Guidelines on Decision rules and

Conformity with Requirements

Draft Ver.JW4 ILAC-G8:20XX

 © Copyright ILAC 20XX
ILAC publications may not be copied for sale by any individual or
body other than ILAC member organisations
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

1
2 TABLE OF CONTENTS page
3
4 Preamble 4
5 Purpose 4
6 Authorship 4
7 Definitions 5
8 Decision Rules and Statements of conformity in ISO/IEC 17025:2017 6
9 Measurement Decision Risk Overview 7
10 Guard Bands and Decision Rules 7
11 Taking Measurement Uncertainty into Account 10
12 Decision Rule Flow Chart 13
13 Decision Rule Documentation and Application 15
14 Summary 16
15 References 17
16 Annex A – Sample Checklist for meeting the requirements 18
17 Annex B – Examples of Documented Agreements 19
18

Page 3 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

19 PREAMBLE

20
21 This guidance document has been prepared to assist laboratories in the use of decision rules
22 when declaring statements of conformity to a specification or standard as required by ISO/IEC
23 17025:2017 [1].
24
25
26 Since ISO/IEC 17025 was first published in 1999, the need for statements of conformity with
27 specifications or standards has developed greatly, together with documents on the concept of
28 decision rules used to make such statements.
29
30
31 The revised ISO/IEC 17025:2017 recognizes that no single decision rule can cover all situations
32 applicable to statements of conformity. Further, a survey conducted by ILAC demonstrated that
33 statements of conformity in the regulated area usually did not adopt the decision rules found in
34 ILAC G8:2009 but instead to other rules.
35 This document provides:
36
37 a) overall guidance on how to select appropriate decision rules; and

38 b) guidance for required elements of a decision rule if none of the standard published rules
39 apply.

40
41 NOTE: Where further information is required regarding the mathematics of various decision
42 rules covered in this document the reader is referred to the JCGM 106:2012 [2].
43
44 PURPOSE
45
46 This document aims to provide an overview for assessors, laboratories, regulators and customers
47 concerning decision rules and conformity with requirements. It does not enter into the details
48 regarding underlying statistics and mathematics but refers readers to the relevant literature. This
49 means that some laboratories, their personnel and their customers will be required to improve
50 their knowledge related to decision rule risks and associated statistics. For cases where
51 legislation mandates certain decision rules to be applied, those must be followed by laboratories.
52
53 It must also be noted that there is a difference between overall “Laboratory Risk” and the “Risk”
54 which is associated with a decision rule (Measurement Decision Risk in this case). The latter is
55 directly in the control of recipients of statements of conformity as it is they who specify the
56 decision rules to apply by laboratories. Accordingly, it is the recipient who takes the risk
57 associated with statements, that is, false acceptance or rejection of results.
58
59 AUTHORSHIP
60
61 These guidelines were prepared by the Accreditation Issues Committee of ILAC with significant
62 support and help from members of the ILAC Laboratory Committee.
63
64
65
66

