Interpreting quality assurance examines whether the meaning conveyed through spoken or signed language was faithful, professionally appropriate and usable in the actual interaction. The gold standard is evidence-led review with fair treatment of participants and interpreters—not a general fluency score or an automatic transcript compared word for word.
This guide follows the fictional Alder View subsidy appointment, where ‘not yet approved’ is rendered as though approval were final. It distinguishes a material semantic error, an unclear source, a failed audio path and the separate authority of the organisation making the underlying decision.
For service delivery, see Spoken Language Interpreting Services and Sign Language Interpreting Services. Interpreter Assignment Coordination covers booking and staffing. This article owns evidence-based assessment of the interpreted exchange.
The four questions every interpreting QA client should settle
- **Source:** What reliable, authorised audio or video captures what was actually spoken or signed?
- **Evaluator:** Who understands the particular languages, interpreting mode and subject matter?
- **Standard:** What professional criteria and contextual severity rules govern this review?
- **Remedy:** Who may correct a service defect, and who alone may decide the underlying case?
Choose a reading route
22-chapter evidence and correction index
Purpose, standards and trustworthy source · chapters 1–7
- A polished interpretation can conceal a material misunderstanding
- Define the interaction being evaluated before selecting a test method
- Use an interpreting standard rather than a translation standard by mistake
- Distinguish service quality assurance from individual accreditation
- Protect confidentiality before obtaining recordings or transcripts
- Establish trustworthy source and interpretation records
- Select reviewers who understand both languages and professional context
Accuracy, uncertainty and professional roles · chapters 8–13
- Compare communicative intent, not surface word order
- Identify the errors that carry disproportionate consequences
- Assess how ambiguity and uncertainty were handled
- Evaluate additions, omissions and changes of emphasis fairly
- Check impartiality and role boundaries without erasing empathy
- Include turn-taking and timing in the assessment
Settings, technical conditions and evaluation · chapters 14–18
- Separate the quality of the interpreting platform from the linguistic result
- Evaluate specialist settings through appropriate experts
- Include signed languages and visual context within suitable reviews
- Build a severity rubric connected to meaningful outcomes
- Choose samples that reveal important patterns without claiming perfection
Reviewer agreement, correction and fairness · chapters 19–22
1. A polished interpretation can conceal a material misunderstanding
A fictional neighbourhood service centre, Alder View, helps a resident understand a decision about a home-improvement subsidy. The officer says the proposed work ‘has not yet been approved’, but the interpreter inadvertently leaves out the qualifying words ‘not yet’. The resident believes approval is final and arranges a contractor. The meeting recording sounds courteous and the appointment took the planned time, yet the most consequential message changed.
The gold standard of interpreting quality assurance is a fair, technically informed review of whether the interpreted communication accurately conveyed the source meaning within the agreed professional role. It is not an attempt to punish every hesitation or convert interpreter evaluation into a review of the underlying subsidy decision. Alder View and its people are fictional. The scenario shows why a single missed negation can outweigh an otherwise fluent, friendly performance.
2. Define the interaction being evaluated before selecting a test method
Interpreting includes community appointments, conferences, healthcare conversations, legal settings, customer support, remote calls and signed languages. Each setting affects terminology, professional boundaries, context and the harm of a misunderstanding. Evaluation must identify the specific language pair, direction, mode, subject, audience and authorised outcome.
For Alder View, a reviewer should know who issued the decision, what the resident needed to understand, whether the conversation was consecutive or remote, and which source materials were authoritative. A generic score for language fluency alone does not answer whether the interpreter met that task. A quality-assurance contract should define scope and limitations so the client understands what has actually been assessed.
3. Use an interpreting standard rather than a translation standard by mistake
Interpreting is the real-time rendition of spoken or signed meaning; translation principally concerns rendering content between languages in another written or signed form. Their processes and evaluation needs differ. A source transcript can help reconstruct an interpreted session, but evaluating the transcript alone may miss timing, turn-taking or visual elements.
