[ablob]

Sigh, I don't know why a blog-post against micro benchmarks should alter my opinion on this topic. There are more metrics than just "speed" regarding an interface. You also want it to be discoverable and visually distinct.

How someone interacts with your software is absolutely measurable and the results will vary by how a user is likely to use it in frequency and variety of function. Someone that needs to do something specific with your software every day will interact with it quite differently than someone that just hops onto it every now and then to do a different task each time.

All of this requires actual studies and observation of users over time. Micro benchmarks have no space there. Testing how fast a find and replace is is meaningless. In case of software for writing text you'd test a user actually writing prose, changing font sizes, title colors, and maybe replace a word over the file too. You would have commonly used functions mixed in with less commonly used functions over how the software is used under a specific use case. (For example, writing text, revising text, and polishing a graph representation are different use cases)

This is not easy, which probably why it's not done all too often, but it is also most definitely unlike a micro benchmark (which your link argues against).

All that being said, I don't know of any person strictly pitting mouse against keyboard when testing UI for possible improvements.