What could go wrong? You make the parts, check them 100% with your gages and ship them to the customer. You know your gages are good because they’ve just been calibrated so all that’s left to do is collect the money. Until the phone call from your customer comes in advising that their gages have rejected the parts.

A check with your people confirms they really did inspect each part with your gages and they passed with flying colors. A check with your customer indicates they did the same with their gages, but they rejected the parts. Considering their inspection people are a shifty looking bunch, you have your doubts. So you ask the customer to send their gages so you can confirm their results using your staff. Unfortunately, your staff-using their gages-rejects the parts.

Both sets of gages have been calibrated recently and there were no red flags on either calibration report. I know this will sound pretty radical, but if ever there was a time to actually study the reports, this is it.

The first thing you have to check is the uncertainty claimed by each laboratory that calibrated the gages. If there’s a difference between them, it would suggest one set of data may be more trustworthy than the other. Of course, I’m assuming that both laboratories have uncertainty budgets to back up their claims. If one doesn’t, consider the calibration invalid and get someone else to calibrate the gage.

Next, create a picture using graph paper to see where each lab’s reading plus uncertainty plots out. If the uncertainties overlap, that’s probably where the true value or size of the gage is likely to be. If their uncertainties don’t come close to such a situation, it would seem that someone’s uncertainty budget may be lacking. Review them to see where they take you with respect to the tolerance on the feature they are being used to check.

Note that a reading that is right at the top product limit, for example, means that the gage could be over the top limit when measurement uncertainty is taken into account. To add more spice to your life, consider that the plug gage was probably not checked for roundness beyond ovality and could be out of round making it larger than what the calibration report shows. This scenario can be quite real when you consider that gage makers are allowed up to half their size tolerance for variations such as roundness or taper.

Situations such as those I’ve outlined here are not uncommon and are usually resolved by reviewing the calibrated gage sizes, uncertainties attached to that calibration and the product limits. Often you’ll find that while there may be a difference between potential gage sizes, it is small enough to be ignored. This is particularly so if the hole they are checking is clearance for a stove bolt holding a name plate on the component. Having written that, there are situations where people launch into great rants and spitting contests over such differences and refuse to accept the reality of measurement uncertainty.

If someone is not about to admit their gage may be outside limits from a functional point of view or believe your comments regarding uncertainty are simply your way of trying to bamboozle him into submission, there is another possibility. That possibility is that the calibration of both gages is not that good or you are fighting over something that is not worth the effort. This brings about the need for a “put up or shut up” test.

Simply put, you send the gages to the supreme authority on measurement-NIST-with the loser picking up the tab. It will only cost a few hundred dollars and a month or two of waiting, but it will settle most arguments when the numbers come from such a respected source. Of course, NIST will show their uncertainty on the reports they issue but it will only be a few millionths of an inch or a small part of a micron. I have found that such tests are rarely entered into because one or more of the parties is battling from a position that changes when it will cost them more if they’re wrong.

When you get down to what’s important on the grand scale of it all, what’s a few millionths between friends?