ISO 18841:2018 remains the published general interpreting-service standard in October 2026, with a replacement in final-draft development. ISO 13611:2024 addresses community interpreting. ISO 5060:2024 concerns evaluation of translation output and explicitly does not apply to interpreting. An assurance provider should never claim that passing a document-translation test proves a live interpreter’s competence.
4. Distinguish service quality assurance from individual accreditation
An evaluator may review one assignment, analyse a provider’s management process, assess training needs or conduct a formal qualification process. These are different products. An internally issued ‘QA passed’ label does not automatically confer nationally recognised interpreter accreditation or permission to practise in a regulated setting.
Define the purpose in the statement of work. A recorded session review may show whether specific information was conveyed; it cannot certify that the person is qualified for every future language direction or subject. A competent reviewer should describe the observed evidence and meaningful scope instead of selling a universal credential. When statutory or contractual accreditation is needed, identify the actual responsible authority or recognised scheme.
5. Protect confidentiality before obtaining recordings or transcripts
Quality reviewers may encounter sensitive personal circumstances, legal advice, medical issues, employment disputes or private financial information. A service owner cannot assume an existing recording was authorised for all possible training and AI analysis. The lawful purpose, relevant consent, access rights and retention conditions need assessment before transferring the source.
Use appropriately authorised and minimised material where possible, removing unrelated identifiers from evaluation copies without destroying the evidence needed to judge meaning. Singapore’s PDPC guidance on the Personal Data Protection Act provides relevant private-sector context; public-sector and other regimes can differ. Reviewers need secure communication routes and must not turn a resident’s dispute into a public training example without proper permission.
6. Establish trustworthy source and interpretation records
Review quality depends on knowing what the original speaker said and what the other person actually received. Separate audio channels, speech overlap, missing fragments or an edited recording can change the evaluator’s understanding. A transcript created automatically may confidently add a word that neither speaker said.
Document the source file, relevant version, channel or video information, recording conditions, participant roles and any known gaps. If one critical phrase is inaudible, describe the uncertainty rather than conclude the interpreter omitted it. For the fictional subsidy case, compare the officer’s exact statement with the interpreted rendition in appropriate original recordings. Never make a sanctioning recommendation based on a reconstruction with no credible source.
7. Select reviewers who understand both languages and professional context
Someone may read both languages well yet lack the listening or signed-language competence to evaluate a live dialogue. A subject expert may identify a technical issue but misunderstand the interpreter’s linguistic choices. Appropriate quality review can require complementary linguistic and domain expertise.
Match evaluators to language varieties, working direction, interpreting mode and subject matter. A resident’s Hokkien conversation should not be assessed solely by someone familiar with Mandarin. Consider conflicts of interest and independence when assigning reviewers, particularly after a dispute. The objective is a fair assessment of meaning and professional conduct rather than a review designed to support a manager’s preferred narrative.
8. Compare communicative intent, not surface word order
Faithful interpreting need not repeat every source word in an identical order. Languages differ in syntax, speech acts, implied meaning and ways of marking politeness. An evaluator should ask whether the listener received the essential proposition, conditions and tone, within the reasonable limits of the mode and setting.
The Alder View officer’s sentence about pending approval could be rendered naturally in another language without literal English word order. That is acceptable if the provisional status remains clear. A rigid word-matching score may punish a competent interpretation and miss a smooth but misleading one. Document the semantic decision and relevant context, including what the source speaker meant and how the recipient reasonably understood it.
9. Identify the errors that carry disproportionate consequences
Negations, numbers, dates, conditions, permissions, dosage information, safety instructions and proper names often carry high-impact meaning. An omission affecting one of these may have a larger practical consequence than several minor shifts in style. Error assessment should therefore consider materiality and the task, not merely the count of deviations.
In the fictional subsidy meeting, ‘not yet approved’ becoming ‘approved’ is a critical error even though the rest of the discussion is fluent. A reviewer should identify the original phrase, interpreted meaning and reasonable consequence without pretending that the interpreter alone authorised the later contractor purchase. Record material uncertainty and avoid extrapolating one error into unsupported conclusions about every appointment the professional has ever handled.
