Marcin Borkowski (mbork) has posted an instructive article on benchmarking Elisp code. He had a theory on the run time of alternative implementations for some code he’d written and rather than wave his arms, he did what we all tell ourselves we’ll do but often don’t: he benchmarked the run time of the two implementations.
He used the benchmark library to do this. It’s worth reading his post just to remember (or familiarize yourself with) the details of benchmark. In any event, he confirmed his suspicions about the implementations but discovered that he was right for the wrong reasons.
Take a look at his post to see the details and discover what he learned about working with strings.