"Theompirical" Research of Service Systems
Avishai Mandelbaum
Technion
Wednesday, Feb 12, 2020
4:30 - 5:30 PM
Location: Shriram 262
Abstract:
I shall describe my personal research journey through service systems (e.g. hospitals, telephone and chat centers, banks, courts). I view these systems through MS/IE/OM lenses, often more specifically as a queueing scientist (e.g. "enjoying" congestion and flows), and sometimes using operational characteristics as surrogates for other performance indicators (e.g. clinical, psychological, financial). The goal of the research is to create principles and tools that support the above viewpoints; and the means for achieving this goal is the marriage of theory with data.
To be more concrete, I am modeling complex service systems as relatively simple (robust) processing networks. My theoretical framework is (asymptotic) queueing theory, specifically parsimonious fluid models and their diffusion refinements: queueing theory is ideally suitable for capturing the operational tradeoff that is at the core of any service, namely quality vs. efficiency (possibly augmented with fairness or profitability); and asymptotic analysis accommodates complex service characteristics that are otherwise mathematically intractable, for example transience, scale and scope. My data/empirical framework builds on an extensive data repository of service event-logs, at the level of the individual customer-server transactions (e.g. patient-physician or customer-agent).
Marrying theory with data, as I see it, will culminate in the creation of models directly from data - automatically and in real-time - and consequently the validation of the models' value againts actual service systems. (This is in contrast to still prevalent OR/MS/IE/OM practice, where models are too often too remote from data, and approximations are validated for merely accuracy relative to their originating models.) More specifically, data-based models - simulation, empirical, statistical and mathematical - will be created via process-mining of their primitives, structure and protocols. Simulation models will then serve as a validation ground for the other models, and all will be universally accessible for applications by researchers, students and in the longer‐run practitioners.
The above research agenda has been advanced, over 15 years or so, at the Technion SEE Laboratory (SEE = Service Enterprise Engineering); SEELab data will hence be used, throughout my lecture, to make it concrete.
Operations Research Colloquia: http://or.stanford.edu/oras_seminars.html