Incoming Inspection in Practice: Sampling, Tests, and Records

Receiving inspection is not a moral exercise about counterfeit parts. It is a sampling decision, a short list of tests, and a record that lets someone else reproduce your conclusion.

1. Sampling: what n = 125 actually buys

ISO 2859-1 (GB/T 2828.1) links lot size to a code letter, and the code letter to a sample size. At general inspection level II, a lot of 1,201–3,200 pieces takes code K, which is 125 samples.

With n = 125 and AQL 1.0, the acceptance number is 3: accept the lot if 3 or fewer samples fail, reject if 4 or more. That is the standard's table, not a rule of thumb.

That curve is the most useful thing to show a manager who asks why a bad lot was accepted. Sampling is a probability, not a guarantee: a plan that accepts 96 % of 1 %-defective lots still accepts 4 % of them, and that is by design.

Worked example: a 2,000-piece lot of an active IC, level II, AQL 1.0. Code K, n = 125, Ac = 3. Inspect 125 pieces taken from both ends of the reel, find 2 failures, accept the lot and record the 2 failures. Find 4 failures, reject, quarantine, and open a supplier claim. The difference between those two outcomes is one part, which is why the rule must be decided before the inspection, not after.

The plan also has a switching rule: after several consecutive accepted lots, the standard allows reduced inspection; after a rejection, normal inspection resumes and tightened inspection may apply. Use the rule rather than an informal "we know this supplier".

Two sampling details decide whether the result means anything. First, the lot must be homogeneous: parts from two different date codes are two lots, even if they arrived on one purchase order. Second, the sample must be random across the reel — take pieces from the beginning, the middle and the end, not the first 125 in the tape.

2. Choosing the AQL

Two practical adjustments. First, reduce inspection after a run of clean lots (switching rules in the standard allow it), and restore it after a failure. Second, do not apply a sampling plan to a single reel of a critical part — 100 % inspection of a 500-piece lot costs minutes and removes the question.

3. What to inspect, by part class

4. Tools, and what each one is for

A workable minimum kit: a 10–20× loupe or stereo microscope, a pair of calipers, an LCR meter, a curve tracer or a simple pin-to-pin tester, and a controlled DC supply. Add XRF if RoHS declarations matter to you, and X-ray if you buy BGAs or QFNs in volume.

Solderability is its own test (J-STD-002) and is worth doing on any lot older than 24 months or stored outside a dry cabinet. MSL status follows J-STD-033: if the bag was opened, the floor-life clock is running.

Calibration matters as much as the instrument. An LCR meter that has not been calibrated in two years is a random number generator for tolerance checks, and an uncalibrated caliper cannot support a dimensional rejection. Put both on a calibration schedule and keep the certificates with the inspection records.

For passives, the cheapest useful test is a value check on a sample plus a visual check for mixed reels. For crystals and oscillators, measure frequency and ESR; a crystal with the wrong load capacitance will be several tens of ppm off and still look perfect.

5. Records: the part that survives an audit

An inspection that leaves no record is an opinion. The record should let a different person, six months later, reproduce the conclusion:

• Lot and purchase order, supplier, manufacturer, date code, quantity.
• Sample size, AQL, and the standard used.
• Photos of the label, the reel, and 2–3 parts at the ends of the strip.
• Measured values (not just pass/fail) for the sample.
• The decision, the date, and who approved it.
• Any deviation and the disposition (accept with concession, rework, return).

Store photos with the lot record, not in a chat thread. If a dispute starts, the photo timestamp and the sample retention are the evidence.

Retention is a policy decision, not a default. A practical rule: keep the records for the warranty period plus one year, keep the physical sample for at least the warranty period, and keep photographs forever — they cost nothing and answer questions that come up long after the lot is gone.

If the product ships to regulated markets, the record may be audited years later. In that case the record format matters as much as its content: a scanned PDF with a date and a name is worth more than a spreadsheet row that nobody signed.

