Conditional Gage R&R: Turn 10–30% %GRR Into Controls

"It came in at 18%. The customer SQE said we can live with it, but only if we tighten things up." Anyone who has run a Gage R&R on a difficult bore has heard some version of this. The study sits in the middle band of the AIAG criteria, the PPAP deadline is close, and a better gauge is a capital request that won’t be approved this quarter. So the team writes down a few conditions: a shorter calibration interval, a setting-master check before each shift, the gauge stays in cell 4, a repeat study after the next calibration. The PPAP goes in. Element 8 gets accepted.

Eleven months later the gauge is on its original 12-month interval, it has been borrowed by cell 2 twice, and nobody can find the email with the conditions in it. An 18% result was acceptable in February, under those conditions. In January of the next year, without them, it is just a weak measurement system.

This article is about the part after the study: turning a conditional acceptance into controls that hold without anyone having to remember them.

What "conditional" actually means in the AIAG MSA manual

The AIAG MSA reference manual (4th edition, still the current edition) gives three bands for %GRR. Below 10% is generally acceptable. Above 30% is unacceptable, and the measurement system needs improvement. The 10–30% band may be acceptable depending on the application: how important the characteristic is, what the gauge costs, what repair or improvement would cost. The manual also says the decision should be approved by the customer. Alongside %GRR, the number of distinct categories should be 5 or more.

Two things follow from this.

First, conditional acceptance is a decision with a rationale, not a pass. The rationale is always something like "18% is tolerable because we will do X, Y and Z." If X, Y and Z stop happening, the rationale is gone, even though the study report hasn’t changed.

Second, IATF 16949:2016 clause 7.1.5.1.1 requires statistical studies for every type of inspection, measurement and test equipment system identified in the control plan, using methods and acceptance criteria from the customer’s reference manuals. When an auditor samples that bore gauge, they will read the MSA record and then the conditions attached to it, and then check whether you actually did them. Findings in this area usually aren’t "your %GRR was 18%." They’re "your MSA record says six-month calibration and the gauge history shows twelve."

A worked example: an 18% bore gauge on a valve body

Here is a realistic case. A dial bore gauge measures a Ø22 mm spool bore on an aluminium valve body. The tolerance is ±0.010 mm, 0.020 mm in total. The Gage R&R, crossed design, 10 parts × 3 appraisers × 3 trials, analysed in the team’s statistics package, gives:

• %GRR (to tolerance): 18%
• %GRR (to study variation): 24%
• ndc: 6
• Repeatability (EV) dominates; reproducibility (AV) is small

The customer approves use, subject to four written conditions:

1. Calibration interval cut from 12 months to 6 months, with as-found data reviewed at each calibration.
2. An extended calibration procedure for this unit: as-found and as-left readings at three points across the working range, not one.
3. The gauge is used only on the valve-body line in cell 4, by the appraisers who took part in the study.
4. The R&R is repeated after the next calibration, or immediately if the gauge is adjusted or repaired.

Each of these is reasonable. None of them happens on its own.

Mapping each condition to a control

The test for each condition: if the person who negotiated it leaves the company tomorrow, does it still happen?

Condition 1: the shorter interval

A note in the PPAP file doesn’t enforce anything. What does is setting the interval on that serial number in the equipment register, so the next calibration due date is calculated from the six-month interval, not the default for the gauge type. In PrüfDesk the due date is auto-calculated from the interval on the instrument record, so shortening the interval for one unit moves its due date, its reminders and its position on the overdue list. The other bore gauges of the same type stay on 12 months.

Condition 2: the extended procedure

This is where reusable calibration procedure templates earn their keep. The bore-gauge template covers every unit of the type. For this one serial, the template is overridden at unit level to add the extra measurement points and as-found/as-left recording. The technician gets the step-by-step procedure for this unit without having to know the history behind it, and the pass/fail record and certificate are stored against the serial. When the as-found data from each six-month calibration is on file, you have evidence of whether the gauge is drifting. That is what the customer wanted when they asked for it.

Condition 3: restricted use

"Only in cell 4" becomes a permanent allocation to the cell, not a verbal agreement. If someone checks the gauge out to cell 2, the possession audit trail records who had it, when, and where. That doesn’t stop every misuse, but it turns an invisible deviation into one you can see and act on. It also tells you whether the conditions of use still match the conditions of the study, which the MSA manual expects.

Condition 4: repeat the study after calibration

This condition depends on condition 1. With the due date driven by the shortened interval, the calibration event is the trigger for scheduling the repeat R&R. PrüfDesk won’t run the study; your statistics tool does that. What PrüfDesk gives you is the calibration record that shows the gauge was in calibration, with traceability, on the day the repeat study was run.

The backstop: overdue blocking

PrüfDesk lets each tenant choose how overdue instruments are handled: a hard block or a soft warning. For a site that carries conditionally accepted gauges, a hard block is the setting that makes condition 1 stick. A gauge past its six-month due date can’t be checked out, so it can’t be used for a production measurement or a study. A soft warning only records that someone was warned.

The objection: "Calibration doesn’t fix repeatability"

That’s correct, and it is worth being honest about. In this example the variation is mostly EV. A shorter calibration interval does nothing to the within-appraiser spread. Calibration addresses bias and detects drift; it does not improve %GRR.

But the conditions were never meant to lower the 18%. They exist so the system stays the one that was studied and judged tolerable: the same gauge, in calibration, with its bias checked more often, used where and by whom the study assumed. An 18% system that has also drifted, or moved to a line with a tighter tolerance, is no longer the system the customer approved. The controls protect the approval. The improvement plan, whether that’s a better gauge, a fixture or an air gauge, is a separate action and should have its own date.

Checklist for every conditionally accepted gauge

• Conditions written, customer-approved, and filed with the MSA record
• Interval shortened on the serial, not the gauge type
• Unit-level procedure override with the extra as-found/as-left points
• Permanent allocation to the line or cell named in the conditions
• Tenant overdue mode set to hard block
• Repeat-study date tied to the next calibration event
• Improvement action with an owner and a date, if one was agreed

Where this leaves you

A 10–30% result is a legitimate outcome under the AIAG criteria, but only while its conditions are met. An auditor sampling that gauge in Q4 will test those conditions, not the number. If you’d like to see how interval overrides, unit-level procedures and overdue blocking look for one of your own conditionally accepted gauges, PrüfDesk can walk through it with your data.