Or, more generally, for ubiquitous names, weigh the desiderata and come up with something good. For nonce names, it doesn’t matter. In your example, I’d maybe go with “wage” as something short/pronounceable/greppable/etc for pay_rate.

If you’re just computing weekly pay in one place then the self-documenting “weekly_pay” is fine, as is the other extreme of, say, p = hours_worked * wage. The opaqueness of “p” is fine if you’re just using it on the next line and never again.

I’m wondering if this system can be distilled down into a tool to measure these criteria objectively and score a name in realtime. Somewhat like a password strength indicator (eg. http://www.passwordmeter.com/) for names. I’d love to see and use such a tool.

I think Evocativity would be the toughest to measure, with verbability a distant second. Some of the easier criteria to systematic assess are brevity, greppability (eg. against some common corpus) and googability (how many exact phrase matches does google have).