Requirements can be ambiguous. In fact, any English-language (or any other language) statement is probably ambiguous in some way, as well as being subjectively understood. This is obviously not the way that we should write requirements. As a consequence, in addition to the natural-language description of the requirement, we write a fit criterion. The fit criterion is a quantification of the requirement, which, as far as possible, uses numbers (numbers have no ambiguity) instead of words to express the requirement.
The fit criterion is a measurement of the requirement that enables the testers to determine, without recourse to subjective judgments, whether the delivered solution meets, or fits, the requirement.
The fit criterion is a measurement of the requirement.
It is crucial to accurate requirements specification that each requirement carries a fit criterion. The first question to ask is, "Does the requirement have a correctly defined fit criterion?" Any requirements that do not must be considered to be incomplete, as a large degree of uncertainty remains about the requirement. It is really this simple: If you can't measure it, you don't understand it. The next question for the fit criterion is, "Can it be used as input to designing acceptance tests?" You may need help from the testing people to answer this question. The fit criterion is not the design for the test, but indicates what needs to be tested so as to ensure the delivered solution complies with the original requirement. At this stage you should also consider whether there is a cost-effective (within the project's constraints) way of testing a solution to this requirement.
Does each requirement have a fit criterion that can be used to test whether a solution meets the requirement? The fit criterion must also meet the purpose of the project. We have discussed how the requirement must conform to the project purpose, so it makes sense that the measurement of the requirement must likewise conform to this purpose. While the fit criterion uses numbers to express the requirement, the numbers themselves must not be subjective, but must be based on evidence. For example:
The question to ask about this fit criterion is, "Where did the 30 minutes come from?" Is it simply the whim of a stakeholder or the requirements analyst? Or is it based on evidence that a learning curve longer than 30 minutes means the users will become discouraged and give up? It is, of course, useful if the requirement writer has included a reference to the evidence in the Supporting Materials component of the requirement. The fit criterion uses a scale of measurement. For example, if the requirement states the product is to be fast, then the scale of measurement for the fit criterion is probably time. Check whether the scale of measurement is appropriate for the requirement, and whether acceptance tests run using that scale will, in fact, show whether the product meets the requirement. The fit criterion measurement may include some business tolerances. For example, if the fit criterion specifies that a number of people have to perform certain tasks within certain time limits, then the business tolerance allows a small percentage of the test panel to fail. This tolerance exists because you cannot expect 100 percent of humans to be able to do everything. Similarly, products that handle material or deal with imperfect outside world artifacts might be allowed a certain margin of error, or business tolerance. Consider whether the business tolerance is based on appropriate evidence. |