Page 4 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

67 1. DEFINITIONS
68 For this document the JCGM 106:2012 Evaluation of measurement data - The role of
69 measurement uncertainty in conformity assessment is taken as the primary reference.
70 There are however more documents that provide information on how to decide on
71 conformity with requirements. (See paragraph 9 for reference documents).
72
73 1.1 Tolerance limit (TL) (specification limit)
74 specified upper or lower bound of permissible values of a property
75
76 1.2 Tolerance interval
77 interval of permissible values of a property
78
79 1.3 Measured quantity value
80 quantity value representing a measured result in VIM 3.2.10 [6]
81
82 1.4 Acceptance limit (AL)
83 specified upper or lower bound of permissible measured quantity values
84
85 1.5 Acceptance interval
86 interval of permissible measured quantity values
87 NOTE 1 Unless otherwise stated in the specification, the acceptance limits belong to the
88 acceptance interval.
89 NOTE 2 An acceptance interval is called an “acceptance zone” in ASME B89.7.3.1 [3].
90
91 1.6 Rejection interval
92 interval of non-permissible measured quantity values
93 NOTE 1 A rejection interval is called an “rejection zone” in ASME B89.7.3.1 [3].
94
95 1.7 Guard band (w)
96 interval between a tolerance limit and a corresponding acceptance limit where w = TL-AL
97
98 1.8 Decision rule
99 documented rule that describes how measurement uncertainty will be accounted for when
100 accepting or rejecting an item, given a specified requirement and the result of a
101 measurement. JCGM 106:2012 [2]
102
103 rule that describes how measurement uncertainty is accounted for when stating
104 conformity with a specified requirement in ISO/IEC 17025:2017 3.7 [1]
105
106 1.9 Simple Acceptance
107 a decision rule in which the acceptance limit is the same as the Tolerance, i.e. AL=TL [3]
108 ASME B89.7.3.1 [3].
109
110 1.10 Indication
111 quantity provided by a measuring instrument or measuring system
112 NOTE 1 An indication is often given as the position of a pointer for an analogue output or
113 the displayed or printed number for a digital output.
114 NOTE 2 An indication is also known as a reading
115
116 1.11 Maximum permissible error (MPE) (of indication)
117 for a measuring instrument, maximum difference, permitted by specifications or
118 regulations, between the instrument indication and the quantity being measured

Page 5 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

119
120 1.12 Test Uncertainty Ratio (TUR)
121 the ratio of the span of the tolerance of a measurement quantity subject to calibration, to
122 twice the 95% expanded uncertainty of the measurement process used for calibration. [7]
123
124 1.13 Expanded Uncertainty of Measurement (U)
125 U is interpreted as defining an interval about the measurement result that encompasses a
126 large fraction p of the probability distribution characterized by that result and its
127 combined standard uncertainty, and p is the coverage probability or level of confidence of
128 the interval. GUM 6.2.2 [4].
129 For this document U should be taken to be the expanded uncertainty which corresponds to
130 a coverage probability of approximately 95%.
131
132 1.14 Specific risk
133 is the probability that an accepted item is non-conforming, or that a rejected item does
134 conform. This risk is based on measurements of a single item.
135
136 1.15 Global risk
137 is the average probability that an accepted item is non-conforming, or that a rejected item
138 does conform. It does not directly address the probability of false accept to any single
139 item, discrete measurement result or individual workpiece
140
141 2. DECISION RULES AND STATEMENTS OF CONFORMITY IN ISO/IEC
142 17025:2017
143
144 ISO/IEC 17025:2017 includes criteria related to decision rules and conformity with
145 requirements in resources and processes related to personnel, contract review and
146 reporting as described below.
147
148 2.1 Clause 3.7: a decision rule is defined as “a rule that describes how measurement
149 uncertainty will be accounted for when stating conformity with a specified
150 requirement”
151
152 2.2 Clause 6.2.6 requires that the laboratory shall authorize personnel to “analyse results,
153 including statements of conformity or opinions and interpretations”.
154
155 2.3 Clause 7.1.3 requires that “When the customer requests a statement of conformity to a
156 specification or standard for the test or calibration (e.g. pass/fail, in-tolerance/out-of-
157 tolerance) the specification or standard, and the decision rule shall be clearly
158 defined. Unless inherent in the requested specification or standard, the decision rule
159 selected shall be communicated to the customer.”
160
161 2.4 Clause 7.8.3.1b) states “where relevant, a statement of conformity with requirements
162 or specifications” and clause 7.8.3.1c) states “where applicable, the measurement
163 uncertainty presented in the same unit as that of the measurand or in a term relative
164 to the measurand (e.g. percent), when it is relevant to the validity or application of
165 the test results, when a customer's instruction so requires, or when the measurement
166 uncertainty affects conformity to a specification limit”.
167
168 2.5 Clause 7.8.4.1a) states “the measurement uncertainty of the measurement result
169 presented in the same unit as that of the measurand or in a term relative to the
170 measurand (e.g. percent).”

