Overbooking with bounded loss

We study a classical problem in revenue management: quantity-based single-resource management with no-shows. In this problem, firm observes sequence of T customers requesting service. Each arrival is drawn independently from known distribution k different types, and the needs to decide irrevocably whether accept or reject requests an online fashion. The has capacity resources B, wants maximize its profit. accepted service request yields type-dependent probability requiring resource once all arrivals have occurred (or, be no-show). If number that require at end horizon greater than pay fixed compensation for each it cannot fulfill. With clairvoyant, knows ahead time, as benchmark, we provide algorithm uniform additive loss bound, i.e., expected independent B T. This improves upon prior works achieving Ω(√B+T) guarantees.