10. Assess how ambiguity and uncertainty were handled
Speakers sometimes use unknown terms, mumble, change meaning mid-sentence or refer to documents unavailable to the interpreter. Professional practice includes appropriate clarification and transparent communication of uncertainty rather than guessing. A reviewer should distinguish an error the interpreter caused from ambiguity originating in the source.
Suppose the officer uses an unfamiliar internal abbreviation for a programme stage. The interpreter’s request for clarification can be a strength rather than a failure to demonstrate perfect fluency. Review whether the parties received the same corrected explanation. A scoring system that punishes every clarification may incentivise confident invention. Good assurance rewards faithful handling of unclear information and identifies source-briefing problems the organisation itself must fix.
11. Evaluate additions, omissions and changes of emphasis fairly
An interpreter may condense repetitions or rephrase for naturalness without changing essential meaning. But unjustified additions, omitted objections or shifts in tone can materially affect a conversation. An evaluator should account for the service mode and actual communicative goal rather than counting every word as a separate contract obligation.
For Alder View, deleting an emotional complaint might make the resident appear satisfied when they were not; adding a reassurance about approval could create a promise the officer did not make. Compare representative passages in full context. Explain whether a finding concerns inaccurate content, professional boundary or communication management. Avoid using a single undefined error percentage as a substitute for a reasoned assessment.
12. Check impartiality and role boundaries without erasing empathy
Interpreters are not usually the substantive decision-makers in the meetings they assist. They may appropriately facilitate clarification while remaining faithful and professionally impartial. Quality review should examine whether the interpreter crossed into persuading the participant, altering a service rule or expressing personal judgement as though it were the official’s.
An interpreter at a community meeting can speak respectfully and show ordinary human warmth without becoming the resident’s adviser. If the service official gives unclear information, the interpreter should not quietly repair the policy by inventing an answer. Consider whether the original speakers understood the interpreter’s role and whether the meeting structure allowed proper turn-taking. Some boundary problems begin with client instructions, not only interpreter choices.
13. Include turn-taking and timing in the assessment
Consecutive interpreting requires manageable segments, pauses and chances to clarify; simultaneous interpreting has different demands and working arrangements. Overlapping speech, interruptions and rushed meeting schedules can cause omissions regardless of vocabulary knowledge. Review those conditions before assigning responsibility for a missed detail.
The subsidy officer might deliver a two-minute explanation of complex conditions without pausing. The interpreter’s workload and opportunity to render it faithfully become part of the evidence. If quality problems recur, changing the meeting format may be more effective than instructing one interpreter to ‘pay closer attention’. ISO 23155:2022 sets requirements and recommendations for conference interpreting within its scope, not every one-to-one appointment.
14. Separate the quality of the interpreting platform from the linguistic result
A recording may be incomplete because a connection dropped, microphone failed or language channel was switched incorrectly. Those issues differ from the interpreter selecting an inaccurate word, even if the recipient experiences both as a misunderstood message. Assurance should distinguish technical root causes from linguistic root causes.
ISO 24019:2022 addresses remote simultaneous interpreting delivery platforms in defined settings. It is not proof that all ordinary video meetings meet the standard. For Alder View, reviewers should examine whether the source audio was audible and whether the participant received the interpreted channel. A ‘connection successful’ indicator alone is not sufficient technical evidence.
15. Evaluate specialist settings through appropriate experts
Legal, healthcare, educational and other high-consequence interpreting can involve terms whose meaning depends on procedure and professional context. Reviewers must avoid treating ordinary-language fluency as proof of specialist accuracy. The provider should select competence appropriate to the assignment, and the relevant organisation remains responsible for its official decisions.
For example, a disputed legal undertaking requires qualified legal and interpreting expertise rather than a casual manager’s review of a machine transcript. Where interpretation affects an important outcome, seek a legally and professionally appropriate remedy through the institution that made that decision. A quality-assurance company cannot automatically reverse a court order or medical diagnosis. Keep the review’s powers clear and its factual findings usable.
