An Advertiser Centered Approach to Improve Sponsored Search Effectiveness
- Author(s): Vattikonda, Bhanu Chandra
- et al.
Sponsored search is a form of advertising where advertisers pay a search engine to show their ads on the search engine results page. The ads, also known as sponsored results, are chosen and presented to the user in response to a user query alongside organic search results. Sponsored search holds the promise of allowing advertisers to precisely target their ads to the large number of users of a search engine. The rise in use of search engines and the opportunity they provide to target ads using fine- grained criteria has led to a 20% annual growth in sponsored search revenues over the last decade. The targeting criteria chosen by an advertiser for their ads allow a search engine to deliver the ads to the right users. At the same time, it also puts the onus on the advertiser to identify the right ad targeting criteria. In this dissertation, we take a two-pronged approach to improve the effectiveness of sponsored search in delivering value to advertisers and improve the quality of results shown to users. First, we improve the ability of a search engine to interpret the targeting criteria specified by the advertiser. As part of the targeting criteria advertisers submit ad keywords which specify the user queries for which they would like to advertise. We leverage the search engine itself to interpret an ad keyword by submitting the ad keyword as an independent query. Using the search results of the ad keyword associated with an ad we determine if the ad is suitable for the original user query. We then analyze the effectiveness of different targeting strategies followed by advertisers. We develop a simple metric called net acquisition benefit (NAB) that admits comparisons between the efficacy of different ad targeting strategies. Using this metric, we conduct the first large-scale measurement of different targeting strategies used by advertisers--- measured in terms of incremental conversion gains. Considering data from a month in early 2015, we employ NAB to identify cases where these targeting strategies are justified