PPAP Layout vs Shop Gage: Hold Rules When Dimensional 6 Disagrees

The hold starts on the floor, not in the meeting. PPAP Layout vs Shop Gage is the kind of event quality teams treat as a paperwork delay when it is actually a reconstruction problem. The first hour decides whether you still have a lot, a load, a cook, or a board you can defend.

Write the time, the asset ID, the last passing check, and the name of the person who called the hold. If those four facts are missing, every later explanation becomes a story. Keep the product, the sample, or the part under the same identity it had when the signal fired. Do not relabel to make the warehouse happier.

What the record has to show

In automotive work, the record is the product for the duration of the hold. That means the electronic log, the paper traveler, the chart, and the photograph of the setup all have to point at the same serial, lot, or cook number. If one of them points elsewhere, stop adding opinions and fix the identity first.

List the last qualified state. For equipment that is the last calibration, the last IQ OQ PQ fragment that still applies, and the last change that touched the recipe or the program. For people it is the last observed competency for that exact method, not a generic annual training slide.

Photograph the as-found condition before anyone 'fixes' it. A cleaned probe, a redressed tip, or a reseated logger destroys the only evidence you had. If production insists on recovering the line, take the photos first and move the evidence under chain of custody.

How to reconstruct without averaging

Two numbers that disagree are not a reason to average. They are a reason to reconstruct sampling, timing, and the instrument. Ask who pulled the sample, which vessel or nest it came from, how long it sat, and which method version was open on the bench. Write those answers in the hold record before anyone proposes a mean.

If a second observer is required, they use the same light, the same magnification, and the same acceptance definition. They do not get a coaching session on the first observer's call. Independent means independent. Record both calls even when they match, because matching calls with no method still fail an audit.

Pull genealogy one level back and one level forward. Supplier lot, in-process split, and downstream units that already left the cell. If downstream units already shipped, the hold becomes a field action discussion. Do not wait for the weekly quality meeting to notice that.

Hold criteria that survive an audit

Write the hold criteria in the language of the specification, not in the language of the shift. 'Looks off' is not a criterion. 'Nugget diameter below 4.2 mm on coupon 3 of 5' is. 'Chart come-up exceeded scheduled process by 4 minutes' is. 'Viability 72 percent on flask, 91 percent on bag, same passage' is.

Name the owner of the hold. Name the owner of the investigation. They can be the same person on a small site. They cannot be 'quality' as a department. If the owner goes home, the next person on the roster inherits the same four facts from the first hour.

Set a clock for the first documented decision: scrap, rework, continue under deviation, or request more data. Open clocks without a decision date become silent scrap. Silent scrap becomes a 483 when the inspector asks why the lot sat for eleven days with no disposition.

What usually goes wrong

Teams reopen the line because the customer is waiting. That is a commercial fact, not a quality decision. If you restart, you restart under a documented rationale that names residual risk. If you cannot name residual risk, you are guessing.

Teams also mix containment with correction. Sorting the remaining units is containment. Changing the recipe is correction. Both may be needed. They are not the same signature. Keep them on separate lines of the record so the CAPA later has something true to attach to.

A third failure is using last year's protocol as if the machine, the method, and the people were frozen. If the last change control on that asset is newer than the protocol, the protocol is a draft in practice. Say so in the hold record instead of pretending the old IQ still covers the new probe.

Evidence pack

• As-found photos and the asset ID plate
• Last calibration or verification printout
• Sample or coupon identity and chain of custody
• Recipe, program, or process revision in force at the time of the signal
• Names and times of both observers if a second read was required
• Downstream units already packed or shipped

Disposition language

Disposition language should be boring. 'Lot 24B17 remains on hold pending integrity retest under protocol VF-17 after controlled rewet at 20 C.' That sentence can be defended. 'Looks fine, running' cannot.

If you release, say what residual risk remains and who accepted it. If you scrap, say which identity was destroyed and how you will keep a retain. If you rework, say the rework method, the yield expectation, and the inspection that will prove the rework did not hide the original defect.

Training the next shift

The next shift will inherit this hold. Leave them a one-page as-found, not a verbal briefing. Include the four facts from the first hour, the photos, and the open questions. If the only transfer is a Slack message, you have already lost the investigation.

This pattern holds across automotive sites because the physics change and the record rules do not. Identity, as-found, independent second check, genealogy, and a dated decision. Miss one and the hold becomes a story. Keep all five and you can still ship, scrap, or rework with a straight face.

