Fair tests and repeats
Change one variable at a time and repeat the trial, so any difference in results can only come from what you changed.
A worked example
Three rolls of the same car down the same ramp give 45, 49 and 50 cm. The mean is (45 + 49 + 50) / 3 = 48 cm, which softens the effect of one odd trial.
The mistake people actually make
If two people get different results, compare the methods before deciding who is right. Record measurements as you take them, because writing them up later from memory loses detail.