Page 6 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

171 Clause 7.8.4.1e) also states “where relevant, a statement of conformity with
172 requirements or specifications”
173
174 2.6 Clause 7.8.6.1 states “When a statement of conformity to a specification or standard
175 for test or calibration is provided, the laboratory shall document the decision rule
176 employed, taking into account the level of risk (such as false accept and false reject
177 and statistical assumptions) associated with the decision rule employed and apply the
178 decision rule.”
179
180 2.7 Clause 7.8.6.2 requires that “the laboratory shall report on the statement of
181 conformity such that the statement clearly identifies:
182 a) to which results the statement applies; and
183 b) which specifications, standard or parts thereof are met or not met;
184 c) the decision rule applied (unless it is inherent in the requested specification or
185 standard).”
186
187 3. MEASUREMENT DECISION RISK OVERVIEW
188
189 When performing a measurement and subsequently making a statement of conformity, for
190 example, in or out-of-tolerance to manufacturer’s specifications or Pass/Fail to a
191 particular requirement, there are two possible outcomes:
192 a. A correct decision is made regarding conformance to specification
193 b. An incorrect decision is made regarding conformance to specification

194 Each measurement has an associated uncertainty. Figure 1 shows two identical
195 measurements but with different measurement uncertainties [3]. The expanded
196 uncertainty in the lower result (case A) lies entirely within the tolerance limit. The upper
197 result (case B) has significantly larger measurement uncertainty. The risk of falsely
198 accepting a result in case B is higher due to the larger measurement uncertainty. (i.e. See
199 “What % risk is this?” in figure 1)
200

201
202 Figure 1. Illustration of Measurement Decision Risk

203
204 4. GUARD BANDS AND DECISION RULES
205
206 4.1 Guard Bands
207
208 The use of guard bands provides a way of limiting the probability of making an incorrect
209 conformance decision. It is basically a safety factor built into the measurement decision
210 process by reducing the specification limits to a desired acceptable limit. This is often
211 done to cater for measurement uncertainty as is described later in this section.
212

Page 7 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

213 This guidance document refers to Guard Bands where the Guard Band (w) is the
214 Tolerance Limit (TL) minus the Acceptance Limit (AL) or w=TL–AL. This means that if
215 the measurement result is below the Acceptance Limit (AL), then the measurement is
216 accepted as conforming to specification.
217

218
219 Figure 2 Graphical representation of a Guard Band

220 With guard band terminology there are usually upper and lower limits for a tolerance.
221 This document generally deals with an upper tolerance limit for simplicity so the user
222 needs to include the lower limits as well.
223
224 A guard band which is equal to zero, w = 0, infers that acceptance simply is when a
225 measurement result is below a tolerance limit. This is called simple acceptance. Simple
226 acceptance is also called “shared risk” because the probability to be outside the tolerance
227 limit may be as high as 50 % in the case when a measurement result is exactly on the
228 tolerance limit (assuming a symmetric normal distribution of the measurements).
229
230 4.2 Decision Rules
231 A decision rule can either be binary or non-binary. This means either “Pass or Fail” only
232 for binary conditions and “Pass or Fail and some “conditional” terms for non-binary.
233 These are further explained below.
234
235 4.2.1 Binary statement for simple acceptance rule (w=0)
236
237 Measurement results are reported as:
238  Pass - acceptance based on simple acceptance; the measurement result being below
239 the acceptance limit, AL = TL.
240  Fail - rejection based on if the measurement result is above acceptance limit AL =
241 TL
242

Page 8 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

243
244 Figure 3 Graphical representation of a Binary - Simple Acceptance

245
246 4.2.2 Binary statement with guard band
247
248 Measurement results are reported as:
249
250  Pass - acceptance based on guard band; the measurement result being below the
251 acceptance limit, AL = TL – w.
252  Fail - rejection based on guard band; if the measurement result is above acceptance
253 limit AL = TL – w
254

