Skip to content

Measure Before Optimizing

Human intuition about performance bottlenecks is systematically unreliable; measurement must precede optimization because you cannot improve what you haven't measured

Tension Addressed

Simplicity vs. Performance

Recognizable As

Profiling tools, flame graphs, instrumentation, metrics collection, A/B performance testing, baseline measurements before changes

Canonical Sources

Knuth, "Structured Programming with go to Statements" (1974); Gregg, "Systems Performance" (2020); Bentley, "Writing Efficient Programs" (1982)

Transfers To

Scientific method, quality improvement, operations research, any optimization domain

Knuth's famous observation that "premature optimization is the root of all evil" is often misquoted without context. The full insight is that optimizing without data leads to complex code in places that don't matter while missing the actual bottlenecks.