In tackling problems such as those presented in Section 1.2. one must necessarily follow the eight classical steps of a systematic analysis which are summarized in Figure 1.1.

As an illustration of utilizing these steps, let us consider Example 6 in Section 1.2, "Redesigning a City's Ambulance Services." Here a law has been passed mandating upgraded emergency medical skills for ambulance drivers and attendants. Private ambulance companies that are currently providing service are likely to go out of business unless something is done. A study group has been appointed by the mayor to analyze the problem. Let us briefly consider each of the eight steps, as they might evolve in practice:

1. Define the problem. With private ambulance companies likely to go out of business, the problem as stated to the study group is to determine the economic, operational, and service-related ramifications of alternative proposals for the provision of emergency ambulance service. Broadened to include the mayor and his or her staff, the problem is to select and implement a (possibly) new form of ambulance service.

2. Identify the objectives. Often, objectives are stated broadly; for instance: " The objective is to deliver effective and efficient emergency ambulance service equitably to all citizens throughout the city, within given fiscal constraints."

3. Specify performance measures. To analyze the various alternatives, we must have a systematic procedure for evaluating how well each one accomplishes the objective(s) stated in Step 2. Performance measures allow us to undertake the quantitative part of our study. A performance measure is a quantity attributable to system operation; it reflects the quality of system performance, or, at least, it quantifies some aspect of system performance. A good performance measure is understandable to both citizens and agency personnel, is relatively stable under some set of operating conditions, is dependent .on operating policy, is readily measurable, and is not easily subvertible (i.e., it is a true measure of system quality or at least service quantity). For some relatively straightforward urban services, such as door-to-door pickup and delivery services, valid performance measures are easy to identify. For others, however, one must rely on surrogate measures that represent an intermediate reading of service quality; this is particularly true of services for which we do not now know "production functions" that relate service characteristics to achievement of fundamental objectives. Examples are the relationships (or lack of known relationships) among police and crime, transportation services and economic growth, and-in our case here-ambulance deployment and mortality and morbidity. Thus, we may have to settle for such surrogate measures as response times to various types of ambulance calls for various neighborhoods, workloads or utilization rates of the ambulances, costs of training and retaining personnel, and other system costs. (As argued in Chapter 8, there are many important nonquantifiable features of a systematic analysis that require consideration in addition to performance measures.)

4. Identify the alternative courses of action. Here the basic alternative courses of action are prespecified:
a. Incorporate ambulance service into police department operations.
b. Have a separate city-sponsored ambulance fleet.
c. Subsidize the existing companies or a merged version of those companies.

Of course, each of these can be implemented in numerous different ways, thus implying that there are, in fact, many alternatives that we should consider. A reasonable strategy would be to compare the best option under each of (a), (b), and (c) in order to select the best basic strategy.

5. Analyze the alternatives to understand the consequences of each. At this step the methods of Chapters 2-7 are most appropriate and necessary for examining the operational consequences of alternative courses of action. As one example, we could employ our analysis to examine the implications of "equal service provision" for each of the three basic options. Equal service could imply for each option the achievement of, say, a 5-minute average citywide ambulance response time, with no neighborhood incurring an average response time greater than 8 minutes. We would then construct mathematical models of each of the three systems proposed-one augmenting the current police patrol force, the second representing an independently operating ambulance fleet, and the third representing the current companies or a merged version of those companies-and analyze each of the models, while maintaining as a constraint the equalservice provision. The geometrical probability methods of Chapter 3 would probably be useful in depicting geographical interactions, and the spatial queueing models of Chapter 5 would be necessary for analysis of system congestion. One might also require the network methods of Chapter 6 to study the point-to-point transportation characteristics of the city, and perhaps even the simulation techniques of Chapter 7, especially if the proposed operations are too complicated to yield a mathematically tractable model.

6. Compare the consequences and select an alternative. This is simply an extension of Step 5. If service levels are held equal for each option, then selection will probably be based on cost and on nonquantifiable issues such as political feasibility, likely implementation time, and long-term impact of the decision on other services.

7. Present the results and conclusions. The study group must present its findings to the mayor and his or her staff. Usually, an oral presentation highlighting the main features of the analysis is more effective than a long formal report (although thorough written documentation is obviously necessary, as well). In the operations research profession, there is a school of thought which argues that the analysts should not select an option in Step 6 but should present the characteristics and consequences of each to the decision maker in Step 7, who, in turn, will make the selection. Another school of thought argues that the analysts probably know more about the problem than anyone else is likely to, and thus-provided that the agreed-upon performance measures are adequate-are in the best position to recommend an alternative. In many cases, a hybrid strategy seems most appropriate, in which a constructive dialogue is established between the analysts and the decision makers, and the selection is eventually made jointly.

8. Implement and evaluate. As argued in Chapter 8, this crucial step is often the most difficult. Especially in urban services, rarely are all the constraints and other relevant features of a problem visible before attempted implementation. One should anticipate a period of "shakedown," "debugging," or "fine tuning" during the initial stages of any implementation program. Information obtained during this phase will often "feed back" to one of the earlier seven steps in the analysis, making the entire analysis process iterative in nature. Rarely is the analysis process strictly linear or sequential. This is why we have indicated possible feedback paths in Figure 1. 1. For instance, under any of the three basic ambulance options, once the spatial and temporal deployment of ambulances is made known, various neighborhood groups might raise objections because of perceived or actual undercoverage of their area by the proposed ambulance service.

If the complaints are found to be valid, the analysts may have to go back to Step 4 and redefine the alternative deployments that are feasible. Problems could also arise in labor-management negotiations, in purchasing equipment, in training, or in a host of other cases. Even if visible problems do not occur, it is necessary to monitor and evaluate the implementation process to assure that the system is functioning as planned and designed.

Much of the work in Steps 3-6 will involve working with models, idealized mathematical representations of the urban service systems of interest. A primary purpose of Chapters 2-7 is to develop skills in formulating such models. It is important to emphasize at this early point that the type and complexity of models must be adjusted to the types of questions that will be addressed during the eight steps outlined above. A given urban service system may need to be represented by two entirely dissimilar models in the process of answering two entirely dissimilar sets of questions about that system. In addition, just like the entire eight-step procedure, the formulation of a model only rarely turns out to be a "single-pass" sequential process. It is often necessary to revise a mathematical model several times in the course of an urban operations research study before that model is tailored to the particular needs of that study.

Many of the people who will be constructing models of urban service systems will also have responsibility for implementing them or the policy results of their use in one or more cities. Thus, although most of the methods of this book pertain to Steps 3-6, it is important to understand the larger framework in which a public-sector analysis is carried out. Our brief treatment here is sufficient at this time to provide the perspective necessary for Chapters 2-7. A fuller treatment is given in Chapter 8.