When I'm feeling obsessive-compulsive over distance and/or speed, the wired (or wireless) cyclocomputer with a wheel sensor is my go-to measurement device. I've got a nice downhill ridge near home I can coast down, and compare the computer's measured distance with the mile markers. If I'm within 1%, it's fine; more than 2%, time to adjust the calibration. I've gotten it within 0.5% some months, and then the tire gets worn and/or replaced, and the next time it's up to 1% match.

A few key features of the above. First, I'm coasting downhill, so there's minimal side-to-side pedal tracking to affect the distance measured. Second, because it's a U.S. highway, I trust the mile markers to be accurate. (Don't try this in Kansas or Missouri -- I think they sell mile markers at Walmart in those states, and farmers pick out a pretty green sign to put up near their farms without regards to location or numbers) Finally, curves are pretty wide, even on the ridge.

GPS devices sometimes match the cyclocomputer pretty closely, as long as the comparison is on a straight, treeless road with no large buildings nearby. On my more typical rides (curvy, trees, hillsides blocking some satellite signals), the GPS distance is off. It's still good if I need help mapping or navigating, but I don't trust the distance the GPS measures. Ergo, I figure the instantaneous speed measurement is in the ballpark, but it's more for entertainment than precise measuring.