Back Translation Services with Full Reconciliation

What Our Back Translation Service Actually Delivers

Back translation is a verification method, not a second translation project. After your document has been translated and revised through our standard workflow, a second, fully independent linguist who has never seen the source text translates the target version back into English. A third linguist, the reconciler, then places the back translation next to your original and examines every point where the two diverge. Some divergences are harmless artifacts of the method. Others reveal a real shift in meaning that must be fixed in the translation before anyone relies on it.

You receive four things: the back translation itself, a reconciliation report listing every discrepancy with its classification and resolution, the corrected final translation where changes were warranted, and a signed statement describing the procedure and the qualifications of each linguist involved. That last item matters more than clients expect. When a sponsor's clinical team or a notified body asks how a translation was verified, "we had someone look at it" is not an answer. A documented, blind, three-role procedure is.

We run this service most often for medical device translation projects, clinical research materials, and other content where a comprehension error puts a person or a filing at risk. It pairs naturally with certified technical translation when the end use also demands a formal accuracy statement.

Who Expects Back Translation, and Where It Is Written Down

No U.S. statute contains the phrase "back translation." The expectation lives one level down, in how oversight bodies interpret their mandates. Institutional Review Boards must verify that consent information is presented "in language understandable to the subject" under 21 CFR 50.20, and many IRBs have translated that duty into standard operating procedures that name back translation as the accepted verification method for informed consent forms. Sponsors follow suit because a rejected consent package stalls enrollment at every affected site.

The pattern repeats across the clinical world. ISPOR's good practice principles for translating patient-reported outcome measures (Wild et al., 2005) build back translation and reconciliation into the linguistic validation sequence for clinical outcome assessments, and licensors of copyrighted instruments frequently refuse to authorize a translated scale without evidence that the sequence was followed. ISO 14155:2020, which governs clinical investigations of medical devices, pushes sponsors toward documented translation verification for subject-facing materials. And beyond trials, we see back translation required by risk managers for device labeling and IFU claims, by legal teams before foreign-language safety warnings go to print, and by sponsors preparing a regulatory submission where a reviewer question about translation quality would be expensive to receive.

The common thread is accountability, not translation quality per se. These bodies want evidence they can file. Back translation produces evidence.

The Five-Step Process, with a Deliverable at Each Step

Our procedure is fixed, because a validation method that changes shape from project to project defeats its own purpose. Each step closes with a dated deliverable:

1. Forward translation and revision. Specialized translation plus independent revision under our ISO 17100-compliant workflow. Deliverable: the candidate translation. If your translation already exists, from us or anyone else, we start at step two.
2. Blind back translation. An independent linguist, native in the source language and screened for no prior contact with the project, translates the target text back into English with instructions to render meaning literally rather than elegantly. Deliverable: the back translation file.
3. Comparison and discrepancy logging. The reconciler aligns original and back translation segment by segment and logs every divergence with a severity class. Deliverable: the draft discrepancy log.
4. Reconciliation. For each logged item, the reconciler determines whether the translation, the back translation, or neither is at fault, consults the forward translator where intent matters, and specifies the correction. Deliverable: the reconciliation report with resolutions.
5. Final implementation and sign-off. Corrections are applied, the corrected translation is re-checked, and the signed procedural statement is issued. Deliverable: final files plus the complete validation package.

Independence is enforced, not assumed. Before assignment, we check that the back translator has no record on the forward project, no access to its translation memory or termbase, and no shared reviser with the original team; the assignment record naming all three roles goes into the validation package. Blinding covers reference materials too. The back translator receives the target file stripped of the English graphics, captions, and metadata that could leak source wording, because one visible English table header can quietly turn a blind exercise into a sighted one.

For a typical informed consent form of 3,000–4,000 words, the full cycle takes five to seven business days from receipt of the approved English source. Multi-language studies run in parallel: a recent 12-language consent project, anchored by our Spanish technical translation team for the largest patient population, delivered all validation packages inside two weeks.

