Dynamic capacity allocation for group bookings in live entertainment

• Published in: European journal of operational research Vol. 287; no. 3; pp. 975 - 988
• Main Authors: Maclean, K.D.S., Ødegaard, F.
• Format: Journal Article
• Language: English
• Published: Elsevier B.V 16.12.2020
• Subjects: Customer choice; Dynamic capacity allocation; Group bookings; Live entertainment; Revenue management; Revenue management; Live entertainment; Customer choice; Group bookings; Dynamic capacity allocation
• ISSN: 0377-2217; 1872-6860
• DOI: 10.1016/j.ejor.2020.02.017

•A dynamic model for small group request in live entertainment.•Three customer choice based modeling approaches for seat selection.•Optimal capacity allocation algorithms for live entertainment.•Empirical analysis of sporting event to show-case model parameters.•Revenue enhancement of 2–3 by incorporating analytics in seat allocation. A persistent problem within live entertainment is lost revenue due to unsold seats. One reason behind this problem is that venues generally permit customers, of varying group size, to freely choose seats, and thus causing a sub-optimal seating allocation with sparsely stranded single seats. Due to the experiential attribute of live entertainment, ticket requests are predominantly groups wishing sets of contiguous seats. Consequently, the sparse single seats remain unsold. To solve this operational problem we analyze a capacity based revenue management control problem that explicitly accounts for group size and customer choice. We formulate the problem as a discrete-time Markov Decision Process with the objective to maximize total expected profit. Each period, and for a given arriving group size, the manager decides which price-differentiated segments to make available. Given the offered segments, customers select seats from a particular segment or choose not to purchase any. We discuss three selection models and provide algorithms to determine the optimal solution for each. Motivated by ad hoc provisions observed in practice and due to the curse of dimensionality we provide and analyze via simulation a heuristic. Finally, based on transactional sales data from a large annual North American sporting event we showcase how the model parameters can empirically be estimated.