RAIL YIELD MANAGEMENT. TRENITALIA CASE

Abstract

The Yield Management System (YMS) described in this thesis has been developed by a joint group from IBM and Trenitalia, that I have joined since its inception. It has been implemented gradually to most trains `Le Frecce' at Trenitalia, main Italian and 3rd European railway undertaking, with 10 Million passengers and more than 260 High Speed trains offered daily, on average, in the first three months of 2018. The operating YMS aims at maximising revenues through a two-stage stochastic optimization model which forecasts the unconstrained demand, optimizes the capacity allocations per Origin & Destination (O&D) and fare cluster, sets the protection levels using a nesting technique, develops the constrained forecasts and simulates the results. During these years many and continuous improvements to the system have been designed, tested and deployed. Among the others: (i) a proportional correction to potential demand forecast, which is here a fundamental part of the overall optimization; (ii) the development of a methodological framework for assessing the YMS performance over time, consisting of a set of Key Performance Indicators, a Monitoring module developed from post-departure computation of the Revenue Opportunity; (iii) the design, run and analysis of a live test of a new prototype compared to the incumbent algorithm. They are subjects of this thesis, too. Since 2005, the YMS has forecasted and optimized approximately 4 Million model instances: nearly 120 Billion decisions on combinations of fares and O&Ds, per each train, class and departure date. This automatic optimization tool has provided a powerful support to the Revenue Management team, with a high degree of productivity and solid results, even in a period of major changes in the mobility landscape, with the beginning of open track competition in Italian High Speed rail services.