An incoming inspection sampling plan is a documented, risk-based procedure for deciding how many units to pull and examine from an incoming supplier lot — reagents, consumables, components, or devices — and how many defective units in that sample are enough to reject the whole lot. It exists because inspecting every unit in every shipment (100% inspection) is usually too slow, too expensive, or in the case of destructive testing, impossible, while inspecting nothing at all leaves a lab exposed to a bad lot reaching patients, researchers, or downstream processes undetected.
This guide covers what a sampling plan actually specifies, how the ANSI/ASQ Z1.4 standard (and its international counterpart, ISO 2859-1) structures that decision, how to set one up for a lab’s own incoming supplies, and where it fits alongside the other receiving-quality controls a medical or research lab already runs.
What an Incoming Inspection Sampling Plan Actually Specifies
A complete sampling plan answers three questions for a given incoming lot:
- How many units get inspected (n) — the sample size, drawn from the lot at random.
- What quality level is the plan built around (AQL) — the Acceptable Quality Limit, expressed as a percentage or defects-per-hundred-units, that represents the worst tolerable long-run defect rate for that item.
- What sample result accepts or rejects the lot (Ac/Re) — the acceptance number (Ac) is the maximum number of defective units the sample can contain and still pass; the rejection number (Re) is the number that fails it.
Everything else in a sampling-plan standard — lot-size brackets, inspection levels, code letters, switching rules — exists to make those three numbers repeatable and defensible instead of ad hoc.
Why Sample Instead of Inspecting Every Unit
100% inspection sounds safer than sampling, but in practice it has two well-documented weaknesses that make a sampling plan the more reliable choice for most incoming lots:
- It’s often not possible. Sterility testing, tensile-strength testing, and other destructive or consumptive tests destroy the unit tested — you cannot inspect 100% of a lot of sterile swabs without using up the lot.
- Inspector fatigue lowers detection rates. Quality-engineering literature on inspection accuracy has repeatedly found that 100% manual inspection of large lots misses a meaningful fraction of defects simply because attention degrades over a long, repetitive task — a statistically designed sample, inspected carefully, can outperform an exhausted full inspection.
A sampling plan trades certainty about any single lot for a known, quantified, and documented risk across all lots — which is exactly what a quality system audit or regulatory inspector wants to see: not zero risk, but risk that was deliberately chosen and can be explained.
ANSI/ASQ Z1.4 and ISO 2859-1: The Standard Most Plans Are Built On
ANSI/ASQ Z1.4, Sampling Procedures and Tables for Inspection by Attributes, is the American national standard most incoming-inspection plans in the US are built on. It descends from MIL-STD-105E, the US military sampling standard that was cancelled in 1991; industry adopted the same statistical tables under the civilian ANSI/ASQ designation, and they have stayed in continuous use since. ISO 2859-1 is the international standard covering the same territory, built on an equivalent statistical structure, so a lab operating under both a US quality system and an ISO-based one (ISO 13485, ISO 9001) can generally satisfy both with one Z1.4/ISO 2859-1–based program rather than running two.
The standard organizes a plan around four elements:
- Lot size — the number of units in the incoming shipment or batch being evaluated.
- Inspection level — how much discriminating power the sample needs. General levels I, II, and III trade off sample size against how confidently the plan distinguishes a good lot from a bad one (Level II is the standard default; Level I uses a smaller sample when less discrimination is acceptable; Level III uses a larger one when more is needed). Four additional special levels (S-1 through S-4) exist for cases such as destructive or costly testing, where keeping the sample small matters more than statistical power.
- Sample-size code letter — a lookup value determined jointly by lot size and inspection level, which in turn points to the actual sample size (n) in the standard’s tables.
- Acceptable Quality Limit (AQL) — chosen separately, per characteristic, based on how severe a defect of that type is. Combined with the sample-size code letter, the AQL determines the acceptance number (Ac) and rejection number (Re) from the standard’s master tables.
