Harnessing the Strengths of Computational Advertising and Experimental Methods to Enhance Research Design

Abstract

Computational methods such as those used in computational advertising research are often considered at odds with experimental methods. This divergent view of these two paths has frequently led to researchers examining these methodologies in isolation. In this article, we take the view that each path has value and that the strengths of one can inform the other. To this end, we provide a guide to conducting research that focuses on four strengths inherent to experimental research—the principle of control, random assignment of treatments, the exploration of underlying mechanisms, and transparency—that can provide lessons to those using computational advertising data. Similarly, we focus on three strengths inherent to computational advertising research—the collection of behavioral data, the realistic nature in which the data are obtained, and the ability to study dynamic changes over time—that can provide lessons on how to enhance experimental research. In doing so, we assist researchers in strengthening their study designs regardless of their preferred methodology.