Is ISO 17100 certification enough to approve a translation vendor?

No. ISO 17100 evidences a defined translation-service process; it does not show that the process has been applied to your pair, files, data route or deadline. Use it as the entry check, then audit the six evidence areas for the named scope.

Should every finalist run a pilot?

Only where the written evidence leaves a gap that a pilot can close. When a pilot runs, every finalist gets the same files, instructions and acceptance rule, or the results cannot be compared. A pilot cannot repair a failed mandatory condition such as a disallowed data route.

Can a supplier be approved with an open gap?

Yes, if the gap is named, owned and dated, and the approval is conditional on it closing before live work in the affected area. An unnamed gap folded into an overall score is how a critical risk reaches production.

Does an approval cover new languages or content types?

No. It applies to the scope that was audited. A new pair, domain or file type reopens the areas it touches, usually pair fit, review and handoff, rather than the whole audit.

What should the audit record contain?

The brief, each finalist's response, the status of every mandatory area, pilot results including correction effort, and the decision with its conditions. That record lets procurement explain the decision later without relying on memory.

What happens if you cannot staff one of my language pairs?

For the proposed project, ask for pair-by-pair availability or a recruitment window in writing before agreeing a date. Define qualification and pilot approval for any new contributor before live work. A coverage claim should be checkable before the scope is signed.

How to audit a translation vendor: a procurement checklist

Why a shortlist of yes answers still leaves the choice open

A vendor questionnaire tends to produce the same result every time. Each finalist confirms the languages, the subject areas, the review step, the security controls and the capacity. The certificates look alike. The only column that moves is the rate, so the decision drifts toward it by default.

The gaps surface later, in the first live batch. The reviewer qualified for the pair is on another project. The files come back with broken tags. The source was passed to a subcontractor nobody approved. A correction dispute has no owner. None of this was hidden; the questionnaire never asked for the evidence that would have shown it.

An audit moves from what a supplier says it can do to what it can show for this project: this language pair, this content risk, this file format, this review model and this acceptance owner.

Start from a tender brief, not a questionnaire

Send every finalist the same short package: source and target languages with any regional variant, domain and audience, representative sample files, the volume pattern, timing constraints, terminology and style sources, security requirements, and the definition of acceptance, including who accepts. Mark each condition as mandatory or preferred. A vague brief produces responses that cannot be compared, and the comparison collapses back to rate.

Ask each response to name its assumptions and exclusions. Does the price include independent revision? Are layout and tag handling in scope? Which source changes after lock trigger rework, and at whose cost? Procurement does not need a capability deck here. It needs a response that could become the first page of a statement of work.

Treat certification as the entry check

If the brief requires ISO 17100, confirm that the certificate's scope covers the translation service being bought, then move on. The ISO 17100 vendor qualification guide covers how to read the scope of a translation certificate and what it does and does not evidence. A certificate shows that a process exists. It does not show that the process has been applied to this pair, these files or this deadline, which is what the six evidence areas are for.

Check six evidence areas against the brief

Pair and domain fit: ask who will do the work, how their qualification for this pair and subject was checked, and who steps in if that person is unavailable. An organisation-wide language count is not evidence that the reviewer a rare pair needs is free on the start date; the pair-by-pair coverage test sets out the written answer to request.

Review and acceptance: ask for the proposed translation, revision and final-file checks for this content, an anonymised example of the error-log format, and the route for resolving a disagreement. Agree error categories, severity rules and the unit of measurement before any sample is scored. If the review model itself is still undecided, settle it first with choosing a translation QA model. The audit only asks whether the supplier can produce the evidence that model requires.

Security and data route: map where source files will be stored, who can open them, whether subcontracting is permitted and to whom, which tools touch the text, and how files are returned or deleted under the actual contract. A statement that the process is secure is not a data route. The secure multilingual delivery guide lists the evidence to request; the buyer's own security and legal owners decide whether it is sufficient for the material.

Production handoff: test the real file type, tag and layout handling, terminology updates issued mid-project, source-version control and the delivery package the buyer's systems expect. A supplier that translates well but returns files that fail import has added a production step to every batch.

