M

The weight of each criterion can be interpreted as the "percent of the total weight accorded to that particular criterion."

A special caveat is in order. It is quite possible with this type of model to include a large number of criteria. It is not particularly difficult to develop scoring scales and weights, and the ease of gathering and processing the required information makes it tempting to include marginally relevant criteria along with the obviously important items. Resist this temptation! After the important factors have been weighted, there usually is little residual weight to be distributed among the remaining elements. The result is that the evaluation is simply insensitive to major differences in the scores on trivial criteria. A good rule of thumb is to discard elements with weights less than 0.02 or 0.03. (If elements are discarded, and if you wish Wi = 1, the weights must be rescaled to 1.0.) As with any linear model, the user should be aware that the elements in the model are assumed to be independent. This presents no particular problems for these scoring models because they are used to make estimates in a "steady state" system, and we are not concerned with transitions between states.

It is useful to note that if one uses a weighted scoring model to aid in project selection, the model can also serve as an aid to project improvement. For any given criterion, the difference between the criterion's score and the highest possible score on that criterion, multiplied by the weight of the criterion, is a measure of the potential improvement in the project score that would result were the project's performance on that criterion sufficiently improved. It may be that such improvement is not feasible, or is more costly than the improvement warrants. On the other hand, such an analysis of each project yields a valuable statement of the comparative benefits of project improvements. Viewing a project in this way is a type of sensitivity analysis. We examine the degree to which a project's score is sensitive to attempts to improve it—usually by adding resources. We will use sensitivity analysis several times in this book. It is a powerful managerial technique.

It is not particularly difficult to computerize a weighted scoring model by creating a template on Lotus 1-2-3 or one of the other standard computer spreadsheets. In Chapter 13, Section 13.3 we discuss an example of a computerized scoring model used for the project termination decision. The model is, in fact, a project selection model. The logic of using a "selection" model for the termination decision is straightforward: Given the time and resources required to take a project from its current state to completion, should we make the investment? A "Yes" answer to that question "selects" for funding the partially completed project from the set of all partially finished and not-yet-started projects.

Rather than using an example in which actual projects are selected for funding with a weighted factor scoring model (hereafter "scoring model") which would require tediously long descriptions of the projects, we can demonstrate the use of the model in a simple, common problem that many readers will have faced—the choice of an automobile for purchase. This problem is nicely suited to use of the scoring model because the purchaser is trying to satisfy multiple objectives in making the purchase and is typically faced with several different alternative cars from which to choose.

Our model must have the following elements:

1. A set of criteria on which to judge the value of any alternative;

2. A numeric estimate of the relative importance (i.e., the "weight") of each criterion in the set; and

3. Scales by which to measure or score the performance or contribution to value of each alternative on each criterion.

The criteria weights and measures of performance must be numeric in form, but this does not mean that they must be either "ob jective" or "quantitative." (If you find this confusing, look ahead in this chapter and read the subsection entitled "Comments on Measurement" in Section 2.5.) Criteria weights, obviously, are subjective by their nature, being an expression of what the decision maker thinks is important. The development of performance scales is more easily dealt with in the context of our example, and we will develop them shortly.

Assume that we have chosen the criteria and weights shown in Table A to be used in our evaluations." The weights represent the relative importance of the criteria measured on a 10-point scale. The numbers in parentheses show the proportion of the total weight carried by each criterion. (They add to only .99 due to rounding.) Raw weights work just as well for decision making as their percentage counterparts, but the latter are usually preferred because they are a constant reminder to the decision maker of the impact of each of the criteria.

* The criteria and weights were picked arbitrarily for this example. Because this is typically an individual or family decision, techniques like Delphi or successive comparisons are not required.

 Appearance