255
256 Figure 4 Graphical representation of a Binary statement with a guard band

257
258
259
260 4.2.3 Non-binary statement with guard band
261
262 Measurement results are reported as:
263  Pass; the measurement result being below the acceptance limit, AL = TL – w.
264

Page 9 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

265  Conditional Pass; the measurement result being inside the guard band and below the
266 tolerance limit, in the interval [TL – w; TL].
267
268  Conditional Fail; the measurement result being above the tolerance limit but below
269 the tolerance limit added the guard band, in the interval [TL; TL + w].
270
271  Fail; the measurement result being above the tolerance limit added the guard band, TL
272 + w.
273

274
275 Figure 5 Graphical representation of a non-Binary statement with a guard band

276
277 It should be considered that a measurement may result in a decision on conformity
278 (acceptance) using one guard band and rejection if a larger guard band is used. Therefore
279 conformity with a requirement is inherently connected to the decision rule employed. It is
280 therefore expected that the decision rule is agreed before the measurements are taken.
281
282
283 5. TAKING MEASUREMENT UNCERTAINTY INTO ACCOUNT
284
285 5.1 Measurement uncertainty taken indirectly into account
286
287 If measurement uncertainty is taken directly into account, the acceptance interval will be a
288 restricted part of the tolerance as described above. Often regulators do not want such
289 restriction for many reasons. One is that the larger the measurement uncertainty is, the
290 smaller the acceptance interval gets. If the measurement results in this case do not have a
291 bias, then acceptance will result in a narrower distribution of accepted results than if
292 measurement uncertainty had been smaller. See figure 6.
293

Page 10 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

A) Small relative expanded uncertainty U =T/10 and w=U

Acceptance interval AL TL

B) Large relative expanded uncertainty U=(T/2) and w=U

Acceptance interval AL TL
294
295 Figure 6 Acceptance interval for a case where expanded uncertainty is small compared to tolerance
296 A) and large B) for the same tolerance limit TL. A large guard band narrows the distribution function
297 of accepted items.

298
299 In order to avoid dependency on guard bands among laboratories, regulators often take
300 measurement uncertainty indirectly into account. This may be done in numerous ways
301 depending on the area of testing or calibration. Some examples are:
302
303  OIML R76-1:2006 (NAWIs) cl. 3.7.1 where it is required that “…the standard
304 masses used for the type examination or verification of an instrument …. shall
305 not have an error greater than 1/3 of the MPE. If they belong to class E2 or
306 better, their uncertainty is allowed to be not greater than 1/3 of the MPE of the
307 instrument (the tolerance)”

308  OIML R117-1:2007 Dynamic measuring systems for liquids other than water
309 Part 1: Metrological and technical requirements A.2 Uncertainties of
310 measurement: When a test is conducted, the expanded uncertainty of the
311 determination of errors on indications of volume or mass shall be less than
312 one-fifth of the maximum permissible error (MPE) (the tolerance)

313  WADA Technical Document – TD2014DL The decision limit DL shall be

314 calculated as the sum of the value T and the guard band (g), where (g) is
315 calculated based on the relevant WADA maximum acceptable value of the
316 combined standard uncertainty (uc Max)

317 DL = T + g, and g = k x u_c Max, with k = 1.645

318
319 In most cases, u_c Max is assigned using data from the combined participant
320 results obtained from relevant rounds of the External Quality Assessment
321 Scheme (EQAS). This corresponds to a negative guard band w that is fixed for
322 all laboratories no matter their own measurement uncertainty.

323  Cases where the conviction of a person is based on the passing of a special
324 tolerance e.g. for speeding or drunk driving, where the acceptance limit is
325 defined by the applicable local legislation, to be the tolerance limit with some
326 safety contribution added to it. This is to ensure that the probability of false
327 convictions is very small. This corresponds to a negative value for the guard
328 band, w < 0 in the sense that acceptance (e.g. not speeding and not drunk

Page 11 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

329 driving) is if you are below the acceptance limit which happens to be above the
330 tolerance limit.

