A novel bidding strategy based on dynamic targeting in real-time bidding market

Driven by big data analytics technology, real-time bidding (RTB) advertising has emerged as a mainstream paradigm in online advertising. Demand-side platform, acting as advertiser agents, develop bidding strategy for advertiser to match target audience and to make appropriate bid price in RTB market. Most bidding strategies target potential audience via cookie-based data analysis and market segmentation. However, due to the granularity choice dilemma, segmenting market in advance according to historical data may lead to targeting deviation. In addition, it cannot capture the changes of users' interests in time. A novel bidding strategy with dynamic targeting is proposed in this study. A user similarity model is constructed by calculating the distance between reaching user and typical target audience in a feature space. On this basis, bid price is determined. The profile of target audience is depicted based on prior information and then adjusted dynamically according to users' responses. Hence, targeting deviation caused by market segmentation which relies heavily on historical data is reduced. Additionally, we theoretically reveal a negative correlation between bid price and revenue along with a positive correlation between bid price and impression rate. Computational experimental results demonstrate the superiority of our strategy over the existing strategy in this regard.