What is variation?

Variation is the spread of a characteristic across parts made by the same process: a diameter, a flatness or a time per part. Turn fifty shafts on the same machine with the same program and you will not measure the same diameter fifty times. Variation is usually expressed as a standard deviation or as a range, the difference between the largest and smallest value. It is separate from location: a process can vary little and still sit off the middle of the tolerance.

What does variation depend on?

Variation has causes that act at the same time: wear of the cutting edge, differences in the stock, the temperature of machine and shop, clamping force and position, and the operator's hand. Part of it is random and small, part is systematic, such as a dimension that drifts over the life of the tool. Statistical process control separates common causes, which belong to the process, from special causes, which can be identified and removed. What you see is not the process alone either: the variation of the measuring instrument adds to it.

How does variation work in practice?

A tight tolerance needs a process whose variation fits comfortably inside it. Where it does not, nonconforming parts appear without anyone making a mistake. That can call for more inspection, earlier tool changes, a finishing pass or a different process, with time and cost effects that differ per feature and machine. Times vary too. An estimate usually works with the average over the batch, and because breakdowns only ever make times longer, that average tends to sit above the most common time.

What is variation often confused with?

Variation is not an outlier: variation is the band, the outlier a value outside it. Nor is it measurement uncertainty: uncertainty concerns the doubt attached to a measurement result, variation the differences between parts, although the instrument adds to the variation you observe. And variation is not process capability: capability sets the variation against the tolerance.