Programmers love to use benchmarks to figure out whether a program is running quickly enough. Or to find the bottlenecks in a program. Or, sometimes, just to figure out why users are experiencing slow behavior. But programmers who try to compare programs, or algorithms, or languages, or technologies, are often accused of making mistakes with their benchmarks. In this talk, Aysylu Greenberg looks at benchmarking, and points to ways in which our benchmarks may well be flawed — and how we can try to avoid such problems.