HPL-2012-234Remixing the Marketing Mix: Accounting for Dynamic Cross-Channel Spillover Effects
Jamal, Zainab; Wang, Jianqiang C.
Abstract: In today's marketing landscape, firms are using multiple online channels (e.g., search, display) to reach their customers with their marketing messages. In terms of spending, the two major channels are display advertising (also called `banner ads') and paid search advertising (sponsored links on search engines like Google) (Bucklin and Sismeiro, 2009). A third channel, shopbots or comparative search engine, offers customers the ability to compare the same product at different online sites and click on the site from which they want to purchase it -- the site that offers the best value. These sites have become popular as consumers' demand for comparison shopping has grown (Wan and Peng 2010). The popularity of these sites is expected to grow with `show-rooming' among customers where they examine products in offline retail stores and then to purchase them online at the best price available. In this study, we look at shopbots (comparative search engines) from the website's perspective where the shopbot is another marketing channel (like paid search or display advertising) where the website advertises and pays the shopbot owners on a per click basis. We examine shopbots and their effectiveness in generating online orders -- not done to our knowledge in other prior research. Specifically, we examine HP's e-commerce site and compare the effectiveness of three online marketing channels -- search, display and shopbots (called comparative search engines, CSE) in garnering online orders at this site. Further, the level of spending in these channels is affected by how well the channels perform in bringing in customers and purchase revenues. The full value of a customer's purchase during a website visit is typically attributed to the referring marketing channel for that visit. The part played by other channels earlier in the shopping process (prior visits for research et al) is largely ignored, resulting in a non-optimal marketing mix allocation with some of the channels undervalued while others are overvalued. In this study, we estimate the correct value that should be attributed to an online marketing channel in the presence of dynamic cross- channel spillover effects as well as endogeneity. We estimate a Vector Autoregressive Model (VAR) as well as a Dynamic Linear Model (DLM) and compare them with a benchmark Autoregressive Distributed Lag Model (ARDL). We find significant direct and indirect effects of spending in the different channels. We also find interesting asymmetries of their impact.
External Posting Date: November 21, 2012 [Abstract Only]. Approved for External Publication - External Copyright Consideration
Internal Posting Date: November 21, 2012 [Fulltext]