A few guidelines should be considered when constructing the list of metrics:
• Metrics should be complete. Ideally each customer need would correspond to a single metric, and the value of that metric would correlate perfectly with satisfaction of that need. In practice, several metrics may be necessary to completely reflect a single customer need.
• Metrics should be dependent, not independent, variables. This guideline is avariant of the what -not-how
principle introduced in Chapter 5 . As do customer needs, specifications also indicate what the product must do, but not h ow the
specifica tions will be achieved. Designers use many types of variables in product development; some are dependent, such as the mass of the fork, and some are in dependent, suchas the material used for the fork. In other words, designers cannot control mass directly because it arises from other independent decisions the designers will make, such as dimensions and materials choices. Metrics specify the overall performance of a product and should therefore be the dependent variables (i.e., the performance measures or output variables) in the design problem. By using dependent variables for the specifications, designers are left with the freedom to achieve the specifications using the best approach possible.
o Metrics should be practical. It does not serve the team to devise a metric for a bicycle suspension that can only be measured by a scientific laboratory at a cost of $100,000. Ideally, metrics will be directly observable or analyzable properties of the product that can be easily evaluated by the team.
o Some needs cannot easily be translated into quantifiable metrics. The need that the suspension instills pride may be quite critical to success in the fashion-consciou mountain bike market, but how can pride be quantified? In these cases, the team simply repeats the need statement as a specification and notes that the metric is subjective and would be evaluated by a panel of customers. (We indicate this by entering "Subj.'" in the units column.)
The metrics should include the popular criteria for comparison in the marketplace. Many customers in various markets buy products based on independently published evaluations. Such evaluations are found, for example, in Popular Science, Consumer Reports, on various Internet sites, or, in our case, in Bicycling and Mountain Bike magazines. lf the team knows that its product will be evaluated by the trade media and knows what the evaluation criteria will be, then it should include metrics corresponding to these criteria. /llfountain Bike magazine uses a test machine called the Monster, which measures the vertical acceleration (in g's) of the handlebars as a bicycle equipped with the fork runs over a block 50 millimeters talL For this reason. the team included "maximum value from the Monster" as a metric. If the team cannot find a relationship between the criteria used by the media and the customer needs it has identified, then it should ensure that a need has not been overlooked and/or should work with the media to revise the criteria. In a few cases, the team may conclude that high performance in the media
evaluations is in itself a customer need and choose to include a metric used by the media that has little intrinsic technical merit.
In addition to denoting the needs related to each metric , Exhibit 6-4 contains the units of measurement and an importance rating for each metric . The units of measurement are most commonly conventional engineering units such as kilograms and seconds. However, some metrics will not lend themselves to numerical values. The need that the suspension "works with fenders" is best translated into a specification listing the models of fenders with which the fork is compatible. In this case, the value of the metric is actually a list of fenders rather than a number. For the metric involving the standard safety test, the value is pass/faiL (We indicate these two cases by entering "List" and "Binary" in the units column.)
The importance rating of a metric is derived from the importance ratings of the needs it reflects. For cases in which a metric maps directly to a single need, the importance rating of the need becomes the importance rating of the metric. For cases in which a metric is related to more than one need, the importance of the metric is de termined by considering the importances of the needs to which it relates and the nature of these relationships. We believe that there are enough subtleties in this process that importance weightings can best be determined through discussion among the team members, rather than through a formal algorithm. When there are relati vely few speci fications and establishing the relative importance of these specifications is critically important, conjoint analysis may be usefuL Conjoint analysis is described briefly later in this chapter and publications explaining the technique are referenced at the end of the chapter.