16. Include signed languages and visual context within suitable reviews
Signed-language interpreting has language-specific grammar, spatial information and nonmanual features that cannot be assessed adequately from a voice-only transcript. Camera angle, visibility of signers, screen content and video quality may affect the information available to the interpreter and recipient.
Choose reviewers competent in the actual signed languages and intended setting. Use properly authorised video evidence and preserve privacy. The eduKate Sign Language Interpreting Services guide discusses service delivery; the QA provider should identify which aspects of fidelity, timing, ethics and access it can evaluate. A hidden signing window is a technical exclusion problem, not evidence of poor language performance.
17. Build a severity rubric connected to meaningful outcomes
A report needs a way to distinguish minor stylistic variation, moderate misunderstandings and material changes affecting a party’s rights or decisions. A suitable rubric states how errors are classified, which contexts matter and how reviewer judgements are calibrated. Numerical results can be useful when defined, but should not pretend all errors have the same consequence.
In the Alder View case, the omitted provisional status is high impact. An evaluator might separately record a harmless alternative phrasing of the centre’s name. ISO 5060:2024 illustrates human analytic error evaluation for translation output, but expressly excludes interpreting. It may inspire transparent reasoning but cannot be claimed as the interpreting assessment standard. Build the actual interpreting rubric from appropriate professional expertise and the assignment’s risk.
18. Choose samples that reveal important patterns without claiming perfection
An organisation may review randomly selected sessions, targeted complaints, new-interpreter onboarding and high-risk interactions. Each sample serves a different purpose. Reviewing only recordings that triggered complaints can exaggerate general error rates, while examining only easy calls can hide serious risks.
Document the sampling rationale, coverage, period, language pairs, service modes and limits. Consider whether a small sample can support a claim about the whole provider network. Where sensitive recordings cannot lawfully be retained, explore other appropriate observation or feedback arrangements. A transparent evaluation statement can say that certain high-risk calls were reviewed and problems identified without inventing a universal accuracy percentage.
19. Resolve reviewer disagreement through evidence and adjudication
Two bilingual reviewers may disagree whether a local expression preserved a particular condition or tone. Quality assurance should not suppress disagreement merely to issue a neat score. Define a route for second review, relevant specialist input and reasoned adjudication, including conflicts of interest.
Keep the original excerpt and the linguistic arguments in an appropriately protected record. A review lead should explain why the final decision is supported, including remaining uncertainty. Staff should not be punished for identifying ambiguous cases, nor should an evaluator inflate findings to justify a commercial retainer. Fair and documented adjudication makes the result more useful to both the interpreter and the commissioning organisation.
20. Turn findings into training and process corrections
Not every quality problem is solved by retraining the interpreter. Source speakers may need clearer turn-taking, coordinators may need better dialect matching, technology vendors may need to repair channel failures, and the client may need to provide authorised terminology before the meeting.
Assign each corrective action to the actual owner and verify that it changes subsequent interactions. If similar provisional-approval errors occur across several sessions, investigate whether the client’s standard explanation is confusing. Record what was changed, when it was tested and which relevant staff were briefed. A completed training attendance sheet is not evidence that the underlying misunderstanding disappeared.
21. Provide a fair complaints, appeals and correction route
Participants and interpreters should have a way to report and contest material findings under appropriate confidentiality. The QA provider should not publish identifying examples unnecessarily or present an allegation as proof. Serious concerns require prompt, impartial handling matched to their actual risk and jurisdiction.
Where a resident may have acted on a wrong interpretation, the community organisation needs to review the underlying case appropriately and communicate the verified outcome in a language the resident understands. The evaluator supplies evidence and recommendations within scope. A formal sanction, contractual remedy or legal decision belongs to the authorised party. Good review supports fairness for the affected participant and the interpreter rather than serving a predetermined disciplinary goal.
22. The gold standard is a review that makes future dialogue safer and clearer
Return to Alder View. The QA team identifies the missing ‘not yet’, confirms what source and interpreted audio actually contain, distinguishes the linguistic issue from any technical problem and submits a protected factual report. The centre reviews the resident’s situation under its actual rules, improves its briefing and provides an appropriate language-access route for clarification.
