Title: Estimating the effectiveness of AdWords geo-targeting
Authors: Iman Al-Hasani - Durham University (United Kingdom) [presenting]
Abstract: AdWords is an online advertising service, allows advertisers to place advertising. AdWords allows advertisers to track the performance of their ads by receiving detailed data such as website visits, number of clicks, sales and the revenue. These metrics help to measure the effectiveness of the advertising. In practice, however the actual effects of the ads campaign is difficult to quantify due to the noise in users behaviour. A comparison geo experiment approach has been used to measure the impact of the advertising . In this experiment a region of interest is partitioned into a set of geographical areas, called geos. These geos are then assigned randomly to serve the ads. The aim is to provide extended statistical methodology for geo experiment. It is expected to suggest improvements in generation the geos, construct a proper design of how to allocate these geo to treatment group and estimate the size of the effect of targeting within each geo. Adwords geo-targeting are linked to governmental areas using the shortest distance over the earth's surface. A simulation structure has been constructed to simulate the online searches and purchases occured at each target geo. Logistic regression is used to estimate the probability of search and conditionally the probability of purchase. The size of the effect of the ad campaign is computed for different campaign design structures. The estimation are computed based on covariates attributable to individuals, regions and ad campaign.