331  Cases where test standards have taken typical uncertainty into account when
332 setting the tolerance limits and the acceptance limit then equals the tolerance
333 limit.

334  Cases where a customer specifies a guard band to be used for deciding on
335 conformity with specification. Such guard bands may be fixed, but may also be
336 based on the measurement uncertainty which is detailed below.
337
338 As can be seen from the bullet list, decision rules can be not only very different but also
339 very complicated.
340
341 5.2 Measurement uncertainty taken directly into account
342
343 ISO/IEC 17025:2017 requires that laboratories evaluate measurement uncertainty and that
344 it be taken into account when making statements of conformity.
345
346 As mentioned previously, the approach adopted may significantly vary dependent on the
347 situation and different guard bands may be applied.
348
349 Often the guard band w=rU is based on a multiple r of the expanded measurement
350 uncertainty U. For a binary decision rule, a measured value below the acceptance limit
351 AL = TL – w is accepted.
352
353 While it is common to use a guard band w = U, there may be cases where a multiplier
354 other than 1 is more appropriate. The table below provides examples of different guard
355 bands to achieve certain levels of specific risk, based on the customer application.
356
357
Decision rule Guard % Specific Risk
band w
6 sigma 3U <1 ppm PFA.
3 sigma 1,5 U < 0.16% PFA.
ILAC G8:2009 rule 1U < 2.5% PFA.
ISO 14253-1:2017 [5] 0,83 U < 5 % PFA.
Simple acceptance 0 < 50 % PFA
Uncritical -U Item rejected for measured value greater than AL = TL+ U
< 2.5% PFR
Customer defined rU Customers may define arbitrary multiple of r to have applied
as guard band.
358 (PFA – Probability of False Accept and PFR – Probability of False Reject)
359
360
361 5.3 Specific vs. global [average] risk in calibration
362
363 Specific consumer risk is “the probability that a particular accepted item is non-
364 conforming” (See 3.3.13 of JCGM 106 [2]). Specific producer risk is “probability that a
365 particular rejected item is conforming.” (See 3.3.14 of JCGM 106 [2]).
366

Page 12 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

367 If the laboratory only measures a single instrument and has no history of calibration
368 results for that serial number, or if it has no information on the behaviour of that model as
369 a population, that can be considered to be a situation with “meagre prior information” (see
370 7.2.2 of [2]). Some take the view that when a laboratory receives an instrument for
371 calibration (and subsequent verification to manufacturer’s tolerance) with meagre prior
372 information, that the laboratory can only provide specific risks.
373
374 Some customers take steps to actively reduce the probability that instruments submitted
375 for calibration and verification will be returned “Failed”. They do so by operating a
376 “Calibration system” (See 5.3.4 of Z540.3 [7]) such that calibration records (measurement
377 reliability) are monitored by model number and serial number and calibration intervals are
378 actively managed to achieve a desired target reliability (See 5.4.1 of Z540.3 [7]), where
379 target reliability refers to the percentage of instruments that “Pass” calibration. The end
380 result is a process by which the instrument submitted is part of a customer device
381 population. If that process “rarely results in an instrument whose property of interest is
382 near the tolerance limits, there is less opportunity for incorrect decisions to be made” (See
383 9.1.4 of JCGM 106 [2]).
384
385 Thus, average probability of false accept and false reject (global risk) can be applied by
386 evaluation of the joint probability density consisting of customer managed device
387 population and laboratory managed calibration process uncertainty (See equations 17 and
388 19 of JCGM 106 [2]). Several references, [8] and [9] provide simple techniques for
389 estimating global risk.
390
391 When a customer actively manages calibration intervals as mentioned here, during
392 contract negotiation with laboratories for services compliant with the new ISO/IEC
393 17025:2017, they can direct the laboratory to use the average global risk associated with
394 decision rules when reporting results per clause 7.8.2.2. As already clarified in definition
395 1.12 an instrument passing a global risk criteria e.g. 2 % probability for false acceptance
396 (2 % PFA) may not pass a specific risk with a guard band equal to the expanded
397 uncertainty and may have a specific risk for false acceptance that can be as high as close
398 to 50 %. This is similar to the criteria for approval of instruments mostly utilized in legal
399 metrology. Generally the output from decision rules based on OIML principles (e.g. TUR
400 > 3:1 or 5:1) and global risk with approximately 2 % PFA may provide the same results in
401 terms of number of falsely rejected instruments.
402
403 5.4 Consideration of both false accept and false reject risks
404
405 “Binary decision rules, acting to reduce the consumer’s risk, will always increase the
406 producers risk”i(page 31 of [2]). This statement applies to any decision rule which applies
407 a guard band to improve or set a minimum false accept risk.
408
409 Initially, a customer submitting an item for calibration or testing to a laboratory may only
410 care about their “consumer false accept risk”. However, when a laboratory returns an
411 item as “Failed”, the customer will need to investigate the impact on the products their
412 organization produces which can often lead to expensive recalls.
413
414 6. DECISION RULE SELECTION FLOW CHART
415
416 Where choices of decision rules are available, customers and laboratories will need to
417 discuss levels of risk regarding the probability of false acceptance and false rejects

