Answers › AQL switching rules

AQL switching rules: when does inspection move between normal, tightened and reduced?

Reviewed by the JCH tooling team · Updated

Inspection starts on normal: 2 rejected lots within 5 or fewer consecutive lots switch it to tightened, and 5 consecutive accepted lots switch it back. Reduced inspection must be earned and ends at the first problem; 5 lots not accepted while on tightened stop sampling inspection.

Which text these rules come from

The triggers on this page are quoted from two texts we read clause by clause: MIL-STD-105E (10 May 1989), the military predecessor of ANSI/ASQ Z1.4 and now in the public domain (a US government document), and the first edition of ISO 2859-1 (1989), whose clause 9 appears in the free sample pages iTeh publishes. On the switching rules the two agree, in places word for word. We have not read the current paid texts. Quality Magazine calls ASQ/ANSI Z1.4-2003 (R2018) "the current active standard for sampling procedures and tables for inspection by attributes" and reports that "the final draft of ISO 2859-1 was approved in December 2025." Where a later edition is known to differ, this page says so and names the source. If your contract names an edition, its wording governs.

Three ground rules sit under every switch. First, inspection starts on normal; ISO 2859-1:1989 clause 9.1 adds "unless otherwise directed by the responsible authority." Second, each defect class switches on its own. MIL-STD-105E section 4.6 says the procedures "shall be applied to each class of defects or defectives independently," so one shipment can sit on tightened for major defects and on normal for minor ones. Third, only original inspection counts. Section 4.7.1 tells you to ignore "resubmitted lots or batches for this procedure," so a lot that fails, gets sorted and passes on re-inspection is still a rejection in the switching record.

The four switches and the stop rule

MoveTrigger (MIL-STD-105E and ISO 2859-1:1989)Clause
Normal to tightened2 lots not accepted out of 5 or fewer consecutive lots, on original inspection105E 4.7.1; ISO 9.3.1
Tightened to normal5 consecutive lots accepted on original inspection105E 4.7.2; ISO 9.3.2
Normal to reducedAll four at once: the preceding 10 lots (or more, per the note to Table VIII) were on normal and all were accepted; total defectives in those 10 samples at or under the limit number in Table VIII; production at a steady rate; reduced inspection considered desirable (ISO adds "by the responsible authority"). ISO 2859-1:1999 uses a switching score instead, see below105E 4.7.3; ISO 9.3.3
Reduced to normalAny one of: a lot not accepted; a lot accepted with a count above Ac but below Re; production irregular or delayed; other conditions that warrant it105E 4.7.4; ISO 9.3.4
Discontinue inspectionThe cumulative number of lots not accepted on original tightened inspection reaches 5105E 4.8; ISO 9.4

Notice the asymmetry. Two bad lots put a supplier on tightened; five good lots in a row are needed to get off it. Reduced inspection takes ten clean lots to earn and one doubtful lot to lose. The system reacts quickly to bad news and slowly to good news, which is the point: it protects the buyer from a process that is slipping, while one unlucky sample on its own never moves a supplier off normal.

State diagram of the switching rules: inspection starts on normal; 2 lots not accepted in 5 or fewer consecutive lots moves it to tightened; 5 consecutive accepted lots on tightened moves it back to normal; reduced inspection is earned from normal and lost on the first doubtful lot; 5 lots not accepted while on tightened discontinues inspection until corrective action, after which it restarts on tightened Switching between severities (MIL-STD-105E 4.6-4.8; ISO 2859-1:1989 clause 9) start REDUCEDsmaller sample NORMALdefault plan TIGHTENEDlower accept number DISCONTINUEDfix, restart on tightened 2 not accepted in5 or fewerconsecutive lots 5 consecutivelots accepted 10+ lots accepted,limit number met,steady production,approved 1 lot not accepted,count between Ac and Re,or irregular production 5 not accepted(cumulative)on tightened aftera fix Triggers count original inspection only and run separately for each defect class. ISO 2859-1:1999 replaces the 10-lot test with a switching score of at least 30 (secondary sources).

