Hi Jayfass, from what you have said the part data has been rearranged for the study. If we select 10 parts and perform repeated measurements on each with multiple operators, then go on to compute a standard deviation representing repeatability it's valid to compare this against the specification width to compute a consumption of tolerance but only if the measurement process is consistent.

However, if we want the result of the study to represent what we see in the production process and not just a measurement repeatability study between operators then we need to select parts in a way which best represents the variation in the production process. For a torque study we have aspects like variation in nuts and bolts, hole, lubrication, washers etc etc to rationally take into consideration

As long as we understand what we are trying accomplish and what the results will portray against what the customer requires.

For your studies:

**Front** - measurement process is not consistent, particularly affecting parts 6 and 7 across 2 operators, this requires looking at before making any characterisations about the measurement process because the standard deviation is clearly not stable and therefore not representative.

**Rear** - Measurement is consistent, no detectible operator bias (reproducibility), the real consumption of repeatability at a minimum of 96% confidence for the samples chosen is 12%.

Its true that the average torque values for the parts in both studies appear to climb but as the data has been re arranged into Low Med and High that's hardly surprising. This type of study doesn't care about the patterns of averages or ranges as long as they are not detectibly different.

I hope this helps