Page 13 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

418 associated with available decision rules. There is no one single decision rule that covers
419 the many testing and calibration cases involving ISO/IEC 17025:2017.
420
421 Some disciplines / industries have determined decision rules appropriate for their use and
422 published them in a specification or requirement or they are provided by
423 regulators/authorities.
424
425 Figure 7 provides general guidance for the selection of decision rules.
426
427 Here are suggestions on how to use the flow chart:
428
429 a. Some calibration or testing applications do not require a statement of
430 conformance to a metrological specification. Examples may include some
431 precision masses, efficiency of power sensors, etc. In these cases, you should
432 report the measured result and the [GUM] uncertainty.
433
434 b. If the measurement results are governed by legal or regulatory standards or
435 rules, then use the decision rule as prescribed in the appropriate standard.
436
437 c. The next scenario to consider is if your application already has measurement
438 decision rules governed by a published standard. (Examples, ISO 8655, ISO
439 6508). Generally, in these cases, standard test methods are prescribed and
440 often compliance limits already have a guard band built in to the limit, so any
441 further guard banding to limit risk is not necessary.
442
443 d. If you reach decision box 4, it generally means that no specific published
444 decision rule already governs your application. Laboratories and customers
445 may select from standard decision rules shown or choose to self-document
446 their own rule. (See appendix B)
447
448 Note: If you select a rule using TUR ≥ N:1, be sure to specify what action to
449 take for any measurements that result in a TUR below the defined rule.

450
451

Page 14 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

452
453
454 Figure 7. Pass/Fail Conformity Decision Rule selection flow chart.

455
456 7. DECISION RULE DOCUMENTATION AND APPLICATION
457
458 It is the responsibility of the laboratory to agree on a service with the customer. In
459 ISO/IEC 17025:2017 para 7.1.3 it indicates that the request for a statement of conformity
460 must come from the customer. However calibration laboratories may offer standard
461 service deliverables with differing amounts of guard band (including zero) to provide the
462 customer with choices of levels of risk.
463

