Revenue management systems as symbiotic analytics systems: insights from a field study

Abstract

Revenue management is a complex operational planning process involving predictive and prescriptive analytics. As real-world implementations strongly rely on the joint outcomes from both algorithms and analysts, we consider the revenue management system as an example of symbiotic analytics systems. This paper presents insights from a field study observing a natural experiment in revenue management. As a firm updates its automated revenue management systems, it also updates the related processes and the corresponding organizational structure. We use this opportunity to examine the multilevel use of symbiotic analytics systems based in a field study and explore the implications for the design of future systems. Specifically, we identify two different perspectives on the revenue management process. In the functional view, jobs are organized sequentially with a high degree of system-oriented specialization. The process view organizes jobs in a parallel structure, differentiating two perspectives on demand. Depending on what view the firm implements, different structural fault lines turn the communication and training of analysts into keystones of the planning process. Furthermore, as we point out, even implementing more sophisticated algorithms and redesigning planning processes and organization do not seem to reduce the relevance of human analysts.

Decision biases in revenue management revisited: Dynamic decision‐making under stationary and nonstationary demand

“…Second, the opportunity to observe operational RM experts in situ, as described in Schütze et al (2020), is severely limited by company policy and labor councils. These seem to effectively prevent much empirical research on human analysts overruling analytics systems, with a few recent notable exemptions, such as Sun et al (2021).…”

“…In addition, the broad diversity of possible RM system and process designs and the complexity of the related models create a wide range of possible system variants. In the light of this challenge, empirical studies, for example, as described in Zeni (2003) or Schütze et al (2020), might offer a better way of considering RM in its full complexity.…”

“…In contrast to the classical definition of decision support systems (Kasper, 1996), RM systems do not support infrequent, managerial decisions but instead guide dynamic operational planning. Therefore, we view such systems as symbiotic analytics systems (Schütze et al., 2020). System designers need to consider the implications of biases for the human–algorithm interface of symbiotic analytics systems.…”

“…In other words, they let humans override both inputs and outcomes. This idea became clear to the authors in discussions and demonstrations with experts from RM departments and software suppliers and has been considered in some recent contributions on organizational RM aspects (compare Fürstenau et al, 2020;Schütze et al, 2020). Our research is mainly motivated by research-industry cooperation involving a major European airline that had implemented and funded RM research for several years at the time of the project.…”

Detecting outlying demand in multi-leg bookings for transportation networks

“…Furthermore, practical network RM often relies on manual forecast adjustments (Currie and Rowley, 2010;Schütze et al, 2020). However, previous research (Lawrence et al, 2006;De Baets and Harvey, 2020) has shown that the resulting judgemental forecasts can be biased and even superfluous.…”