A customer rejects a lot of shafts for an oversize diameter. Your inspector measured the same parts last week and passed them. Before anyone blames the machine or the operator, there is a simpler question to ask: can your gauge actually tell a good part from a bad one? A Gauge R&R study is how you answer that with numbers instead of arguments.

What a Gauge R&R study measures

Every reading has two parts: the true size of the part, and the error added by the measurement system. That system is the gauge, the person, the method, the fixture and the environment together.

Gauge R&R splits that measurement error into two pieces:

  • Repeatability is the variation when the same operator measures the same part with the same gauge again and again. Minitab's guidance describes it as "the variability when the gage is used to measure the same part by the same operator." Think of it as equipment variation.
  • Reproducibility is the variation from different operators measuring the same part. In simple terms, it is operator-to-operator variation: different feel on a micrometer, different way of seating a bore gauge, different reading of a dial.

Add the two and you get Total Gauge R&R. The rest of the variation in your data is part-to-part variation, which is the real difference between parts that you actually want to see.

How to run a basic study on the shop floor

According to Minitab's Assistant white paper, typical guidelines recommend 10 parts, 2 or 3 operators, and 2 or 3 replicates. A practical routine looks like this:

  1. Pick 10 parts that cover the real process spread. Do not take ten parts from one good batch. Include parts near the lower and upper end of what the process makes.
  2. Number the parts on a hidden face so operators cannot remember earlier readings.
  3. Choose operators who normally do this check, not only the most experienced inspector.
  4. Measure in random order. Minitab's worked example uses 10 parts, 3 operators and 3 trials each, 90 readings in all, taken in random order.
  5. Use the same method every time: same location, temperature and clamping.
  6. Analyse with the ANOVA method in your statistical software or a validated spreadsheet template.

The same white paper also notes that 10 parts do not give a precise estimate of part-to-part variation. If your sample does not represent typical process variation, use a historical estimate of process variation or more parts.

How to read the results

Most reports show three numbers. Here is how the commonly used AIAG guidelines read, as summarised by Minitab's interpretation guide:

  • %Study Variation (Total Gauge R&R): under 10% is acceptable; 10% to 30% may be acceptable depending on the application; over 30% is unacceptable and the system should be improved.
  • %Contribution: under 1% is acceptable; 1% to 9% may be acceptable depending on the application; over 9% is unacceptable.
  • Number of distinct categories (ndc): the number of groups your system can reliably separate. Minitab notes that the AIAG MSA manual recommends 5 or more categories for an acceptable system.

In Minitab's example, Total Gauge R&R was 27.86% of study variation with 4 distinct categories, so that system sits in the "depends on the application" zone and falls short on ndc.

A gauge that cannot see the difference between parts will pass bad parts and reject good ones, and no amount of process control will fix that.

Also check %Tolerance when you use the gauge to accept or reject parts against a drawing limit. A gauge can look fine against process spread and still be too coarse for a tight tolerance band.

What to do when the study fails

The split between repeatability and reproducibility tells you where to look.

  • High repeatability, low reproducibility: the gauge itself is the problem. Check resolution (a 0.01 mm instrument on a 0.02 mm tolerance band will struggle), wear, play in the fixture, and cleanliness of the part and anvils.
  • High reproducibility: the operators differ. Write a clear work instruction: where to measure, how many points, how much force, how long to let the part cool after machining. Train, then repeat the study.
  • Both high: consider a better gauge type, for example an air gauge or a dedicated fixture with a dial instead of a hand-held micrometer.

For example (illustrative): a turned bore of 25H7 has a tolerance band of 21 microns. Checking it with a vernier caliper will almost certainly fail a %Tolerance check, while a calibrated bore gauge set on a ring master has a much better chance. Decide the gauge at the planning stage, not after the first rejection.

Plan your gauges before you quote

Many MSA problems start at the RFQ stage. A tight positional tolerance or a small precise bore needs the right gauge, fixture and time for a study, and all of that is cost. List each critical characteristic with its gauge before you accept the order.

How SpecReview helps

SpecReview reads your 2D drawing PDF or STEP model and builds a quality plan and a layout inspection plan, listing the characteristics to measure so you can decide gauges early. Its DFM and rejection risk list also flags tight tolerances with a short reason why each matters, before you quote.