Page 15 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

464 Similarly in para 7.8.3.1 b) test laboratories shall provide statements of conformity if this
465 is necessary for the interpretation of the results. Regulated areas of testing generally have
466 specific standards for labs to follow, as has been described earlier in the document.
467
468 In all cases, decision rules need to be compatible with the customer, regulation or
469 standard requirements. They need to be agreed and documented, before the work starts. It
470 must be clear that the tolerance limits are consistent with the calibration requirements and
471 that all uncertainty and other calculations are performed consistent to the ISO/IEC
472 17025:2017 requirements. The agreed decision rule employed for statements of
473 conformance must be clearly documented in the measurement report.
474
475 In more complicated decision rules the following must also be included,
476 a) Documentation of other supporting factors such as statistical assumptions
477 including type of risk, specific or global and measurement uncertainty. (clause
478 7.8.6.1)
479 NOTE: for more information on specific and global risk see para 5.3.
480 b) Documentation of the conformity assessment type and statements of
481 conformance. (clause 7.8.6.2)
482 Note: for more info on decision rules and conformity statements see para 4.
483 c) Compatibility of decision rule documentation with test and calibration records.
484 (clause 7.8.6.2)
485 Annex A provides a sample checklist for both the Laboratory and the Assessor to follow
486 and Annex B provides some examples of documentation that may be required
487
488 8. SUMMARY
489
490 The concept of decision rules applicable to statements of conformity with specifications
491 or standards is not new. However, ISO/IEC 17025:2017 provides further clarity and
492 emphasis by requiring laboratories to:
493
494 1) understand the needs of customers relating to statements of conformity they may
495 require and that this be confirmed at the test / calibration request stage. The request
496 review stage is to take into account the application of the statements and agree with
497 the customer the decision rules to apply based on the risk the customer will accept;
498
499 2) include the decision rule in reports covering statements of conformity (unless the rule
500 is inherent in the specification or standard).
501

Page 16 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

502
503 9. REFERENCES
504
505 1. ISO/IEC 17025:2017, General requirements for the competence of testing and
506 calibration laboratories
507
508 2. JCGM 106:2012, Evaluation of measurement data – The role of measurement
509 uncertainty in conformity assessment.
510 Note: this document is also available as ISO/IEC Guide 98-4:2012
511
512 3. ASME, B89.7.3.1-2001, Guidelines for Decision Rules: Considering Measurement
513 Uncertainty in Determining Conformance to Specifications.
514
515 4. JCGM 100:2008, (GUM), Evaluation of measurement data - Guide to the
516 Expression of Uncertainty in Measurement.
517
518 5. ISO 14253-1:2017, Geometrical product specifications (GPS) – Inspection by
519 measurement of workpieces and measuring equipment – Part 1: Decision rules for
520 verifying conformity or nonconformity with specification.
521
522 6. JCGM 200:2012, (VIM), International Vocabulary of Basic and General Terms in
523 Metrology, Third Edition
524
525 7. NCSLI International, ANSI/NCSL Z540.3:2006 Requirements for the Calibration
526 of Measuring and Test Equipment, Boulder, Colorado, USA
527
528 8. David D., and Somppi, J., “A study of and recommendation for applying the false
529 acceptance risk specification of Z50.3”, Proc., NCSL Workshop & Symposium,
530 2007.
531
532 9. Dobbert, M., “A Guard-Band Strategy for Managing False-Accept Risk”, Proc.,
533 NCSL Workshop & Symposium, 2008.
534
535
536
537
538
539
540

Page 17 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

541
542 ANNEX A - SAMPLE CHECKLIST FOR MEETING THE REQUIREMENTS INCLUDES:
543
544 a) Documentation and records reflecting customer agreement requesting statement of
545 conformity to a specification or standard. (clause 7.1.3)
546
547 b) Records of the selection of test limits and associated tolerances and compatibility with
548 customer requirements (clause 7.1.3)
549
550 d) A documented decision rule for calculating, controlling, and reporting levels of risk
551 associated with statement of conformity. (clause 7.1.3)
552
553 c) Documentation of laboratory personnel that include the knowledge, skill, and
554 authorization to apply the decision rule and make statements of conformity. (clause 6.2.6
555 c)
556
557 e) Documentation of the calculation or estimation of level of risk and uncertainty. (clause
558 7.8.6.1)
559
560 f) Documentation of other supporting factors such as statistical assumptions including
561 type of risk specific or global and measurement uncertainty. (clause 7.8.6.1)
562 NOTE: for more information on specific and global risk see para 5 of this document.
563
564 g) Documentation of the conformity assessment type and statements of conformance.
565 (clause 7.8.6.2)
566 Note: for more info see para 4 of this document
567
568 h) Compatibility of decision rule documentation with test and calibration records. (clause
569 7.8.6.2)
570

