Advertisement Allocation and Mechanism Design in Native Stream Advertising.

We study optimal advertisement allocation in native stream advertising, which exhibits notable fundamental differences from traditional web advertising. Our model highlights the influence of distance between consecutively projected ads on user engagement, which implies interesting negative externalities between ads: when there are fewer intervening posts between two ads, there is higher ad fatigue and lower user engagement. We study the problem of advertisement selection and placement in a stream so as to maximize social welfare. We fully characterize the computational complexity of the problem by demonstrating that it is tightly connected to a special form of interval scheduling. Next, we prove that it is strictly NP-hard (SNP-hard) but can be efficiently approximated to within 1/2 of the optimum. We complement this result by studying the mechanism design variant of the problem and develop a 1/2-approximation truthful-in-expectation mechanism. We also consider mechanisms that guarantee a stronger deterministic form of truthfulness. We believe that these results lay a valuable theoretical foundation for further research in the field.