Many around me (including my sponsors) assumed uncritically that assessing human performance was a simple and straightforward matter of counting up successes and errors starting with the benchmark of standard procedures. But watching operators deal with complex emergency situations revealed many challenging forms of cognitive work where expertise and failure were intermingled. Digging behind the veil of procedure-following and everyone’s narrow focus on major failures, I was able to begin to see how practitioners create safety normally despite conflicting goals, pressures, and dilemmas.