Quantum computing for rostering and A/B testing

In this presentation, we considered two problems: 1) Producing efficient personnel rosters that satisfy various constraints and preferences, and 2) Quantifying the impact of ad campaigns. The rostering problem is notoriously difficult, with constraints imposed by the flight schedule, the work force including different types of contracts, and agreements such as collective labor agreements. The goal is to find a roster that is as good as possible while meeting as many constraints as possible. The difficulty of the ad campaign impact is that ideally one wishes to compare revenue with and without the ad campaign. However, only one of the two typically is available. Predicting the impact of the revenue without the ad campaign is possible based on matching time series from other sources. Choosing the most fitting time series is hard.

For the rostering problem, a hybrid quantum annealing approach was used. The problem was formulated as a mathematical optimization problem. The hybrid quantum annealer broke the formulation in smaller pieces, solved each of the smaller subproblems and then reconstructed the overall solution. The work for A/B testing is still ongoing. So far, quantum computing seems to help in better feature selection (which leads to better time series) and quantum Boltzmann machines that better process the selected time series.

Improved rosters that better satisfy the operational requirements are in general cheaper. Additionally, rosters that satisfy the personnel needs tend to result in happier workforce, contributing to a vital organization. Better prediction of the impact of ad campaigns helps in better spending the advertising budget, resulting in a stronger financial position.