I work for a driveshaft manufacturer and we are measuring joint articulation torque on driveshaft universal joints. The tolerance is 0.3Nm to 0.7Nm so already its a torque measurement with a very small window for variability. What we are seeing is that the measurement is climbing from each operator. For example, operator 1 measurements look like .52, .58, .61, so my range is looking high but the next operator is having roughly the same measurements (.53, .57, .62) so each trial across the operators is consistent but because our measurement is climbing, its making our range look big and I believe its what is causing us to fail. I have been running a nested ANOVA and getting failing results. Am I running the wrong study for this test? Any ideas?