In a prior lifes endeavor I was involved in all manner of lubrication application. Whenever we did fuel economy measures, for whatever product tested, we found that doing fleet testing, e.g. taxi or police fleets was more meaningful. We also did the test blind, i.e. so that the driver didn't know whether he had the "old" stuff, or the "new" stuff. We did this because we found individual references were generally unreliable because people would often unconsciously change their driving habits/patterns/style that would distort the data. We would control the variables such as Denny mentioned so they wouldn't be a modifier of the data. Usually the more vehicles involved in similar operating circumstances, and the more time (to "even out" seasonal effects), the more reliable our results.

All that being said, synthetic oils offered a very small improvement by themselves. Driving style had as much or more impact.