling for a comprehensive set of factors related to media and seasonality. Lastly, the study uniquely utilises an easy to use but effective method that overcomes issues that plague clickstream data characterised by very large volume and low conversion rates. The clickstream data for this study was obtained from the ad-server log files from a large advertising agency based in Singapore. The use of such a data-set for this study is a significant step towards linking the world of information systems and marketing by making effective use of big data in developing optimal IDA impression frequency guidelines that will ultimately contribute towards improving the ROI of internet display advertising (IDA).