What each severity does to one plan

Take a common plan, sample size code letter L at an AQL of 2.5 with single sampling, and read it from the three master tables in MIL-STD-105E. We quote only these cells; the full tables are in the standard.

SeveritySample sizeAccept the lot onReject the lot onTable
Normal20010 or fewer defectives11 or moreII-A
Tightened2008 or fewer9 or moreII-B
Reduced805 or fewer8 or moreII-C

Tightened keeps the 200-piece sample and cuts the acceptance number from 10 to 8. Reduced cuts the sample to 80 and leaves a gap: 6 or 7 defectives is neither an accept nor a reject. The footnote to Table II-C settles it: "accept the lot, but reinstate normal inspection." That gap is the second trigger in the reduced-to-normal rule, and it is why reduced inspection is safer than its small sample suggests.

True % defective in the lotNormal: chance of acceptanceTightened: chance of acceptanceReduced: accepted, stays on reducedReduced: accepted, but back to normal
1%over 99.9%over 99.9%over 99.9%under 0.1%
1.5%over 99.9%99.6%99.9%0.1%
2.5%98.7%93.4%98.5%1.4%
4%82.0%59.3%89.9%8.6%
5%58.3%32.7%78.9%16.4%
6.5%24.3%9.2%58.0%27.2%

These are our binomial calculations for a random sample from a large lot, not figures printed in the standard. Two things stand out. Tightened inspection bites where it should: a lot that is 4% defective, well above the AQL of 2.5, is accepted 82% of the time on normal but only 59% on tightened. And reduced inspection on its own is a weak filter: at 5% defective it accepts 79% of lots outright. It stays safe because another 16% of those lots land in the 6-or-7 band and send the supplier back to normal, and any rejection does the same. For how a single plan's acceptance curve is read, see the AQL sampling note.

A worked example: twelve lots through the rules

An illustrative lot log, with made-up counts rather than data from any supplier, using the code letter L plan at AQL 2.5 above. Two counters matter on tightened: the run of consecutive accepted lots, which earns the return to normal, and the total of lots not accepted since tightened began, which triggers discontinuation.

LotSeverityDefectives in sampleResultWhat the rules do
1Normal4AcceptNothing
2Normal12RejectFirst rejection. If the lot is sorted and resubmitted, the re-inspection does not count
3Normal6AcceptNothing
4Normal11RejectSecond rejection within 3 consecutive lots: tightened from lot 5
5Tightened7AcceptRun of accepts: 1. Not accepted on tightened: 0
6Tightened9RejectRun resets to 0. Not accepted on tightened: 1
7Tightened5AcceptRun: 1
8Tightened3AcceptRun: 2
9Tightened8AcceptRun: 3. Eight is the tightened limit; nine would reject
10Tightened4AcceptRun: 4
11Tightened2AcceptRun: 5, so normal from lot 12
12Normal9AcceptPasses on normal; tightened would have rejected it

Three details that people often get wrong show up in this log. Lot 4 triggers tightened although lots 2 and 4 are not back to back, because the rule is 2 out of 2, 3, 4 or 5 consecutive lots. Lot 6's rejection wipes out the run of accepts, so it is lot 11, not lot 10, that earns the return. And lot 9 passes on tightened with exactly 8 defectives, while lot 12 passes on normal with 9, a count that would have failed one lot earlier on tightened.

Now change the story. Suppose lots 6, 8, 10, 12 and 14 had all been rejected on tightened, with accepted lots in between. No two rejections are adjacent and the supplier never strings five accepts together, yet the fifth rejection stops acceptance inspection. Section 4.8 counts "the cumulative number of lots not accepted in a sequence of consecutive lots on original tightened inspection," which means every lot not accepted since tightened began, not five in a row. Some online summaries word it differently, as five lots spent on tightened or five lots in a row; the primary texts count lots not accepted.

What the rules do to a supplier's odds

Using the same three plans, we calculated two things for a supplier whose process runs at a steady defect rate: the chance that a block of 5 lots on normal contains 2 or more rejections, which is enough to trigger tightened, and, once on tightened, the chance of reaching 5 straight accepts before 5 lots have been rejected.

