Don't forget that some tests are open ended

A lot of testing literature talks about inputs and expected results. And that's all well and good, but you can't forget that there are multiple reasons to test and multiple ways to test, and that for some of them you can't accurately predict expected results. I find that a lot of early tests that happen in a project are focused on expected results. "We built the system to do X, does it do it?" However, over time the nature of the testing changes. It becomes more "what if" focused. Examples include, "How many users can we support on our current production configuration?" or "What happens to our users when that batch process runs?" or "What kind of errors might we see in the logs on a typical day in production?"