Enhancing Revenue Management System Performance through Human-Algorithm Interaction
Authors:
Ibrahim Mohammed Research Assistant Professor, School of Hotel and Tourism Management, The Hong Kong Polytechnic University https://www.linkedin.com/in/ibrahmoh/
Basak Denizci Guillet Professor of Revenue Management, Department of Tourism, Sport and Hotel Management, Griffith University www.linkedin.com/in/basakdg
Purpose
To provide a comprehensive guide on the effective use of Revenue Management Systems (RMSs) by understanding common biases in override decision-making and applying best practices. The authors extend their gratitude to all interviewees and survey participants for their valuable contributions, without which this research would not have been possible. We would also like to acknowledge the support from HSMAI Asia Pacific and ARMA, The Australian Revenue Management Association. We hope this report serves as a helpful resource when working with revenue management systems and performing overrides. Additionally, we trust it will support efforts to develop override-related metrics within your organization. For further information, please feel free to contact the authors.
Executive Summary
This report examines the role of cognitive biases— anchoring, overconfidence, and other decision-making biases—in shaping revenue management executives’ interactions with RMSs. These biases, if not managed carefully, can lead to override decisions that impact revenue outcomes. Through a combination of interviews with industry experts and a survey of revenue management professionals, this report identifies the cognitive biases in human-system interactions and outlines effective practices and guidelines for maximizing RMS performance. The report concludes with a detailed set of “10 Dos and 10 Don’ts” that serve as a comprehensive guide to effectively balancing human judgment and RMS recommendations.
Introduction
RMSs are vital tools in the hospitality industry, leveraging data and algorithms to optimize pricing and inventory management decisions. However, RMS recommendations are often overridden by managers based on their judgment or specific market insights not fully captured by the system. In practice, there are two types of overrides in this context, namely input and output overrides. Input override refers to adjusting the underlying variables or inputs in the forecasting model, such as modifying demand assumptions or correcting data inaccuracies.
Output override, on the other hand, involves altering the final recommendations produced by the system, such as changing the suggested room rates or inventory controls. Input overrides are generally preferable when time allows for recalibration, as they maintain the integrity of the RMS’s analytical process, resulting in more consistent, data-driven recommendations. Output overrides are better suited for immediate adjustments, where speed and flexibility are critical, though they should be applied selectively to avoid undermining the RMS’s predictive accuracy over time. While these overrides can add significant value when used correctly, they introduce potential risks, especially when driven by biases or subjective interpretations. Despite the widespread use of overrides across the industry, an objective metric for measuring the success rate of these adjustments remains largely undeveloped. In the absence of such a metric, measuring the true effectiveness of overrides becomes challenging, as does understanding the impact of manual interventions on long-term revenue performance. Recognizing and managing biases in override practices can play a key role in enhancing both RMS effectiveness and overall revenue outcomes.