UC Berkeley

This dissertation contains three essays that examine different aspects of online marketing activities, the ability of marketers to measure the effectiveness of such activities, and the design of experiments to aid in this measurement.

Chapter 2 examines the impact of search engine optimization (SEO) on the competition between advertisers for organic and sponsored search results. The results show that a positive level of search engine optimization may improve the search engine's ranking quality and thus the satisfaction of its visitors. In the absence of sponsored links, the organic ranking is improved by SEO if and only if the quality provided by a website is sufficiently positively correlated with its valuation for consumers. In the presence of sponsored links, the results are accentuated and hold regardless of the correlation.

Chapter 3 examines the attribution problem faced by advertisers utilizing multiple advertising channels. In these campaigns advertisers predominantly compensate publishers based on effort (CPM) or performance (CPA) and a process known as Last-Touch attribution. Using an analytical model of an online campaign we show that CPA schemes cause moral-hazard while existence of a baseline conversion rate by consumers may create adverse selection. The analysis identifies two strategies publishers may use in equilibrium - free-riding on other publishers and exploitation of the baseline conversion rate of consumers.

Our results show that when no attribution is being used CPM compensation is more beneficial to the advertiser than CPA payment as a result of free-riding on other's efforts. When an attribution process is added to the campaign, it creates a contest between the publishers and as a result has potential to improve the advertiser's profits when no baseline exists. Specifically, we show that last-touch attribution can be beneficial for CPA campaigns when the process is not too accurate or when advertising exhibits concavity in its effects on consumers. As the process breaks down for lower noise, however, we develop an attribution method based on the Shapley value that can be beneficial under flexible campaign specifications. To resolve adverse selection created by the baseline we propose that the advertiser will require publishers to run an experiment as proof of effectiveness.

Chapter 4 discusses the type of experiments an advertiser can run online and their required sample sizes. We identify several shortcomings of the current prevailing experimental design that may result in longer experiments due to overestimation of the required sample sizes.

We discuss the use of sequential analysis in online experiments and the different goals of the experiments to make experiments more efficient. Using these techniques we show that a significant lowering of required sample sizes is achievable online.