Quality Escape Prevention: The Four Ways Parts Get Out

A quality escape is almost never the result of someone deciding to ship a bad part. It is the result of a system that was pointed at the wrong question. The inspector measured a bore with a two-point micrometer, the drawing meant a three-point one, and the part passed a check that did not test the requirement. By the time the defect surfaces it is at your assembly line, in a lot of 500, and the cost of sorting it is 10 to 100 times the price of the part. Understanding the four mechanisms that let defects through is the fastest route to closing them.

Escapes are process failures wearing an operator costume

When a bad lot arrives, the tempting response is to blame the person at the machine. That person is usually doing exactly what the work instruction says. The instruction was written for a different failure mode, the gauge was chosen for convenience, or the sampling plan was designed for a process that no longer exists.

Look at where the defect was created, where it could have been detected, and where it actually left the building. Those three points are almost never the same, and the gap between the second and third is the escape.

A shop running without an outgoing check typically holds 500 to 2,000 defects per million parts on a stable job. With method-matched gauging, a meaningful outgoing sample and packaging validation, that falls to 100 to 400. The difference is process design, not effort.

Mode one: the measurement method disagrees with the drawing

This is the most common escape and the hardest to see, because the inspection report is full of passing numbers. A bore specified at 12.000 +0.021 / 0 measures round on a two-point micrometer and can be 0.015 mm oval. A position tolerance measured with a height gauge and trigonometry is an approximation of a true-position requirement. A surface finish compared against a visual comparator is not a roughness measurement.

The fix is to agree the method at the first article and write it down per feature: bore measured on a bore gauge or CMM with a known probe radius; flatness on a surface plate with a dial indicator; thread checked with a go/no-go gauge at the specified class; radius with a profile projector. That list becomes the inspection plan, and the inspection plan follows the part through every lot.

See also: How to Build an Incoming Inspection Plan for Machined Parts and How to Qualify a Second Source for a Running Part.

Mode two: sampling gaps at the edges of a run

Sampling plans assume a stable process. Real lots have edges: the first part after a setup, the first part after a tool change, the last part before the bar runs out, and the first part after a shift handover. Those are statistically the most likely to be out, and they are exactly the parts a simple "one in twenty" sample will miss.

Mode three: damage after the last inspection

Damage happens after the last inspection, during counting, bagging, boxing and freight. Packaging validation is cheap and rarely done. Drop a filled carton from 800 mm onto a corner, open it, and inspect. If the parts survive, the packaging is adequate for parcel freight. If they do not, add a layer of foam, a divider or an individual bag. For parts above 2 kg, add a load test: stack three cartons for 72 hours and check for deformation.

Mode four: documentation that hides the problem

A part can be correct and still be an escape if the paperwork is wrong. The material certificate names a different heat than the parts. The finish certificate covers a different lot. The inspection report was copied from the previous run and the values were not updated — a failure that happens more often than anyone admits, because the spreadsheet makes it easy.

Three checks catch most of it. The lot number on the box must match the lot number on the inspection report and the work order. The heat number on the material certificate must match the heat number on the work order. The quantity on the packing list must match the quantity inspected. These take 90 seconds per shipment and are the highest-value 90 seconds in the whole process.

The four controls that stop most of it

The first control is a signed method sheet. Every checked feature has a named instrument, and the inspector uses that instrument. This single control prevents the method-mismatch family, which is the largest single source of escapes.

The second is edge sampling. First part after setup, first part after any tool change, first part after shift change, and last part of the run. Four checks per lot, placed where risk is highest, rather than ten checks placed randomly.

The third is a validated pack. One drop test, one photo, one written pack specification that the packer follows. The specification should name the bag, the divider, the foam and the carton size, not just say "protect the finish."

The fourth is a release gate with authority. One person confirms lot, heat and quantity match, confirms the inspection report is complete and current, and signs the shipment. The gate must be able to stop a truck. A gate that can only annotate is not a gate.

When a defect does arrive, do not ask for a corrective action on the part. Ask which of the four controls failed, and what the fix is at that control. The answer is usually specific and cheap: a different gauge, an extra check at shift change, a divider in the carton, a second signature on the release.

If you want to see how these controls are applied on a running job, our how we work page covers the inspection and release sequence, and the contact page is the place to raise a specific escape you are trying to close.

Frequently asked questions

What is a quality escape in machining?

A defect that reaches the customer despite the supplier's inspection process. It is not the same as a defect: most defects are caught internally, and the escape is the subset that got past every check because the check was testing something else.

What is the most common cause of escapes?

Measuring the wrong way. A gauge that cannot detect the failure mode, such as a two-point micrometer on an out-of-round bore, will report a passing number and the defect will ship with a clean certificate.

How do I stop shipping damage on machined parts?

Validate the pack with a drop test from 800 mm onto a corner, photograph the packed carton, and write a pack specification naming the bag, divider and foam. Individual bagging for cosmetic parts is usually cheaper than one rejected lot.

How often should outgoing inspection be performed?

At minimum on the first part after setup, the first part after a tool change, at shift change and on the last part of the run. These edge points carry most of the risk, and four checks placed there beat ten placed randomly.

Related reading

• A Realistic First Article Inspection Timeline2026-10-15
• A Supplier Audit Checklist for Uncertified Machine Shops2026-10-13
• Machining Purchase Order Terms That Actually Protect the Buyer2026-10-18