The features included in a prioritization matrix are the discriminated union of features we deem to be particularly important, plus all of the tool features gathered in the interviews. The structure of this matrix is very similar to the feature comparison matrix described in the 308 document here:

http://users.csc.calpoly.edu/~gfisher/classes/308/handouts/feature-matrix.html

The rows in the prioritization matrix are elements of a taxonomic feature list, with precisely the same structure as in the comparison matrix. Each column in the prioritization matrix corresponds to a requirements interview, with one or more interview attendees.

Each matrix cell has an integer "desirability value" from -2 to 2, defined as follows:

Value	Meaning
-2	Really don't want the feature
-1	Would rather not have the feature
0	Don't care about the feature
1	Would like to have the feature
2	Really want the feature

Values in the rightmost matrix column are the sum of all the column values in a given row. These values represent a weighted desirability value for each feature.

The desirability values are determined by reading the interview transcripts and listening to the audio. The project team member who attended the interview will play a key role, being able to remember the tone of interviewee responses, in terms of how strongly certain opinions were expressed.

Determining desirability values is clearly a subjective exercise. Those who attended the interview may recall language like "we must have this feature", "we would love this feature", or "this feature is very important to us". These are indications of a desirability value of 2. In contrast, there may be interview language like "we hate this feature in the tool we use", or "there simply must be a better way to do this". These are indications of a desirability value of -2.

A desirability 0 value can easily be determined if a feature was not mentioned at all during the course of the interview, by either the interviewers or interviewees. Values of -1 and 1 are determined subjectively, based on the lukewarmedness of the interviewees' language.

The prioritization of features is accomplished by sorting the rows by the weighted desirability total.

Some Additional Discussion Points (These Ideas Need to be Refined)

• Look carefully at rows that have both -2 and 2 values. Use some subjective judgment to determine in any given case whether a feature should be in or out with a 0 value that comes from a sum of -2 and 2.

• Pick the features with the top n values, for n = the number of people we have to work on the scenarios.

• As we see fit, we can increase the values of features that we think are critically important, even when those features have client values of "pure" 0, where "pure" means all or nearly all column values are 0.