6. How deep to inspect

The matrix in Figure 4 is a planning tool, not a standard. Its value is that it forces the question: is this a high-risk part from an unknown supplier, or a commodity from a supplier with two years of clean history? The answer changes the plan, and writing it down stops the discussion from restarting every lot.

A workable default for a small team: full visual and paperwork for every lot, electrical sample for new suppliers and critical parts, X-ray or XRF only for disputes or first articles. Review the matrix once a quarter against the scorecard, and move suppliers between cells based on evidence rather than impressions.

7. The decision, and who owns it

Quarantine means a separate location with a label and a hold tag, not a different shelf. The holder of the hold should not be the person who bought the parts — that separation is what makes the decision credible.

When a lot is rejected, notify the supplier in writing with photos and measurements, and ask for a corrective action rather than a replacement only. A replacement fixes this lot; a corrective action fixes the process.

A useful corrective-action request has four parts: what was found (with the measurement), how many of the sample, what the supplier's process should have prevented, and what evidence will show it is fixed. A reply that says "we will be more careful" is not a corrective action.

Track the response time. A supplier that answers in two days with a process change is a different risk from one that answers in three weeks with a replacement reel, and the scorecard should reflect that difference.

8. Metrics that make the program improve

A simple rule: if the escape rate is rising while incoming DPPM is flat, the sampling plan is too shallow or the tests are not measuring the right thing.

An example of the arithmetic: 5,000 pieces received, 2 defects found in the sample, and 3 more found on the line. Incoming DPPM is 400, but the true defect rate is higher because the sample only found part of it. That gap is the signal to widen the sample or add a test, not to congratulate the supplier on a low DPPM.

9. Receiving checklist

This is the same evidence base used in the counterfeit checklist; receiving is where that checklist becomes a routine. For lot-level traceability on our side, see the traceability process.

10. FAQ

Do I need AQL sampling if I buy from an authorized distributor?

You can reduce it, but not to zero. Authorized channels have far lower counterfeit risk and still ship wrong-date-code, damaged or mispackaged lots. A reduced visual check is cheap.

How many samples for a small lot?

Follow the table: 281–500 pieces gives code H and n = 50 at level II; 501–1,200 gives J and n = 80. For very small lots, 100 % inspection is often simpler.

What if only one sample fails?

Compare with the acceptance number for your plan. With n = 125 and Ac = 3, one failure still accepts the lot — but record it and watch the supplier trend.

Do I have to keep samples?

Keep a small retention sample (5–10 pieces) and photos for the warranty period. If a field failure appears, the retained parts answer the first question.

How do I inspect a BGA without X-ray?

You cannot inspect the joint, but you can inspect the part: ball diameter consistency, coplanarity, marking, and a sample X-ray from a lab. For production volumes, budget for X-ray or 5D CT on first articles.

What does "tightened inspection" mean in practice?

One level down in AQL for the next lots, or a larger sample size, until a defined number of clean lots restores normal inspection. The rule should be written before the failure, otherwise it becomes a negotiation.

Should I inspect every lot of a cheap passive?

Usually no. Sample at a looser AQL, inspect the packaging, and rely on the supplier scorecard. The cost of inspecting a 1-cent resistor exceeds the cost of the few that fail.

Can inspection replace supplier qualification?

No. Inspection catches defects at the lot level; qualification decides whether the supplier's process is capable. They are complementary, and the second is cheaper in the long run.

11. Sources

• ISO 2859-1 — sampling procedures for inspection by attributes (AQL tables, code letters, switching rules).
• JEDEC, J-STD-002 — solderability tests for component leads and terminations.
• JEDEC, J-STD-033 — handling, packing and floor life of moisture-sensitive devices.
• SAE, AS6081 — counterfeit avoidance for distributors.
• IDEA, IDEA-STD-1010 — inspection standard for electronic components.