A B2B travel technology company intermediating car rental between suppliers and partner sites.
Millions of automated decisions are made on the client’s platform every day. These decisions include computing the price of an individual transport product or optimising the order of these products shown to customers. Because of the scale at which these decisions happen, small changes to them can have large aggregated effects; as a result, being able to make optimal decisions is critical to the business.
The client’s in-house data science team had already developed a number of tools to maximise the revenue while performing A/B testing of various decision-making algorithms. The client asked Faculty to develop a ‘virtual revenue management lab’ to test these tools.
To achieve this, we modelled a series of options that would cover the full range of agents acting on behalf of both customers and suppliers. The models were flexible to allow for the different preferences and priorities of particular agents. For example, agents representing customers might prioritise price and quality, while those representing suppliers might prioritise having as much of their fleet booked as possible at any one time.
The virtual lab is implemented as a Python package so that any member of the client’s technical team can install, use and improve it by making additions. It is already in use by the client’s data science team, and is giving recommendations on how to improve the existing algorithms.