If your sampling procedure still cites ISO 2859-1:1999, it cites a withdrawn standard. ISO published the third edition on 22 January 2026 and withdrew the 1999 edition the same day. Here are the ISO 2859-1:2026 changes that matter to a food or packaging site, and a plain explanation of how an AQL plan works and what it cannot tell you.
ISO 2859-1:2026 (3rd edition, 22 January 2026) replaces the 1999 edition and its 2011 amendment. It adds skip-lot inspection, moves OC curves into a method in Annex E, and reworks the tables for usability. The switching-rule table also now counts internal suppliers as suppliers.
The QC Inspector Course for Food and Beverage Manufacturing teaches AQL plans, micro reports and what sampling cannot prove on a working snack and beverage site, with a free AQL calculator: about 20 hours online, R2 950.
ISO 2859-1:2026 changes: the key facts
| Standard | ISO 2859-1, Sampling procedures for inspection by attributes, Part 1: sampling schemes indexed by acceptance quality limit (AQL) for lot-by-lot inspection |
|---|---|
| Edition | 3rd edition, published 22 January 2026, 82 pages |
| Replaces | ISO 2859-1:1999 and Amendment 1:2011; Corrigendum 1:2001 is also withdrawn |
| New | Skip-lot inspection, added as a further step in the switching rules |
| Moved | Per-plan OC curves removed; Annex E gives methods to build OC and ASN curves |
| Clarified | “Supplier” in the switching-rule table includes internal suppliers |
| Tables | “Modified to enhance usability”; confirm every value in your licensed copy |
| United States | A revision of ANSI/ASQ Z1.4-2003 (R2018) is planned |
| SANS adoption | Not confirmed as at September 2026; check with SABS |
| Not for | Critical defects, on-line process control and microbiological testing |
What changed in ISO 2859-1:2026, and what stayed the same?
What changed is the scheme around the tables: skip-lot inspection as a further switching step, OC curves built by a method instead of printed per plan, and internal suppliers counted in the switching rules. What did not change is the principle: an AQL plan judges a lot from a random sample, and a pass is a probability.
- Skip-lot inspection. The 1999 scheme moved a supplier between normal, tightened and reduced inspection. The 2026 edition adds skip-lot as a further step, so a supplier with a good record may have some lots skipped. Take the entry and exit conditions from your licensed copy, not from a summary.
- OC curves. Instead of printing an operating characteristic (OC) curve for each plan, the 2026 edition gives a method in Annex E to build OC and average sample number (ASN) curves, which is what a calculator does for you.
- Internal suppliers. If your own blow-moulding or printing department feeds the filling hall, it can now be treated as a supplier under the switching rules.
- The tables. ISO says the sampling plan tables were modified to enhance usability. No public source says whether any code letter or acceptance number changed, so do not assume they did not.
For most sites the job is short: update the edition your sampling procedure cites, and confirm every plan you run against the licensed 2026 copy before your next audit.
What is an AQL, and what is it not?
The acceptance quality limit (AQL) is the worst average defect rate a sampling scheme tolerates over a continuing series of lots. Lots at or better than the AQL are usually accepted, which protects the producer. It is not a promise that any one lot is that good, and it is not a defect rate the customer has agreed to receive.
AQL sampling for food packaging and ingredients suits incoming lots (film reels, closures, cartons, preforms), final lot audits before dispatch and customer lot inspection, such as a retailer’s label audit. It does not suit on-line control, which uses fixed-frequency samples and control charts; critical defects; or microbiological testing, which has its own plans.
How do you read a sampling plan?
You need three inputs: the lot size, the inspection level (general level II unless your plan says otherwise) and the AQL for the defect class. The lot size and level give a code letter and a sample size n; the code letter and the AQL give the acceptance number Ac and the rejection number Re.
