ZenTrans
b2b law-firms vendor-management certified-translation

How Immigration Law Firms Should Evaluate a Translation Vendor

A mistranslation costs a firm an RFE and months of delay. The criteria that actually predict vendor reliability on immigration filings โ€” and the questions procurement usually forgets to ask.

7 min read ZenTrans

For an immigration practice, a translation vendor is a risk position, not a line item. A single mistranslation triggers a Request for Evidence, and an RFE delays a case by months โ€” an unacceptable outcome when a paralegal is running dozens of matters against fixed deadlines.

Yet vendor selection is usually run on price and turnaround, which are the two criteria that predict reliability least well. This piece covers what actually does.

Written for firms evaluating vendors โ€” and equally useful to LSPs who want to know what a serious buyer will ask.

Start with the destination matrix

The first question is not about quality. It is about credential coverage, because a translation can be linguistically perfect and still be refused.

Ask the vendor to state, in writing, which of these they can deliver:

DestinationRequirement
USCISCertification with competence statement
UKVICertification with translator credentials and contact
IRCCProvincially certified translator or sworn affidavit
Home AffairsNAATI Certified Translator, current, correctly directed
IQAS / provincial ECACertified translator with seal
Germany, France, Spain, ItalyCourt-appointed sworn translator

A vendor who answers "we do certified translation for any country" has not understood the question. The credentials are structurally different and a US-oriented provider frequently cannot deliver IRCC or NAATI work.

If your practice files in more than one country, this table alone eliminates most candidates.

Ask about language granularity

The failure that reaches your file as an RFE usually starts as a staffing shortcut. Ask directly:

  • Do you distinguish Dari and Pashto as separate certifier competencies?
  • Do you distinguish Hindi and Punjabi by script?
  • How do you handle Chinese romanisation for Hong Kong and Taiwan-origin documents?
  • How is a certifier's language competence recorded and verified?

A vendor whose answer is "we cover Persian" when you asked about Pashto is telling you their pool is indexed too coarsely for immigration work.

Test structural fidelity, not prose quality

Most vendor sample tests use a prose document, which tests the wrong thing. Immigration documents are structured artefacts.

A better test piece: a scanned civil document with a table, a seal, a marginal annotation, one blank field, and one genuinely illegible word.

What you are checking:

  • Is the table reproduced as a table, with values under the correct headers?
  • Is the seal text translated, or silently skipped because it is an image?
  • Is the marginal annotation present โ€” the place where divorce, legitimation, and corrections are recorded?
  • Is the blank field marked blank, or omitted?
  • Is the illegible word marked [Illegible], or guessed at?

That last one is the character test. A vendor who guesses on a faded village name will guess on a date of birth.

Ask the AI question properly

Most vendors now use AI somewhere in production. "Do you use AI?" produces a defensive non-answer. Better questions:

  1. Who signs the certification, and are they named on it?
  2. How do you ensure the certifier actually reviewed the output โ€” is review a workflow gate or a policy statement?
  3. What is your average review time per document?
  4. How do you route documents your tools handle poorly โ€” degraded scans, heavy handwriting, unusual scripts?

The regulatory position is about accountability, not tooling. IRCC refuses machine-generated translations; USCIS, UKVI, and Home Affairs each require an accountable human by a different route. A vendor with a clear three-part answer โ€” what the tool drafts, what the human reviews, who carries accountability โ€” is lower risk than one claiming a purely manual process at a price that makes it implausible.

The genuine risk is a certifier signing work they did not read. Ask how that is prevented structurally. See ai-in-certified-translation-workflows.

The independence trap

One constraint that catches firms directly: the certifier must be independent of the applicant. USCIS treats the applicant as an interested party. Australian Home Affairs excludes the applicant, family members, and the migration agent.

That last exclusion means a firm cannot have its own bilingual staff certify a client's documents for an Australian filing. Confirm your vendor applies an independence check at assignment rather than assuming it.

Consistency across a case file

An RFE cause specific to law firm work: names spelled inconsistently across documents in one bundle.

Ask how the vendor maintains consistency across a multi-document file:

  • Is a name and terminology table built per matter?
  • Is the romanised form locked to the passport spelling across every document?
  • If documents arrive in batches over weeks, how is consistency preserved between batches?

A vendor treating each document as an independent job will produce a bundle that contradicts itself. For a family sponsorship file with sixty documents across four languages, this is the single highest-probability failure.

Operational criteria that actually matter

Per-document pricing transparency. Immigration bundles are many short documents. A per-page rate with an undefined page can double at invoice. Get the page definition in writing โ€” 250 words or part thereof โ€” and ask how transaction-row documents such as bank statements are counted.

Rework policy. If a filing is rejected on translation grounds, what happens? Free correction is the minimum; ask about turnaround on rework, since you will be inside a response deadline.

Deadline behaviour under surge. Ask for their declared daily document capacity per language. A vendor who will not state a ceiling will accept work they cannot deliver.

Escalation contact. A named person reachable inside a response window, not a ticket queue.

Retention and reissue. Can they reissue a certification if a caseworker questions the format, months later? Do they retain the file?

Questions procurement usually forgets

  • What is your rework rate on immigration filings, and how do you measure it?
  • Have you had a translation rejected by USCIS, UKVI, IRCC, or Home Affairs in the last year? What was the cause and what changed?
  • Who verifies credentials, how often, and against which registry?
  • What happens when a NAATI credential expires mid-project?
  • Do you provide a scanning specification to clients, and do you gate intake on completeness?

That last one is a good proxy for operational maturity. A vendor who catches a missing reverse side at intake rather than at review is a vendor whose process is designed rather than improvised.

Frequently asked questions

Is the cheapest vendor a false economy? Not automatically. Risk comes from credential mismatch, incompleteness, and unreviewed output โ€” not from price. But a price implying no real review time is a signal worth probing.

Should we use one vendor or several? One primary with verified multi-destination coverage reduces consistency risk across a bundle. Keep a second for surge and rare languages.

Can we use our own bilingual staff? Not for certification. Staff are not independent of the client relationship, and for Australian filings the migration agent is explicitly excluded.

How do we handle a rejection caused by the vendor? Classify the defect first โ€” many are fixable by reissuing a certification rather than redoing the translation. Then log the cause; a pattern across matters is the actionable signal. See immigration-document-qa-checklist.


Evaluate on credential coverage, language granularity, structural fidelity, and accountability โ€” in that order. Price and turnaround are real, but they are not what causes the RFE.

Preparing a vendor evaluation, or answering one? Open ZenTrans and run a sample batch โ€” the first 5 pages are free, and volume pricing runs to $0.70 per page.