Closing the loop

When the investigation ends, close the hold with the same identity you opened. Update the traveler, the electronic batch or process record, and the warehouse location. If CAPA is required, write the problem statement from the hold criteria, not from the meeting minutes. Then verify the fix on the next run of the same asset, not on a convenient cousin.

For this automotive case, keep the hold tied to the exact asset and lot named in the title. Do not broaden the story into a generic quality essay. Name the instrument, the sample, and the clock.

Field notes from similar holds

Shift supervisors often ask whether they can keep packing while quality 'looks at it.' Packing during an open identity question creates mixed cartons. Mixed cartons become mixed complaints. Stop packing until identity is confirmed, even if the dock appointment is in forty minutes.

Laboratories often ask whether they can rerun immediately. A rerun without preserving the first preparation destroys the only independent check you had. Freeze or retain the first preparation, then rerun. If the method consumes the sample, say so in the hold record and escalate to the method owner.

Maintenance often asks whether they can 'touch it up' so the next audit looks cleaner. Touch-up is a change. Changes need a record. If the touch-up is truly emergency restoration, photograph first, restore, then open change control before the next commercial run.

Purchasing often asks whether the supplier lot can be used on a different line. Different lines have different qualified states. A foil lot that failed pinholes on blister line 2 is not automatically acceptable on line 4. Treat it as a new question with its own as-found.

Customer quality often asks for a date. Give a date for the next documented decision, not a promise of release. Promises of release before the evidence pack is complete are how sites write themselves into a missed notification.

Method version control

Print or export the method or work instruction that was open at the minute of the signal. If the screen showed revision 9 and the binder on the bench is revision 7, that disagreement is part of the hold. Do not quietly swap the binder.

If the method allows a range, write the actual setting used, not the range. 'Mixer 45 rpm' can be reconstructed. 'Mixer 30-60 rpm' cannot.

If the method was executed on a substitute instrument, name the substitute and the bridging study that allowed it. If there is no bridging study, the hold now includes an unapproved method change.

Sampling geometry

Where the sample came from matters as much as the number. Top of tank versus bottom, leading edge versus trailing edge of a web, first cavity versus last cavity of a tool. Draw it. A sketch in the hold record beats a paragraph.

How long the sample sat matters. Water activity drifts as the snack cools. Viability drops as the flask waits on the bench. Torque readings change as the joint relaxes. Write the wait time.

Who held the sample matters. If the only person who can explain the pull already left for the weekend, call them before you invent a narrative.

Independent checks

Independent does not mean a coworker glancing over a shoulder. It means a second qualified person, same method, same lighting, no coaching, written result. If you cannot staff that, say so and keep the hold open.

Where the method uses an instrument, independent may mean a second instrument that is in calibration. Where it uses a visual standard, independent means a second set of eyes on the same retained unit, not a new unit from a different carton.

Record disagreement without blame. Disagreement is data. Blame makes people stop writing.

Genealogy and containment radius

Walk one step upstream: supplier lot, previous process, previous shift's leftover. Walk one step downstream: packed goods, shipped goods, customer locations if known. Set the containment radius in writing. Expand it only with new facts.

If shipped goods are in the radius, notify the function that owns customer communication on the same day. Waiting until 'we know more' is how notification clocks get missed.

If leftover material from the prior SKU is in the radius, treat it as a mix-up until proven otherwise. Mix-ups are not theoretical in changeover holds.

Decision types

Continue under deviation: allowed only when residual risk is named, accepted by the process owner, and the next checks are defined.

Rework: allowed only when the rework method is already qualified or is qualified before the units return to stock.

Scrap: allowed when identity is intact enough to destroy the right units. If identity is not intact, you may have to scrap a larger radius.

More data: allowed as a dated pause, not as an endless state. Name the test, the lab, and the due time.

What inspectors actually ask

They ask for the as-found. They ask who called the hold. They ask whether downstream units were checked. They ask whether the method version matched. They ask whether the second observer was independent. They ask when the disposition was dated. If your hold record answers those without a scavenger hunt, you are done.

They also ask why a similar event last quarter did not prevent this one. Have the prior CAPA number ready, or say there was no prior CAPA and that this investigation will decide whether one is needed. Pretending last quarter's CAPA covered this event is worse than saying it did not.

After the hold

Verify the fix on the next run of the same asset. A cousin asset is a different qualified state. File the evidence pack where the next shift can find it. Update any visual standard, recipe, or program that the investigation proved was wrong. Then stop talking about the hold in hallway terms. The record is the only version that counts.