WHAT ARE METRICS?

Continuing with the hierarchy view, metrics are the information dimension of measurement. A metric comprises one or more measures. Consider the example of the measure of the width of a 30-inch desk. That is certainly a property of the desk; it has quantification and units. However, this measure takes on a distinct meaning if the desk is being moved to another area and the narrowest part of the move is a 32-inch doorway. Now, the measure has meaning, or as we are defining it, metric. Usually metrics combine more than one measure, such as dollars per month or more complex formulas such as earned value analysis (EVA).

Metrics obviously inherit the precision, accuracy, and validity of the measures they comprise, with the metrics’ same properties limited to the lowest of these attributes for the associated measures. Focus and coverage are two additional characteristics of metrics that must be considered in evolving effective measurement.

Measurement can focus on process improvement, progress, benchmarking, project management, technology management, etc. The selected focus defines the structure of the measurement activities. For example, if the purpose of measurement is competitive benchmarking, there is little need to define boundaries internal to the overall activity that is being benchmarked. If the focus of measurement is process improvement, then boundaries need to be defined carefully to ensure that processes are compared and analyzed separately and that all data collected take these boundaries into consideration. Clearly, the focus of measurement is a primary determining factor of the cost and speed with which measurement can be implemented.

Coverage expresses how much of what is being measured is actually measured. In situations where all possible measures are taken, the results are called population measures or full coverage. Measures of less than the full population are samples; coverage is then expressed as the sample percentage of the total population. Coverage is also a primary determinant in the overall cost of measurement.

Metrics can also be characterized into four scales:

1. Nominal

2. Ordinal

3. Interval

4. Ratio.

Nominal measures define categories and relationships. Examples of nominal measures are: mode (most frequent occurrence) of a sample, name of a category, and characterization as “temperature.” Nominal measures for hair color, for example, would be: brown, blonde, red, black, white/gray, and none of the above (N/A).

Ordinal measures are those given to preserve the ranking of categories. For example, a statement that blonde hair is the second most common natural hair color in the company gives no indication of how many people are in the company or how many blondes there are. It simply states that blonde hair is second.

Interval measures look at relationships between measured objects but preserve the order and magnitude of the relationship. For example, it is 20 degrees warmer today than it was yesterday. An interval measure of the hair colors in the company is that there are 12 fewer blondes than there are people with brown hair but 6 more than there are people with red hair.

Ratio measures preserve the order of the relationship, the magnitude of the interval, and the magnitude of the measure itself. For example, if it is 60 degrees Fahrenheit today and it was 40 degrees Fahrenheit yesterday, the nominal measure is temperature, the ordinal measure is warmer, the interval measure is 20 degrees Fahrenheit warmer, and the ratio measure is 60 degrees Fahrenheit today compared with 40 degrees Fahrenheit yesterday. The thermometer reading is absolute in that there is a zero reference on the Fahrenheit scale and an absolute zero reference related to the Fahrenheit scale.