How to Build an Incoming Inspection Plan for Machined Parts

Most incoming inspection plans are copies of a template that checks too much on the wrong parts and too little on the right ones. The result is a receiving area that spends 40 minutes on a bracket and waves through a housing with a 0.02 mm bore fit. A workable plan starts from what happens if the feature fails, assigns an inspection level to that consequence, and then chooses a sample size and a gauge that can actually detect the failure. This is how to build one for machined parts without a quality department of twelve.

Rank the part before you write the plan

Sort every purchased part into one of four levels based on consequence, not price. Level A is a feature whose failure stops an assembly line or creates a safety concern: a bearing bore, a mounting pattern that locates a sensor, a sealing face. Level B affects fit or function but is recoverable by rework at your plant. Level C is cosmetic or non-functional. Level D is commodity hardware and standard fasteners.

Most companies discover that 10 to 15 percent of part numbers are level A, 30 to 40 percent are level B, and the rest can be handled with a dock check. Writing a plan by level instead of by part number is what makes it survivable; a plan with 400 individual part-specific procedures will be followed for six weeks.

What AQL sampling actually buys you

AQL stands for acceptance quality limit, and it is not a defect rate target. It is the worst quality level that the sampling plan will accept 95 percent of the time. A plan at AQL 1.0 with a sample of 80 accepts lots containing about 1 percent defective most of the time, and rejects most lots that are much worse.

The practical mechanics: choose an inspection level (general level II is standard), find the code letter from the lot size, read the sample size, then apply the acceptance number for your AQL.

Two rules matter more than the table. Zero acceptance for any critical characteristic: one nonconforming part rejects the lot regardless of sample size. And never accept on the basis of a sample you did not actually measure — an inspector who pulls 80 parts and measures 12 has run a different plan.

For most level A parts, AQL 0.65 at general level II is a reasonable default; for level B, AQL 1.0 to 1.5; for level C, AQL 2.5 or visual-only inspection.

See also: Quality Escape Prevention: The Four Ways Parts Get Out and How to Qualify a Second Source for a Running Part.

Critical versus general dimensions

Not every dimension on a drawing deserves the same attention. Split them into three groups and attach a different rule to each.

Critical dimensions affect fit, function or safety. They get 100 percent inspection on the first lot and a defined sample on subsequent lots. Typically 5 to 12 dimensions on a machined part, and they should be the ones your assembly actually depends on, not the ones that were hardest to machine.

General dimensions affect form or appearance and are covered by the drawing's general tolerance block. Sample three to five of them per lot, chosen to cover different setups and different tools; that catches a setup error or a worn tool without turning inspection into a full layout. Then give the critical list to the supplier: a shop that knows which six dimensions matter protects them in-process, and you see fewer rejects before your own inspection changes.

Gauge R&R before you trust any number

A gauge that cannot repeat its own measurement cannot tell you whether a part is good. Gauge R&R quantifies that: it compares measurement variation to the tolerance and to the part-to-part variation.

The usual rule: under 10 percent of tolerance is acceptable, 10 to 30 percent is conditional and needs a clear reason, over 30 percent is unacceptable. If your receiving inspection uses a caliper with 0.02 mm resolution on a ±0.01 mm feature, the R&R alone will exceed 30 percent and every decision based on it is a coin flip.

First article versus production lots

These are different exercises and should not share a checklist.

The first article is a full layout: every dimension on the drawing, reported with actual values, plus material and finish certificates and a visual against the sample. It proves the process could make a conforming part and it establishes the reference measurements you will use for comparison later.

Production lots are sampled, focused on critical dimensions, and require a lot file: material heat number, inspection record, quantity and packaging check. A lot that arrives with a first article report attached but no lot-specific data has not been inspected; it has been referenced.

Running the plan without a quality department

Assign one person to own receiving inspection, give them the gauge list, and give them authority to reject. A plan without the authority to hold a lot is documentation, not control.

If a supplier consistently passes your inspection and then fails in production, the problem is usually the plan, not the supplier — you are checking dimensions that do not predict assembly failure. Rebuild the critical list from the failures.

Our team can supply a lot file and full dimensional report structured to slot into a plan like this; start with the contact page and reference the inspection level you need. The capabilities page lists the metrology equipment behind those reports.

Frequently asked questions

What AQL should I use for machined parts?

AQL 0.65 at general inspection level II for critical fit features, 1.0 to 1.5 for general functional dimensions, and 2.5 for cosmetic. Apply zero acceptance to any critical characteristic regardless of the sample size.

How many dimensions should incoming inspection check per lot?

Four to six critical dimensions plus three general ones on a typical machined part. The critical list should be built from assembly failures, not from the drawing's most difficult features.

What is an acceptable gauge R&R?

Under 10 percent of tolerance is acceptable, 10 to 30 percent is conditional, and above 30 percent the gauge cannot support the decision. Calipers on ±0.01 mm features almost always fail this test.

Should I do 100 percent inspection on critical parts?

Only on critical features, not on the whole part. Full dimensional inspection of every unit is slow and adds no value for general dimensions; 100 percent on five features and sampling on the rest is the practical compromise.

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