Z1.4 also supports single, double, and multiple sampling plans (one sample and a single accept/reject decision, versus a first sample that can trigger a second before a final decision, versus a sequence of smaller samples) and defines switching rules between normal, tightened, and reduced inspection based on a supplier’s recent lot history — a supplier with a run of consecutive accepted lots can move to a lighter inspection regime; one with recent rejections moves to a stricter one. Because the exact sample sizes and Ac/Re values in the published tables are precise, load-bearing numbers, don’t reproduce them from memory or a secondary summary — pull the current edition of the standard (available through ASQ or ANSI) or an ISO 2859-1–licensed copy when setting acceptance numbers for an actual receiving program.
How to Build a Sampling Plan for a Lab’s Incoming Supplies
- Define what counts as a “lot.” Usually a single shipment of one part number/catalog number from one manufacturing lot or batch number. Mixing multiple manufacturing lots into one inspection lot defeats the purpose of lot traceability.
- Classify the item’s defect severity. Most programs sort possible defects into critical (could cause patient harm, a failed result, or a safety hazard), major (would likely cause the item to fail its intended use), and minor (a departure from specification unlikely to affect use) — and assign a tighter AQL to critical characteristics than to minor ones.
- Choose an inspection level and AQL per item/characteristic. This is a quality and risk-management decision, not a statistical one — it belongs in the same review that sets a supplier’s risk tier or a reagent’s criticality classification.
- Pull the sample size and Ac/Re from the current published tables for the item’s lot size, inspection level, and AQL.
- Document the plan in the receiving SOP, including who draws the sample, how randomization is done, what’s inspected (visual, dimensional, functional, a certificate-of-analysis cross-check), and what happens on rejection (return, quarantine, supplier notification, corrective action request).
- Apply the switching rules so inspection intensity tracks actual supplier performance instead of staying static regardless of a supplier’s track record.
- Keep records tying each lot’s disposition back to its sampling plan and result — this is what turns “we inspected it” into an auditable, defensible record.
Zero Acceptance Number (c=0) Plans: A Common Alternative
Many medical device and pharmaceutical incoming-inspection programs use zero acceptance number plans instead of, or alongside, a Z1.4-derived AQL plan. A c=0 plan sets the acceptance number at zero — a single defective unit in the sample rejects the whole lot — and derives sample size from the desired confidence and reliability level (a Lot Tolerance Percent Defective, or LTPD, approach) rather than from an AQL. The most widely referenced published tables for this approach come from Nicholas Squeglia’s Zero Acceptance Number Sampling Plans. The appeal for critical items is that Ac=0 avoids the counterintuitive situation an AQL plan can produce, where a lot is accepted despite containing one or more confirmed defective units, which is harder to defend for anything safety-critical. The tradeoff is a larger required sample size to reach an equivalent confidence level, since c=0 statistically concentrates all of the plan’s discriminating power into a single acceptance threshold.
Where This Fits in a Broader Receiving and Supplier-Quality Program
A sampling plan is one control inside a larger incoming-quality program, not a replacement for the rest of it:
- Supplier qualification happens before the first shipment ever arrives — see CASRAI’s guide to verifying a supplier’s ISO/IEC 17025 accreditation for one common qualification check for testing and calibration suppliers.
- Quality agreements with a supplier define contractual quality expectations, including who is responsible for what testing — see CASRAI’s quality agreement guide.
- Lot-to-lot reagent verification is a related but distinct control specific to reagents, checking that a new lot performs equivalently to the one it replaces before it’s released for patient or research use — see CASRAI’s guide to reagent lot-to-lot verification.
- Calibrated measuring equipment used to perform incoming inspection must itself be in calibration — see CASRAI’s guide on what to do with an out-of-tolerance calibration finding.
- A LIMS or quality system is typically where sampling-plan results, Ac/Re outcomes, and lot dispositions actually get logged and trended over time — see CASRAI’s guide to GMP LIMS requirements.
Regulatory Context for Medical and Clinical Labs
For labs manufacturing or distributing medical devices, FDA’s Quality System Regulation at 21 CFR 820.80(b) requires that incoming product be inspected, tested, or otherwise verified against defined acceptance criteria before use or release, and that any sampling plan used be based on a valid statistical rationale and documented. ISO 13485:2016, clause 7.4.3 (verification of purchased product), imposes a parallel requirement for organizations certified to that standard. Neither standard mandates ANSI/ASQ Z1.4 or ISO 2859-1 specifically — they require that whatever method is used be documented, statistically justified, and consistently applied; Z1.4/ISO 2859-1 and c=0 plans are simply the most widely recognized ways to meet that bar without designing statistical tables from scratch.