Process % defective5 normal lots include 2+ rejectionsTightened ends with a return to normalAverage lots spent on tightened
2.5% (at the AQL)0.15%99.8%6.1
3%1.5%94.7%7.9
4%22%32%9.8
5%69%2%7.3

A supplier running at the AQL almost never gets tightened, and if it does, it is back on normal in about six lots. A supplier running at 4% passes 82% of individual lots on normal, which looks tolerable lot by lot. Under switching, it lands on tightened in roughly one five-lot block in five, and once there it is about twice as likely to be stopped as to recover. A fixed normal plan gives no such protection. It is the switching system, not any one sample, that screens out a slipping process. These numbers assume a constant defect rate, random samples from large lots and lots that are independent of each other. Real processes drift, so read them as orders of magnitude, not forecasts.

Earning reduced inspection: ten lots, or a switching score

In MIL-STD-105E and ISO 2859-1:1989, reduced inspection needs four conditions at once: the preceding 10 lots (or more, per the note to Table VIII) on normal and all accepted on original inspection; the total defectives found in those 10 samples at or under the limit number in Table VIII; production at a steady rate; and reduced inspection judged desirable, a decision ISO gives to "the responsible authority." We do not reproduce the limit numbers here; take them from the standard.

The 1999 edition of ISO 2859-1 replaced the ten-lot test with a switching score. We have not read that text, so this description comes from secondary sources that agree with each other. SQC Online's ISO 2859-1 chart lists "Switching score value is at least 30." A 2008 post on the Elsmar Cove forum, quoting a third-party summary, says the score is set to zero at the start of normal inspection. With single sampling and an acceptance number of 2 or more, you "add 3 to the switching score if the lot would have been accepted if the AQL had been one step tighter". With an acceptance number of 0 or 1, you "add 2 to the switching score if the lot is accepted". Otherwise the score goes back to zero. QualityIris gives the same 30-point threshold together with steady production.

Applied to the example plan: the next AQL step tighter than 2.5 is 1.5, and MIL-STD-105E's normal table gives code letter L at AQL 1.5 an acceptance number of 7. So each lot with 7 or fewer defectives in its 200-piece sample would add 3 points, and a lot with 8, 9 or 10 would be accepted but would reset the score to zero. Ten such lots in a row reach 30. By our arithmetic, a process running at 1% defective gets there in ten lots about 99% of the time; a process sitting right at the AQL of 2.5% does so only about 25% of the time. Reduced inspection is meant for suppliers running well inside the AQL, not at it.

We have not seen the ANSI/ASQ Z1.4-2003 clause text, so confirm which test applies against the edition your contract names. SQC Online's MIL-STD-105E and Z1.4 chart describes the ten-lot test; OpenExamPrep describes the ten-lot test and mentions a switching score only in passing for updated revisions. Check the edition your contract names before anyone switches to reduced.

When inspection stops: 5 lots, not 10

Both primary texts set the stop at five. ISO 2859-1:1989 clause 9.4 reads: "If the cumulative number of lots not accepted in a sequence of consecutive lots on original tightened inspection reaches 5, the acceptance procedures of this part of ISO 2859 shall be discontinued." MIL-STD-105E section 4.8 says the same in nearly the same words. Inspection resumes only after corrective action. ISO puts that action on the supplier and adds that "the responsible authority shall agree that this action is likely to be effective." Inspection then restarts on tightened, as if the switch from normal had just been made.

Some study material gives ten. OpenExamPrep's Z1.4 guide says inspection must be discontinued "If 10 consecutive lots remain on Tightened inspection without qualifying to return to Normal." That is a different test (ten lots spent on tightened, rather than five lots not accepted), and we have not found it in any standard text. SQC Online's ISO chart shows 5 lots not accepted while on tightened; its MIL-STD-105E/Z1.4 chart says 5 consecutive lots remaining on tightened, a looser wording. Quote five with the MIL-STD-105E and ISO 2859-1:1989 citations, and confirm it against the edition your contract names. The standards also leave open what happens to shipments while inspection is stopped, so the quality agreement has to say.