Inside the Reconciliation Report

The report is where this service earns its fee, so it is worth showing what a line item looks like. Every discrepancy carries an identifier, the source passage, the back-translated passage, a severity class, a root-cause judgment, and a resolution. The severity classes we use:

Class	Meaning	Typical example	Action
Class 1: Meaning change	The translation says something different from the source	"May cause drowsiness" rendered as "will cause drowsiness"	Translation corrected, re-verified
Class 2: Omission or addition	Content missing from or added to the translation	A dosing condition dropped from a bulleted list	Translation corrected, re-verified
Class 3: Ambiguity	The translation permits two readings where the source permits one	Pronoun reference unclear in a two-device instruction	Translation reworded
Class 4: Back-translation artifact	The divergence originates in the back translation, not the translation	Literal re-rendering of a correct idiomatic choice	Documented, no change

Class 4 findings usually outnumber the rest combined, and that is normal. A back translator working literally will flatten every stylistic decision the forward translator made. The reconciler's judgment, distinguishing an artifact from a genuine defect, is the skilled core of the service and the reason we staff reconciliation with senior linguists rather than project coordinators. On patient-facing work, an in-country clinician reviewer can be added to adjudicate medical terminology in the target language.

The Honest Limits of the Method

Fifteen years of running validations have made us direct about what back translation cannot do, because overselling a quality method is its own kind of quality failure. The method is built to detect meaning-level errors: mistranslations, omissions, additions, ambiguities. It is nearly blind to everything else. A translation can be stilted, hard to read at the target literacy level, terminologically inconsistent, or tonally wrong for patients, and still back-translate perfectly. Readability problems in a consent form are a real enrollment risk that this method will not surface.

It also inspects a sample of one rendering chain. A skilled back translator can unconsciously repair a slightly defective translation on the way back to English, hiding the defect. Our literal-rendering instructions and independence screening reduce that risk; nothing eliminates it. And the economics are what they are: you are paying for a second full translation plus senior reconciliation time, so expect roughly twice the cost of translation alone and added calendar time. For content where the risk does not justify that spend, better instruments exist, which is the subject of the next section. For content where it does, informed consent, dosing instructions, safety-relevant labeling, the cost is small against a single enrollment suspension or field action.

When Expert Review Is the Better Buy

Because we sell both services, we can compare them without a thumb on the scale. A bilingual expert review puts a senior subject-matter linguist directly on the translation, reading it against the source and scoring every error by type and severity. It catches the meaning errors back translation catches, and it also catches the fluency, register, and terminology problems back translation misses. It costs less, typically 25–35% of translation cost rather than 100%, and runs faster. Our third-party translation review service delivers it with a scored quality report.

So why does back translation exist at all? Because expert review produces an opinion, while back translation produces an artifact a non-speaker can inspect. A U.S. study coordinator who reads no Vietnamese can still hold the English back translation of a Vietnamese consent form next to the original and see for themselves that the risks section says what it should. When the audience for your quality evidence cannot read the target language, that property is decisive. Our working rule: regulator or IRB involvement, patient-facing risk content, or a licensor mandate points to back translation; internal quality assurance, vendor evaluation, and engineering content point to expert review. Documents like validation protocols, read by trained professionals rather than patients, usually sit in the second group unless a specific submission says otherwise.

Pricing and Scheduling the Service

We price back translation as its own line items so you can see where the money goes: back translation at our standard per-word specialist rate, reconciliation and reporting at an hourly rate that scales with discrepancy volume, and optional clinician review as a flat per-document fee. As a working figure, budget 90–110% of the forward translation cost for the full validation cycle. Volume behaves differently from normal translation: translation memory discounts do not apply to the back translation pass, because reusing memory would defeat blindness, but reconciliation of a document family with shared boilerplate goes fast, and we pass that time saving through.

Scheduling has one rule worth planning around. Validate after the English source is final. A revised sentence in the source reopens the entire chain for the affected passages, and we re-bill only those passages, but the calendar cost of a reopened validation is harder to contain than the invoice. Sponsors who lock the master document, then launch all languages at once, get the smoothest timelines. The biotechnology and lab equipment manufacturers we work with often coordinate this around a design freeze or a protocol amendment date, and we schedule linguist availability against that date in advance.

One anecdote shows the stakes. A device maker asked us to validate the Spanish IFU for an infusion accessory after their notified body queried translation quality. The back translation surfaced a Class 3 finding: the flushing instruction could be read as "flush before each use" or "flush before first use." Clinically those are different behaviors. The fix took one sentence; finding it took a method designed to look for exactly that. The audit query closed with the reconciliation report attached, and the client now validates every patient-facing document line before release. If you are weighing which of your documents deserve the same treatment, this page lists the adjacent quality services we combine into a risk-based program.