Research suggests that humans are biased, not-very-rational decision-makers. We believe we see things clearly when the evidence shows otherwise. Throw in a big dose of optimism, and it's easy to see how estimating software projects can be problematic. Our best hope is to construct diverse groups with varied viewpoints.

There's a simple rule for the minimum values testers should explore: “none, one, some”—or, how the software behaves if you send it nothing, one thing, or some set greater than one. It's not comprehensive, but it gives a good feel for how the feature works at the moment. Developers can also use this in unit testing.

When you test medical device software, you must be very careful. But when development wants to push a cadence of two weeks per sprint, every sprint, you’ve just got to keep up! Interpret the regulatory requirements not as a set of disabling constraints, but as a challenge to find the optimal route to navigate through.