> What about having several use cases in mind, and give the scores for each of those?

Or assign one score according to the worst case scenario.