It is argued that most usability engineering is qualitative in nature: you observe how users use your product - what they have difficulty doing, what they like, and what they hate - then you refine your design accordingly. Given the severe time pressure in most development projects, it's enough to know that something is good or bad. Rarely is it necessary or possible to take the time to find out how good or bad, since if it's good you want to leave it alone and if it is bad you want to change it (if you have time, that is). There are some cases, though, where you do want numbers. The author does not recommend usability metrics for all projects; qualitative methods are usually sufficient to ensure substantial usability improvements. That would mean you won't know bow much better your new design is, but as long as it is not worse than the old one, going with the new one is clearly the better decision. However, organizations that do extensive software development and care about their engineering maturity should collect metrics for at least some of their projects.