What ISO 2859-1's third edition changes

John Vandenbemden's April 2026 column in Quality Magazine lists the key changes in the third edition: "The incorporation of skip-lot sampling inspection, which adds another step to the switching rules"; the removal of operating characteristic curves for each plan, with methods for building them moved to Annex E; a note that "supplier" in the revised switching rule table covers internal suppliers; and tables modified for usability. He also wrote that the ISO editors had introduced errors "that need to be corrected before publication," and that ASQ/ANSI Z1.4-2003 (R2018) is to be revised once ISO 2859-1 is published. We have not read the third edition's clause text. On the evidence of this column, skip-lot is the switching change to look for. Whether the normal, tightened, reduced and discontinuation triggers were reworded is not settled.

Do your orders need switching rules?

MIL-STD-105E section 4.11 says its plans "were designed for use where the units of product are produced in a continuing series of lots or batches over a period of time." For a lot of an isolated nature it points the buyer to limiting quality protection instead. In practice, many importers never switch at all; Sofeast's QualityInspection.org, a third-party inspection firm, writes that very few buyers follow the rules. Without them, a Z1.4 plan is a fixed normal plan, and the odds table above shows what that costs once a supplier drifts above the AQL. If a defect class has to be held at zero accepted, a separately sized c=0 sampling plan is a different tool with no switching rules of its own.

Clear molded plastic housing sliding off a polished steel mold core on an open injection molding press
A clear molded housing coming off the core, one part in a continuing run from the same mold. Switching rules assume a series of lots like this from one process; a one-off shipment is judged on its own plan.

For a repeat program, these points belong in the quality agreement:

To get the normal plan the log starts from, use the AQL calculator: enter lot size, inspection level and AQL, and it returns the sample size and the accept and reject numbers.

Common questions

Does a resubmitted lot count toward tightened inspection?

No. MIL-STD-105E section 4.7.1 and ISO 2859-1:1989 clause 9.3.1 count lots not accepted on original inspection, "ignoring resubmitted lots or batches for this procedure." A lot that fails, is sorted and then passes on re-inspection still counts as a rejection for switching.

Is it 5 or 10 lots before inspection is discontinued?

Five, in MIL-STD-105E section 4.8 and ISO 2859-1:1989 clause 9.4: acceptance inspection stops when the cumulative number of lots not accepted on original tightened inspection reaches 5. Some exam-prep material says 10; we have not found that figure in a standard text. We have not seen the ANSI/ASQ Z1.4-2003 clause text, so check the edition your contract names.

Do critical, major and minor defects switch together?

No. MIL-STD-105E section 4.6 and ISO 2859-1:1989 clause 9.2 apply the switching procedures to each class independently. One shipment can be inspected on tightened for major defects and on normal for minor ones.

Can a supplier start on reduced inspection?

Not by default under the 1989 texts. Both start on normal, and ISO 2859-1:1989 clause 9.1 allows otherwise only when "directed by the responsible authority." Reduced inspection is earned: 10 accepted lots plus the other conditions in the 1989 texts, or a switching score of at least 30 in ISO 2859-1:1999, according to secondary sources.

What happens when the count falls between Ac and Re on reduced inspection?

The lot is accepted, but normal inspection is reinstated. That is the footnote to MIL-STD-105E Table II-C and the second trigger in section 4.7.4. For code letter L at AQL 2.5, reduced inspection accepts on 5 and rejects on 8, so 6 or 7 defectives lands in that band.

Does the JCH AQL calculator give tightened or reduced plans?

No. It gives the normal single sampling plan, which is where inspection starts. Use it for the sample size and the accept and reject numbers, then apply the switching rules on this page to your lot history, with the tables from the edition your contract names.

Sources

Setting up inspection for a repeat program? Send the drawing and your defect list, and we'll go through the defect classes, sampling plan and switching rules with you before production starts. NDA first if you want one.

Send a part for reviewBook a 30-minute call

Ask JCH about AQL switching rules

Have a question about AQL switching rules? Ask our AI assistant. It answers from this website, and our tooling team follows up if you want.

AI assistant. Answers can be wrong; nothing it says is a quote or a promise.