Correction terms: agree how issues are reported, corrected, re-checked and carried across related files. Separate the two cases behind most disputes: the source changed after lock, or a delivered translation failed an agreed criterion. Decide in advance who carries the cost and schedule impact of each.

Capacity and continuity: for a rolling programme, ask how absences, volume peaks and reviewer continuity per language are handled. Request a staffing plan against the buyer's own forecast, not a statement about the size of a network.

Turn the evidence into approve, test or decline

Mark each mandatory area evidenced, needs a test, or not met, and do not average them. A disallowed data route, a certificate whose scope misses an in-scope requirement, or no qualified reviewer for a required pair each stops the contract on its own. A weighted score that lets strong answers elsewhere absorb a critical gap produces a comfortable number and the wrong supplier.

Where the evidence is incomplete but the risk is reversible, run a representative pilot. Give every finalist the same files, instructions and acceptance rule, and record more than the error count: the corrections requested, how they were handled, and the buyer-side time needed to reach final acceptance. That last measure often separates two suppliers whose scores look identical.

The audit closes with one of three outcomes: approve for the named scope; test, through a pilot or an approval that waits on a named gap closing; or decline, when a mandatory area is not met. Record the outcome against the scope, not the supplier. An approval for legal German does not carry over to marketing Japanese; a new pair or content type reopens the areas it touches. The result is a procurement record, not a public ranking of translation companies.

Where MoniSa fits in this audit

The audit applies to MoniSa as it does to any finalist. The MoniSa engagement model follows the same logic: languages, reviewers and timeline are confirmed against the buyer's brief, acceptance criteria are agreed before production starts, and a pilot batch is reviewed and approved by the buyer before production opens. MoniSa holds ISO 9001:2015 and ISO 27001:2022 certification, and its ISO 17100:2015 certification covers translation services only. Send the tender brief in the evidence format procurement already uses, and ask for the response against the six areas above rather than as a general capability statement.

The other steps in a translation vendor decision

Each of these pages goes deeper on one part of the audit.

ISO 17100 vendor qualification: Use for the certification step, when the scope of a translation certificate and independent review need checking.
Verify language coverage before you sign: Use when pair and domain fit is the open question.
Choosing a translation QA model: Use when review layers, severity and acceptance have not yet been designed.
Supplier evidence matrix: Use to record each supplier claim beside its evidence and the decision it supports.
Secure multilingual delivery guide: Use when the data route is what decides the contract.

What to hold for each finalist before the contract decision

Confirm each line is on file for every finalist before the decision meeting.

The same tender brief, sent unchanged, with mandatory conditions marked
A written response naming assumptions, exclusions and what triggers rework
Certificate scope checked against the service being bought
Named qualification and a backup for each required pair and domain
Review steps, an error-log example and a disagreement route for this content
A data-route map covering storage, access, subcontracting, tools, return and deletion
A handoff test on the real file type and delivery package
Correction terms that separate source changes from translation failures
A staffing plan against the buyer's forecast
Each mandatory area marked evidenced, needs a test, or not met

Answers that need a scoped follow-up

None of these proves a supplier is unsuitable. Each one means the audit for that area is not finished.

A capability yes with no named person, process or record behind it
A language count offered in place of a pair-level answer
A security badge or certificate offered in place of a data-route map
Subcontracting permitted in the response but missing from the data route
Pilot files chosen by the supplier rather than taken from the real content
One weighted score that hides a failed mandatory area

What to send MoniSa for a comparable response

Send the same package every finalist receives, so the responses can be compared line by line.

The tender brief: languages and variants, domain, audience, volume pattern and timing constraints
Representative sample files in the real format, with the delivery package your systems expect
Terminology, style and reference sources, and how updates will be issued
Security requirements, subcontracting rules, and the contract's return and deletion terms
The acceptance definition, the acceptance owner and the evidence format procurement requires

Approve, test or decline? Auditing a translation vendor before the contract