Clinical laboratories operating under CLIA (42 CFR Part 493) have a related but less prescriptive obligation: quality control requirements call for verifying that reagents and supplies perform as expected before use, but CLIA itself does not specify a statistical sampling methodology, leaving the lab to define a risk-based approach appropriate to what it’s receiving. ISO 15189, the international standard for medical laboratory quality, similarly requires that supplies affecting the quality of examinations be evaluated before use without mandating a specific sampling standard. In both cases, a documented Z1.4-, ISO 2859-1–, or c=0–based plan is a defensible way to satisfy an otherwise open-ended “verify before use” requirement.
Common Mistakes
- Using one AQL for every item regardless of criticality. A single blanket AQL across critical and minor characteristics either under-protects the critical ones or wastes inspection effort on the minor ones.
- Never applying the switching rules. Running every lot from every supplier at the same fixed inspection level, forever, ignores the very mechanism the standard provides to reward reliable suppliers and tighten up on unreliable ones.
- Treating “sample passed” as “lot is defect-free.” A sampling plan accepts a lot with a known, quantified probability of still containing defects at or below the chosen AQL — it does not certify zero defects, and staff relying on incoming inspection should understand that distinction.
- No documented rationale for the chosen AQL, inspection level, or plan type. If an auditor or inspector asks why a particular AQL was chosen for a particular item, “that’s what we’ve always used” is not a statistical rationale.
- Letting the sampling plan replace, rather than complement, other receiving controls such as certificate-of-analysis review, visual damage inspection, and temperature/condition checks on arrival.
Frequently Asked Questions
What is an AQL in a sampling plan?
The Acceptable Quality Limit is the worst defect rate, expressed as a percentage or defects per hundred units, that a sampling plan treats as an acceptable long-run process average for a given characteristic. It is a policy choice made per item and per defect severity, not a statistical output — the sample size and acceptance/rejection numbers are then derived from it.
Is ANSI/ASQ Z1.4 legally mandatory?
No. FDA’s 21 CFR 820.80 and ISO 13485’s clause 7.4.3 require a documented, statistically valid sampling rationale for incoming inspection, but neither names Z1.4 specifically. Z1.4 (or its international equivalent, ISO 2859-1, or a c=0/LTPD-based plan) is simply the standard most organizations use to satisfy that requirement without building their own statistical tables.
What’s the difference between Z1.4 and ISO 2859-1?
They cover the same territory — attribute sampling by lot size, inspection level, and AQL — and share a common statistical lineage. Z1.4 is the American national standard (ANSI/ASQ); ISO 2859-1 is its international counterpart. A lab certified to both a US-based and an ISO-based quality system can typically run one program built on either standard and satisfy both.
Do small labs need a formal sampling plan, or can they just inspect everything?
If lot sizes are small enough that 100% inspection is genuinely feasible and non-destructive, some labs do inspect every unit rather than sample. The tradeoff is inspector fatigue on larger lots and the impossibility of 100% inspection wherever testing is destructive. Even a small lab benefits from documenting its inspection approach and rationale, whether that approach is 100% inspection or a formal sampling plan.
Why would a lab use a zero acceptance number (c=0) plan instead of an AQL plan?
A c=0 plan rejects a lot on a single confirmed defect in the sample, which avoids the situation an AQL-based plan can produce — accepting a lot that contains one or more known defective units. For safety-critical items, that’s often easier to defend to an auditor or regulator, at the cost of a larger required sample size for an equivalent confidence level.
How does incoming inspection sampling differ from statistical process control?
Incoming inspection sampling is a lot-by-lot accept/reject decision applied to material a lab receives from someone else. Statistical process control (SPC) monitors a process’s own output over time to detect drift before it produces defects. A mature supplier-quality program typically uses incoming sampling as a gate on received material and, where the lab has visibility into supplier data, SPC-style trending of lot results over time to catch a supplier’s process drifting before it starts failing incoming inspection.
