For my limited experience, and drawing only on memory of observation, rather than any recorded data, I'd break it down approximately as: * 1/3rd of interviewee

For my limited experience, and drawing only on memory of observation, rather than any recorded data, I'd break it down approximately as: * 1/3rd of interviewees freeze up and wouldn't be able to add two numbers together if asked. This is, to me, is mostly indistinguishable from, "have never coded in their life and lied successfully to get this far." * 1/3rd only partially freeze up and produce an incomplete solution and can't talk their way through what they did and why there's a missing piece * 1/3rd are able to get through FizzBuzz but with great frustration (they're usually upset with themselves for struggling so much). None of these failures are acceptable in an engineer with any amount of experience, even though they're not all strictly speaking about ability to work with data structures and algorithms in a vacuum. The thing is, software development is a team sport, and you have to come ready to play. If the goal was to find the person with the absolute best ability to take a set of static requirements and turn them into functioning code by themselves, this would not be a fantastic test. That is not, however, how nearly any software job actually works.