A step in the wrong order
When a translated procedure inverts two steps, the document looks correct until someone follows it.
The instructive failures are the ones where a translated procedure inverted two steps, which is why this trade has a standard at all.
The failure that doesn't look like a failure
A translated document has a particular property that almost no other professional artefact shares: it can be completely fluent, grammatically clean, and terminologically consistent, and still be wrong in a way that only becomes visible at the moment of use. A misprinted circuit diagram will confuse an engineer at the bench. A building plan with the wrong dimensions fails at the first measurement. A translated procedure that inverts two steps does nothing at all until someone picks it up and begins to work through it — and by then the cost of the error is no longer the cost of a correction.
This is the failure mode that gave the translation profession its most serious quality argument, and eventually its formal standard. Not illegible prose, not an approximate word choice, but a document that reads entirely as it should while encoding a sequence that will not work, or will not work safely. The error is structural, invisible to fluency checks, and undetectable by any tool that does not hold the source and the target side by side and read both with equal attention.
What has to be held together
The source text in a procedural document is not just a collection of propositions. It is a sequence, and the sequence is load-bearing. Step seven depends on the state created by step six; if the translation swaps them, the surface reading may still be coherent — both steps might use complete, idiomatic sentences in the target language — but the underlying logic has been broken. A translator working at sentence level, as most do in ordinary practice, has no automatic prompt to verify that the order of steps in the target text corresponds to the order in the source. The sentence has been rendered; it reads well; the work moves forward.
Post-editing is a different job from translating, with a different failure mode: the errors are fluent. What machine output changes
The same applies to conditional structure. "If the valve is closed, open it before proceeding" contains two events and a logical dependency. A translation that renders the condition after the instruction — even while preserving both elements — has inverted the decision tree. The result, again, is a well-formed sentence with a broken procedure inside it.
What prevents this, in professional practice, is not fluency but oversight: a second qualified reader who works not from the target text alone but from a comparison of source and target simultaneously. This is what revision under ISO 17100 specifies — not a read-through for grammar, but a bilateral check that the source and target are aligned in content, structure and sequence. The standard, published by the International Organization for Standardization and first published in 2015, makes revision by a second qualified person a formal requirement, not an optional quality layer. The argument for that requirement is precisely the structural failure described above: fluency does not catch it.
The standard specifies qualifications, a second qualified person revising the whole text, and records — because the failure it exists to prevent is physical. What ISO 17100 asks for
Why the standard exists in the form it does
ISO 17100 was not constructed from first principles in a committee room. It grew from an accumulating record of failure in technical and procedural texts, and from the argument that translation, like other high-stakes professional activities, required a quality structure that could be audited and certified. The formal separation of the translator and the reviser — two distinct qualified persons, neither revising their own work — is the standard's structural response to the structural failure. Self-revision is demonstrably weaker because the translator, re-reading their own text, is prone to reading what they intended to write rather than what is there. The reviser brings a genuinely fresh comparison of source against target.
This is not a recent insight. Technical documentation communities — aerospace, pharmaceutical, nuclear — had long understood that procedural documents required controlled-language practices and independent review. What ISO 17100 did was carry this into general professional translation practice and give it a formal architecture. The Directorate-General for Translation of the European Commission, the largest single translation operation in the world, maintains its own quality framework aligned with these principles, as does the Court of Justice of the European Union in Luxembourg, where the procedural precision required by legal texts has the additional property that a mis-sequenced legal instrument does not merely fail to work — it means something different in law.
The terminology dimension is a further layer. A procedure that correctly sequences its steps but uses two different terms for the same component — because a term base was not applied consistently, or because the term base did not exist — creates ambiguity at exactly the wrong moment. The operator encountering "release the retaining pin" in step four and "remove the locking bolt" in step nine may or may not understand that these are the same part. The discipline of terminology work exists to eliminate that ambiguity before the translation begins: one object, one name, enforced throughout the document set. Without it, even a correctly sequenced procedure can mislead by inconsistency.
This is the failure mode that gave the translation profession its most serious quality argument, and eventually its formal standard
The instructive case
The most durable argument for structured quality in translation is not theoretical. It is the accumulated evidence of domains where sequencing failures have had traceable consequences — pharmaceutical package inserts rendered in a language the label's manufacturer did not speak; operating instructions for powered equipment where a safety step was placed after the action it was meant to precede; maintenance procedures where "before" and "after" were not flagged as sequence markers requiring special attention during translation. These cases are rarely published as failures; they surface in regulatory correspondence, in incident reports, in post-market surveillance data. The European Medicines Agency requires translations of product information to meet defined quality standards precisely because the translated label is a procedural document — it tells a patient or a clinician what to do, and in what order.
What makes the inverted-step failure so persistent is that it survives most of the checks that organisations actually apply. A spell-checker does not see it. A terminology check will not find it if all the terms are correct. A bilingual glossary review will not catch it. Even a careful monolingual read of the target text may not catch it, because the prose is coherent. The only reliable detector is a reader who holds both texts and traces the structure of one against the structure of the other — which is, in exact terms, what a reviser under ISO 17100 is required to do.
The standard, in this reading, is not bureaucratic process for its own sake. It is the formalisation of a lesson learned from watching fluent, finished, apparently correct documents fail at the moment of use. The requirement that someone other than the translator must compare source and target in full, against a defined set of parameters, is the translation profession's acknowledgement that surface quality is not the same as structural integrity. A translated procedure is not done when it reads well. It is done when a second qualified person has verified that it does what the original does, in the order the original does it, using terms that mean the same thing throughout — and has signed to that effect.