Interpreting quality assurance earns its gold standard when findings are competent, proportionate, reproducible and useful for correction. A polished performance is not necessarily accurate; an isolated error does not justify unsupported generalisations. The service’s value is enabling fairer dialogue and stronger professional practice while respecting everyone whose words were entrusted to an interpreter.
Worked example: fluent interpreting with a critical change in meaning
Alder View is fictional. This table identifies what each kind of evidence can and cannot establish before a fair outcome is reached.
| Observation | Potential failure | Valid next check |
|---|---|---|
| Resident says approval was confirmed | Material interpretation or source misunderstanding | Compare authorised source and interpreted rendition |
| Transcript omits ‘not yet’ | Recording or recognition error also possible | Listen to actual audio and establish uncertainty |
| Interpreter was courteous | Fluent style does not prove faithful content | Subject-aware review of key conditions |
| Call quality was poor | Technical barrier may affect performance | Assess source and receiver audio separately |
| Resident incurred an expense | Underlying rights and remedies require review | Refer decision to authorised organisation |
A fair review and responsibility matrix
Use this matrix to avoid assigning every failure to the interpreter or to the platform. The correct legal authority depends on the actual meeting.
| Question | Responsible reviewer / owner | Acceptance evidence |
|---|---|---|
| Did the original statement have this meaning? | Qualified source-language and context reviewers | Verified source record and relevant context |
| Was that meaning faithfully conveyed? | Qualified interpreting QA reviewer | Compared rendition with reasoned findings |
| Did the technology deliver the message? | Platform or sound specialist | Route, channel and quality evidence |
| Did the underlying decision stand? | Authorised institution | Correct applicable case review |
| Was the service corrective action successful? | Provider QA manager | Documented change and later verification |
A Clementi-plus QA procurement checklist
- Define interpreting language, direction, mode, subject and exact reviewed scope.
- Use current published interpreting standards rather than translation-output standards.
- Protect authorisation, privacy, recording custody and reviewer independence.
- Inspect source audio/video quality before judging the interpreted message.
- Assess conditions, numbers, names, tone and relevant professional boundaries.
- Separate technical, institutional and linguistic root causes fairly.
- Use documented severity, sampling, reviewer calibration and adjudication.
- Produce a clear correction, remedy and challenge route without deciding outside authority.
Frequently asked questions
These answers prevent common mistakes in selecting a QA standard, judging interpreters and handling confidential source evidence.
1. Can ISO 5060:2024 be used to certify spoken interpreting?
No. ISO 5060:2024 provides guidance for evaluating human, post-edited and unedited translation output, and its published scope explicitly excludes interpreting. Some general ideas about clear error classifications can inform careful thinking, but one must not label an interpreted meeting ISO 5060-compliant. For interpreting, consult standards such as ISO 18841:2018 and the relevant specialised standard, then define the actual evaluation procedure and professional competence.
2. Is ISO 18841:2018 still current in October 2026?
Yes, it remains the published general interpreting-services standard as of the article’s date. A replacement ISO/FDIS 18841 exists at final-draft stage but is not yet the published replacement. A service provider should describe the correct edition and relevant contractual requirements. The presence of a draft does not automatically invalidate existing published requirements or certify any individual interpreter.
3. How can someone prove a crucial word was omitted?
Start with properly authorised source and interpreted audio or another reliable direct record, confirm the speakers and relevant context, then ask a competent language reviewer to compare the two utterances. Be alert to recording gaps and machine-transcript mistakes. State precisely what evidence supports the conclusion and what remains uncertain. A complaint, an edited excerpt or an automated transcript alone may be insufficient for a consequential adverse finding.
4. Should reviewers match original words one for one?
No. Professional interpreting conveys meaning across languages that may have different word order and expression. A literal matching method can punish competent natural phrasing while missing a fluent but material change. Evaluate essential propositions, conditions, speaker intent and relevant tone in context. A rendered phrase can differ in surface form and still be faithful, while dropping ‘not yet’ from an approval condition is a clear potential semantic error.