All values below are from the ISO 2859-1:1999 master tables; confirm against your site’s licensed 2026 copy.
| Lot size | Code letter (level II) | Sample size n |
|---|---|---|
| 1,201 to 3,200 | K | 125 |
| 3,201 to 10,000 | L | 200 |
| 10,001 to 35,000 | M | 315 |
| Code letter (n) | Ac/Re at AQL 1.0 · 2.5 · 4.0 |
|---|---|
| K (125) | 3/4 · 7/8 · 10/11 |
| L (200) | 5/6 · 10/11 · 14/15 |
| M (315) | 7/8 · 14/15 · 21/22 |
Single sampling, normal inspection. Ac is the most defectives you may find and still accept the lot; Re is the number at which you reject it. The inspection level changes n for the same lot: for a lot of 5,000, the 1999 tables give 80 at level I, 200 at level II and 315 at level III, and 5, 8, 20 and 32 at the special levels S-1 to S-4, which are for costly or destructive tests.
So what sample size do you need for AQL 1.0 on a lot of 5,000? On the 1999 master tables at level II: code letter L, n = 200, accept on 5 defectives, reject on 6. Enter your own lot in the free AQL Sampling Plan Calculator to see the probability of acceptance and the OC curve.
Worked example: a label audit on 5,000 bottles
16:00 on Wednesday 7 October 2026 at Isipho Beverages & Snacks in Hammarsdale, the fictional factory in ASC’s Course 2. A retail customer wants a label audit on 5,000 bottles of Isipho Cola 2 L before dispatch. Priya Naidoo, the QA manager, asks B-shift inspector Ayanda Mthembu to run it to Isipho’s AQL plan for label placement.
Step 1: Define the lot
One batch of 5,000 bottles from one line and shift: the 3,201 to 10,000 row.
Step 2: Find the code letter and sample size
General inspection level II gives code letter L and n = 200 (1999 master-table values; confirm in the licensed 2026 copy).
Step 3: Find Ac and Re
Label placement is a major defect, and Isipho’s AQL for majors is 1.0 (case data). Code L at AQL 1.0 gives Ac 5, Re 6.
Step 4: Draw a random sample
Ayanda draws 200 bottles across all the pallets, not the top layer of the nearest one.
Step 5: Count and decide
She finds 3 misplaced labels. 3 is 5 or fewer, so the lot is accepted for label placement.
Step 6: Keep critical defects out of the count
If any bottle had no date code, that is a critical defect. Ayanda would stop, hold the lot and call QA whatever the major count.
Step 7: Record it
Lot, n, Ac/Re, defects by class, the decision and her signature go on the sampling record.
What did the pass tell Priya? The probability of acceptance (Pa) for this plan, by binomial calculation:
| True defect rate in the lot | Pa (n = 200, Ac 5) |
|---|---|
| 1 % | 0.98 |
| 2 % | 0.79 |
| 3 % | 0.44 |
| 5 % | 0.06 |
Setting up plans for your own intake? Try the free AQL Sampling Plan Calculator, then learn to choose and defend plans in Course 2: R2 950, about 20 hours, with a calculator allowed in the final exam.
Why does a lot at twice the AQL pass 79 % of the time?
Because a sample of 200 cannot reliably tell 1 % from 2 % defective. A lot that is truly 2 % defective gives about four defectives in a sample of 200, inside the acceptance number of 5, so this plan accepts it about 79 % of the time. The scheme is designed to protect the producer at the AQL.
Read the table the other way round. At 3 % the lot passes 44 % of the time; only at 5 % does the plan catch it reliably, passing just 6 % of such lots. “Passed at AQL 1.0” therefore does not mean “no more than 1 % defective”.
What protects the customer is everything around the single decision: switching rules that tighten inspection after rejections, supplier records, control charts on your own lines and, for a one-off lot such as film from a new converter, ISO 2859-2, whose plans target the limiting quality you want to reject.
Why do critical defects never get an AQL?
Because even a zero-acceptance plan lets rare defects through. With n = 200 and c = 0, a lot with 0.5 % critical defects is still accepted about 37 % of the time, which in 5,000 bottles means 25 bad ones reaching the customer. Critical defects are controlled by prevention and 100 % automated inspection with challenge tests.
| Critical defect rate | Pa with n = 200, c = 0 |
|---|---|
| 0.1 % | 0.82 |
| 0.5 % | 0.37 |
| 1 % | 0.13 |
So metal, glass, a wrong allergen label, a missing date code and a wrong artwork never get an AQL. They are controlled by metal detectors and X-ray with challenge tests and code and label checks at every changeover, and one found in any sample stops the lot. The Codex General Guidelines on Sampling, CXG 50-2004 (amended 2023), are the international reference for sampling food lots.