Page 18 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

571
572 ANNEX B – EXAMPLES OF DOCUMENTED AGREEMENTS
573
574 Example 1 Simple acceptance (choice a in figure 7)
575 The customer agrees that Pass/Fail decisions are based on acceptance limits chosen based on
576 simple acceptance (w = 0, AL =TL). The expanded measurement uncertainty calculated per the
577 GUM must be less than 1/3 of the tolerance limits based on the manufactures specifications
578 (TUR> 3:1). Statements of conformity are binary. The measurand is assumed to have a normal
579 probability distribution and specific risk is used for the risk calculation. In this case, the risk that
580 accepted items are outside the tolerance limit is up to 50 %. The risk of false reject is up to 50
581 %1 for measured results outside the tolerance.

582
583 Measurement results are reported as:
584 • Passed - The measured values of the equipment were observed in specification at the
585 points tested. The specific false accept risk is up to 50 % for measurements near the
586 tolerance.
587 • Failed - One or more measured values of the equipment were observed out of
588 specification at the points tested. The specific false reject risk is up to 50 % for
589 measurements near the tolerance.
590
591 Example 2 Non-binary acceptance based on guard band w = U (choice b in figure 7)
592 The customer agrees that decisions are based on guard banded acceptance limits. (w = U, AL =
593 TL -w) where U is the expanded measurement uncertainty calculated per the GUM. Statements
594 of conformity are non–binary. The measurand is assumed to have a normal probability
595 distribution and specific risk is used for the risk calculation. In this case the risk of accepted
596 items to be outside the tolerance limit is < 2.5 %. For rejected items the risk to be inside the
597 tolerance limit is < 2.5%. When the measured result is close to the tolerance, the risk of false
598 accept and false reject is up to 50 %.

1
Since both false accept and false reject risk can be 50 %, this rule is sometimes called “shared risk”.

Page 19 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

599
600 Measurement results are reported as:
601
602 • Pass - The measured values of the equipment were observed in specification at the points
603 tested. The specific false accept risk is up to 2.5 %.
604
605 • Conditional Pass - The measured values of the equipment were observed in specification at the
606 points tested. However, a portion of the expanded measurement uncertainty intervals about one
607 or more measured values exceeded specification. When the measured result is close to the
608 tolerance, the specific false accept risk is up to 50 %.
609
610 • Conditional Fail- One or more measured values of the equipment were observed out of
611 specification at the points tested. However, a portion of the expanded measurement uncertainty
612 intervals about one or more measured values were in specification. When the measured result is
613 close to the tolerance, the specific false reject risk is up to 50 %.
614 • Fail - One or more measured values of the equipment were observed out of specification at the
615 points tested. The specific false reject risk is up to 2.5 %.
616 Example 3 Binary acceptance based on guard band (≤ 2.0 % global risk) (choice c in figure
617 7)
618 The customer agrees that decisions are based on guard banded acceptance limits, AL, to result in
619 less than 2 % false accept [global] risk. For this case the guard band w, is given by
620
621 and U is the expanded measurement uncertainty calculated per the GUM [4]. As before, AL =
622 TL -w. Statements of conformity are binary. The measurand is assumed to have a normal
623 probability distribution. The risk of accepted items to be outside the tolerance limit is ≤ 2.0 %.

624
625 Measurement results are reported as:
626 • Passed - The measured values of the equipment were observed in specification at the
627 points tested with a global false accept risk of less than or equal to 2 %.
628

Page 20 of 21
 ILAC-G8:20xx

Guidelines on Decision rules and Conformity with Requirements

629 • Failed - One or more measured values of the equipment were either observed out of
630 specification at the points tested or the global false accept risk for one or more measured
631 values was greater than 2 %.
632

Page 21 of 21