Revenue Optimization in Real-Time Bidding Based Advertising for Mobile Devices
COM2 Level 2
MR3, COM2-02-26
Abstract:
In the new era of online advertising, Real-Time Bidding (RTB) based advertising is becoming the new norm through which advertisers manage their advertising campaigns. Unlike traditional online advertising practices, RTB maps impressions of advertisements on mobile applications or websites to a particular advertiser through a real-time auction. Understanding how RTB differs from other advertising techniques is important as it has enormous potential to increase both the advertiser's and the publisher's revenues. In the RTB ecosystem, mobile app advertising has become the novel revenue generation channel for publishers through user and context targeted advertisements. The RTB ecosystem consists of two key entities - a publisher that places ads on end users' devices and an advertiser that runs the advertising campaign via a Demand Side Platform (DSP). In this dissertation, we focus on optimizing advertising campaign operations both from the perspectives of publishers and DSPs.
The primary task of the DSP is to represent the advertiser in the bidding process while attaining their target. The advertiser's target is defined by the available budget for the ad campaign, target audience, expected number of clicks/conversions, and the duration of the ad campaign. In the bidding process, the DSP needs to decide the bid prices to win the relevant bid requests that can achieve the advertiser's target. But this becomes challenging due to the impartiality in the data and the dynamism in the RTB ecosystem. The proposed novel bidding algorithm, "Auto pricing strategy", utilizes dynamic programming to cope with these challenges and to outperform state of the art approaches.
A subtask of a DSP's bidding process is to evaluate the audience to whom the best advertisements will be delivered such that a higher Click Through Rate (CTR) is attained. State of the art approaches estimate CTR based on the static historical data of incoming bid requests and click events. Existing approaches to improving CTR have overlooked the inherent dynamism in the RTB ecosystem and as a result, have not considered advertiser target audience selection leading to poor overall CTR during a campaign. To mitigate the above issue, we propose an automated campaign optimization strategy that delivers a higher return to the advertiser.
Besides the DSP's and the advertiser's involvement in the RTB ecosystem, this dissertation looks at the publisher's involvement as well. We consider the unique characteristics of mobile advertisements and the limitations of existing approaches in order to propose a solution at the app instance level to boost the publisher's return. The proposed solution provides a fully-fledged mechanism to select ad networks at the app instance level based on advertisement and ad network attributes. Using such attributes along with the mobile app user's behaviour, we estimate the effectiveness of ad networks, the effectiveness of advertisements and click return rates to determine the optimal ad network which is expected to produce the highest return to the publisher.