What does an auditor actually ask for?
An auditor asks you to justify the plan, not just show it: why this AQL for this defect class, which inspection level, which edition of the standard, and where the switching records are. Then they look at what happened to rejected lots. Resampling a failed lot until it passes is the finding they are looking for.
- Show me the sampling plan for this material, and the edition of ISO 2859-1 your procedure cites.
- Why AQL 1.0 for majors, and who agreed it: your customer, your supplier or you?
- Show me how critical defects are controlled, and confirm none of them has an AQL.
- Show me the switching record for this supplier: normal, tightened, reduced and, if you use it, skip-lot.
- Show me a rejected lot: was it sorted 100 % or returned, and where is the supplier non-conformance report?
- How do you know the sample was random across the whole lot?
Frequently asked questions
What sample size do I need for AQL 1.0 on a lot of 5,000?
On the ISO 2859-1:1999 master tables, general inspection level II gives code letter L and a sample of 200, with Ac 5 and Re 6 at AQL 1.0 for single sampling under normal inspection. Confirm these values against your site’s licensed 2026 copy before you rely on them.
Did the sampling tables change in ISO 2859-1:2026?
ISO says only that the sampling plan tables were modified to enhance usability. No public source confirms whether any code letter or Ac/Re value changed, so treat every number you hold as a 1999 value until you have checked it in the licensed 2026 copy.
Is ISO 2859-1 the same as ANSI/ASQ Z1.4?
They are closely related attribute sampling schemes, and many US customers cite Z1.4. A revision of ANSI/ASQ Z1.4-2003 (R2018) is planned. Cite the standard your customer’s specification names, and keep its edition current.
What AQL should we use for food packaging?
No law sets one. The AQL is set per defect class in the specification agreed with your customer or supplier, with separate values for major and minor defects. Critical defects get no AQL at all: they need prevention and 100 % automated inspection.
Can AQL sampling be used for microbiological testing?
No. Microbiological testing uses its own two-class and three-class sampling plans, and a negative result proves less than most people assume, because contamination is rarely spread evenly through a lot.
What do we do with a rejected lot?
Sort 100 % of the lot or return it, raise a supplier non-conformance report, and record the plan, the sample and the result. Never resample the same lot until it passes. Rejections also feed the switching rules; take the exact conditions from your licensed copy.
Sample with a plan you can defend
A sampling plan is only as good as the reasons behind it. Course 2 teaches you to choose it, read it, know what a pass proves and keep critical defects out of it, on a working factory case.
QC Inspector Course for Food and Beverage Manufacturing: Course 2 of the FS60 series, 8 modules, about 20 hours, self-paced.
- R2 950, no VAT charged; for learners with Course 1 or at least 12 months in food production, QC or a food lab
- Three practice tools: the AQL Sampling Plan Calculator and the Net Content and Weight Control Calculator (both free to try) and the Line QC Check Sheet Builder
- A video-proctored final exam: 40 questions from a bank of 90, 75 minutes, calculator allowed, 75 % to pass, 2 attempts
- A QR verified Certificate of Achievement an employer or auditor can check
- ASC Food Safety Training is a FoodBev SETA accredited training provider (587/00337/1900); the FS60 courses are non-credit-bearing short courses
Weights next? Read average weight rules in South Africa: SANS 458, T1 and T2 errors and the e-mark. Compare all three levels on the food quality control training page, see all courses or contact us about a team of five or more.
Last reviewed: 28 September 2026. Standard status as of September 2026.
Sources: ISO 2859-1:2026 (iso.org); Quality Magazine, “Third revision of ISO 2859-1”, 10 April 2026; Codex CXG 50-2004 (amended 2023), General Guidelines on Sampling. Probabilities of acceptance by binomial calculation.