Analysis of Google Ads data for Mitre 10 New Zealand Limited

Abstract

Sponsored search auctions are commonplace these days and search engine marketing companies such as Google make billions of dollars in revenue every year. Advertisers spend a significant portion of their marketing budget on sponsored search and this is only expected to increase in the future. Mitre 10 New Zealand Ltd. is one such advertiser. This research thesis discusses the experiments conducted on Mitre 10’s sponsored search auction data and the subsequent findings. When a user enters a search term, search engine companies such as Google matches it against a pool of keywords. It then invites advertisers who are interested in the keywords to participate in an auction. It is conjectured that Google solves a complex optimisation problem every time an auction is conducted. Each advertiser would have specified a bid–the maximum amount they are willing to pay if their ads are clicked. Advertisers would be invited to participate in auctions until their daily budgets are exhausted. The existing body of literature on sponsored search can be broadly classified into Search Engine Company focused, User-focused, and Advertisers focused. Search Engine Company focused literature proposes various models that help search engine companies optimise their objectives such as increasing their revenue, improving user experience, or increasing advertisers’ return on investment. User-focused literature proposes models that explain user behaviour such as clicking an ad under different conditions. Advertiser focused literature proposes models that help advertisers plan their budgets, bids and keywords better. Experiments were conducted by changing the bids, bidding strategies, budget, keywords and geographical scope of an ad. Key Performance Indicators for Mitre 10 such as the number of auctions won (Impressions), number of ads clicked (Clicks), the probability of an ad getting clicked (Click-Thru-Rate), and the cost paid per click (Cost-Per-Click) are measured. The data thus gathered is used to test various hypotheses and insights are drawn from them. Linear regression models are constructed to understand the impact of the bid and budget change on various KPIs. The thesis concludes with a summary of the findings and proposes future work in this area.