5. Who should conduct an interpreting QA review?
Choose evaluators competent in the relevant spoken or signed languages, working direction and interpreting mode. Specialist fields may require complementary subject expertise. Manage conflicts of interest and explain reviewer authority. A bilingual manager, general translator or automatic transcription system is not automatically qualified to evaluate every professional interpreting performance. A fair review should have a documented route for reasonable disagreement and adjudication.
6. Can the organisation record every interpreted conversation for quality control?
Not without considering the actual law, contractual commitments, purpose, participant consent where needed and the sensitivity of the discussion. Set proportionate recording, access and retention policies before using source material for QA, especially in healthcare, legal or social-service contexts. The ability to click Record in a meeting platform is not sufficient authorisation. Reviewers should work with the least sensitive material that remains adequate for a fair assessment.
7. Why should a technical failure be graded separately?
If an interpreter could not hear the source or the recipient received the wrong audio channel, it may be unfair to classify the resulting gap as a language competence failure. Distinguish technical availability, professional rendition and the client’s original clarity. That distinction identifies the person or system that must be corrected. The participant’s access problem is still real even if no individual interpreter caused it.
8. How should evaluation scores represent serious errors?
Define severity against the task and intended consequence rather than counting each deviation identically. Names, dates, negation, permissions and material conditions may carry disproportionate importance. Record what was observed, which professional judgement was used and where the sample’s limits lie. A high overall accuracy score can conceal one critical mistake, while a low score caused by harmless surface variation may be misleading.
9. Does one mistaken interpretation establish that the interpreter is unqualified?
Not automatically. A single assignment can identify a material error, but broader conclusions require appropriate evidence, context, scope and fair process. Review source speaker pace, audio quality, working conditions and relevant briefings. A serious error deserves appropriate corrective action for the affected participant without inventing a lifelong professional assessment from one recording. Individual qualification and service-incident review are different products.
10. How should signed-language interpretation be assessed?
Use qualified reviewers of the actual signed language and evaluate visible source and rendition, spatial and nonmanual grammar, timing and the communication setting where relevant. A typed transcript or an audio-only recording cannot capture all necessary evidence. Consider camera framing, visibility and technical quality separately. Protect participant and interpreter privacy because recorded signing can reveal identifiable personal information.
11. Who corrects the underlying subsidy or legal decision after a problem?
The authorised organisation, tribunal or other competent decision-maker, according to the real process, owns that decision. The interpreter or QA reviewer should not quietly reverse or invent it. QA can establish what communication occurred and notify the responsible institution through proper channels. The participant then needs a clear, appropriate language-access route to understand the verified outcome and any legitimate review options.
12. What should a completed QA report include?
Identify the assignment type, language and mode, reviewed materials, authorised purposes, evaluation criteria, evidence, substantive findings, severity and context, reviewer competence, unresolved uncertainty and correction owner. Summarise samples and technical limitations honestly. Protect confidential recordings and contributor details. Provide a process for contesting or correcting material factual mistakes in the report. A useful report helps a buyer take legitimate action without claiming universal certification.
Current professional references and connected eduKate services
- ISO 18841:2018 — General Interpreting Requirements (published; final draft replacement pending)
- ISO 13611:2024 — Community Interpreting
- ISO 23155:2022 — Conference Interpreting
- ISO 24019:2022 — Simultaneous Interpreting Platforms
- ISO 5060:2024 — Translation Output Evaluation (explicitly not interpreting)
- Singapore PDPC — Personal Data Protection Act
- eduKate — Spoken Language Interpreting Services
- eduKate — Interpreter Assignment Coordination
- eduKate — Remote Interpreting Platform Services
- Gold Standard Sectors & Services Directory
Standards and legal note, October 2026: ISO 18841:2018 remains published while ISO/FDIS 18841 is under development. ISO 5060:2024 explicitly excludes interpreting from its evaluation scope. No review of one call automatically certifies a professional or determines a legal or social-service outcome. The applicable contractual, privacy and sector rules must be checked separately.