Service process control: a method to compare dynamic robustness of alternative service processes

https://doi.org/10.1016/S0959-1524(02)00097-5Get rights and content

Abstract

This paper contributes to the service engineering field by developing a procedure to compare the dynamic robustness of alternative service delivery processes. The procedure adapts an existing method for comparing the dynamic robustness of chemical process control systems. It has four steps: (a) characterisation of each type of uncertainty that affects the service plant; (b) Monte Carlo computer simulation of the process flowchart to determine how the service process responds to different combinations of uncertainties; (c) calculation of a quality index for each combination of uncertainties; and (d) plotting of a frequency distribution of these indices. The tighter the resulting distribution, the more robust the system. The procedure can be used to evaluate the effects of design and control changes in the service delivery process. To demonstrate the procedure, a modification to a hospital patient-treatment process is evaluated.

Introduction

Consider the following situation: the manager of a hospital emergency ward is considering modifying the patient-treatment process to meet the requirements of patients, staff and government more effectively. It is important to evaluate how consistently the modified process copes with uncertainties (e.g. unpredictable patient arrival rates, ailments, and treatment times) before a decision can be made to implement it permanently. However, it is impractical to test the modified process 'live’ because it may endanger lives if some uncertainties affect it adversely. This paper helps service managers evaluate modified processes by providing an alternative to ‘live’ tests.

The service manager's problem is really how to assess the robustness of the patient-treatment process. The stochastic uncertainties facing the process (called heterogeneity in service literature) parallel the uncertainties arising in poorly known disturbances and process dynamics in chemical plants. As well, a service process should perform consistently in the face of uncertainties caused by human factors in service delivery [1]. This parallels the requirement of robustness in chemical process control systems. Thus, methods that assess the robustness of chemical process control systems would be useful in assessing the robustness of service delivery systems.

However, firstly the similarities between chemical and services processes should be described to demonstrate how chemical robustness methods can be adapted to services. This has been done by comparing a simple service process, such as a visit to the local doctor, to a simple chemical process, such as two chemical reactors in series [2] and recasting the management of service processes as a feedback control system. Fig. 1 shows the two processes. The service equivalent of the chemical plant are the activities and flows representing the processing of a customer's request. The ‘raw material’ in a service is a customer with an unmet need, the ‘chemical reactions’ are a customer's interactions with the service provider, and the ‘final product’ is a customer whose needs have been met. ‘Outputs’ are process characteristics (e.g. waiting time, accuracy, workpace) that meet requirements of the plant's stakeholders, such as customers, staff and managers. ‘Setpoints’ are the ideal values for these outputs. ‘Manipulated variables’ are variables (e.g. staffing levels) within the service manager's control. The ‘controller’ is the service manager, who can operate in ‘manual’ or ‘automatic’ (e.g. using experience or computerised rostering systems to determine staffing levels). ‘Disturbances’ are external variables (e.g. the nature of customer requests) that are beyond the service manager's control. Service processes even have similar requirements for stability since customers value a service's reliability (i.e. its ability to perform dependably and accurately [1]), and performance since customers value a service's responsiveness (i.e. how quickly it meets their needs [1]).

However, many robustness evaluation methods developed for chemical process control systems cannot be applied directly to service processes because of one major difference between these two types of processes. Service processes are discrete systems that are modelled more easily by flowcharts showing the sequence and duration of activities, whereas chemical processes are generally continuous systems modelled by differential equations. Therefore, robustness-assessment methods that rely on analytical solutions (e.g. unstructured and structured singular value analysis [3], [4], [5], [6]) are not applicable to service processes. Hence, a simulation-based method is necessary.

One recent simulation-based method, developed specifically to compare the dynamic robustness of alternative chemical plant-control systems [7], can be adapted to service processes. This method characterises uncertainties by probability distributions and uses stochastic simulation to determine the effects of various combinations of uncertainties on plant behaviour. (Other studies use a similar approach for robust optimisation [8], [9], [10], [11], [12]). This approach has two main advantages: (a) it can be applied directly to discrete systems such as service plants without model simplification; and (b) it presents results in economic terms (for plants where cost data is available) using a graphic form that is more meaningful to non-control engineers than abstract mathematical quantities such as structured singular values. This paper will describe how such an approach can be adapted to service processes.

By developing a robustness-assessment method for service plants, this paper contributes to the field of service engineering (i.e. the application of engineering principles to the design and operation of service processes), and answers a call by leading service scholars for more research in this area [13]. It augments the other chemical process control technique that has been adapted to services, namely the off-line stochastic model-based optimising controller [2]. Further, this paper adds rigor to the service design process, which has been inadequate [14], [15]; and will help service firms to improve their pre-development activities, which have been inefficient compared to those of manufacturing firms [16], [17].

The following sections describe the proposed robustness-assessment procedure in detail, and demonstrate its use in evaluating alternative designs for a patient-treatment process in a hospital emergency ward.

Section snippets

The robustness-testing procedure

The robustness-testing procedure, developed for chemical process control systems [7], has four steps:

  • 1.

    Characterise each type of uncertainty that affects the plant.

  • 2.

    Conduct simulations to show how the plant responds to different combinations of uncertainties.

  • 3.

    Calculate an economic index of control quality for each combination of uncertainties.

  • 4.

    Plot a frequency distribution of the indices.

Each step will be described firstly as applied to chemical plant-control systems, then as it can be applied to

Existing process

The emergency ward is available to anyone with a medical problem. Except for patients with life-threatening ailments, who are classified as Category 1 patients and given immediate treatment on arrival, other patients must present themselves at the registration desk. A registration clerk fills out an admission form for each patient, with details of his/her medical history. Next, patients wait for a triage nurse to categorise their ailments according to how life-threatening they are. There are

Conclusion

The aim of this paper was to provide a method of testing the dynamic robustness of service delivery processes. It did this by adapting a robustness testing procedure developed for chemical process control systems. The procedure involved four steps:

  • 1.

    Characterise each type of uncertainty that affects the plant.

  • 2.

    Conduct simulations to show how the plant responds to different combinations of uncertainties.

  • 3.

    Calculate an index of control quality, based on Taguchi principles, for each combination of

Acknowledgements

I thank Michael Jacobson, two anonymous reviewers, and the editor, Professor Peter Lee, for helpful comments on earlier versions of this paper; and Debbie Malone, Graydon Davison, Yogesh Deshpande, Liwan Liyanage and Ross Chapman for their involvement in flowcharting the patient-treatment process.

References (21)

There are more references available in the full text version of this article.

Cited by (5)

View full text