Not 85%, but 85th percentile, apparently.

https://metrocount.com/downloads/flyers/Speed_analysis_1.pdf

I'm not a traffic engineer so I'd love an explanation, because on the first look, it seems absurd to apply this measure to determining speed limits, as a) 85th percentile will keep creeping up with each iteration of such speed limit adjustment, and b) the whole rationale seems to depend on "wisdom of the crowds", i.e. trusting greedy optimization to find the right balance in the system that needs to be optimized globally.