Feedback-control based hierarchical multi-constraint ad campaign optimization

Online advertising is typically implemented via real-time bidding, and advertising campaigns are then defined as extremely high-dimensional optimization problems. Advertisers often define a campaign by an order consisting of multiple lines. Campaign delivery constraints may be imposed on the order as a whole and on each ad line. E.g., there may be budget and cost per click constraints on the order and on each line individually. Furthermore, the sum of line budgets may exceed the order budget, and the cost per click constraint on lines may differ. This leaves room for cross-line budget optimization; i.e., budget may be shifted across lines to maximize the advertising value without violating the constraints. This paper derives the optimal bidding mechanism for a large family of constrained optimization problems. It is shown how the optimal bidding strategy can be implemented as scalable non-cooperating agents on the order and the individual lines.