Test them on material from your own meeting: a short passage from your field, interpreted in both directions, heard once and recorded, then judged on content fidelity, target-language fluency and delivery.
As of
Author: Paul Dahoon Kim (Founder, MetaPret. Working Korean-English interpreter, ~100 engagements)
Last updated: 2026-10-08
Read time: ~6 minutes
Summary: Credentials describe an interpreter. A short recorded test shows what they can do. This guide explains what a meaningful skill check looks like, what a client can run without being an interpreter, and the red flags worth catching before the booking.
Test them on material from your own meeting. Ask for a recording of a short passage from your field, interpreted in both directions of the language pair and heard only once. Judge it on three things: did the full meaning come through, does the target language sound precise and natural, and would the delivery work in the room?
Why résumés, certificates, years and price don't show it
Each usual input answers a different question from yours.
- A résumé says where someone has worked. It does not show whether they can carry your speaker, in your field, at your meeting's pace.
- A certificate shows that the interpreter passed an exam at some point, usually in general language ability. It rarely covers your field, and it says nothing about how they perform today.
- Years of experience count time, not accuracy. Ten years of escort work at trade shows is real experience, and it is still a different job from consecutive interpreting in a licensing negotiation.
- Price reflects how the interpreter, or the channel that represents them, positions the service. A higher rate can follow seniority, demand or the channel's own cost structure. It is not a measurement of skill in your field.
All four come from the supply side, and none of them is a sample of the work. Interpreting skill is also domain-specific: an interpreter who is excellent on an investor call can lose the thread at a factory audit, where the vocabulary and the speakers are different. The only direct evidence is the interpreter actually interpreting something close to your meeting.
What a meaningful skill check looks like
A useful test is short. Five features separate a real check from a formality.
- A field-specific passage. The material should come from the domain of the engagement. An investor call gets investor-call language; a medical device training gets device and procedure language. A generic passage on tourism or current affairs tells you little about a board meeting.
- Both directions of the language pair. Most business meetings run both ways: your side speaks, the other side answers. Many interpreters are noticeably stronger into their dominant language. Testing only one direction hides the weaker half of the job.
- Heard once. In a meeting, nobody replays the speaker. A test where the interpreter can listen twice, pause, or read the text first measures translation, not interpreting.
- Recorded. A recording lets you, or a colleague, compare what was said with what came out, line by line, after the fact.
- Scored against a written rubric. Decide in advance what you are judging, so the result does not rest on a pleasant voice. Three axes cover most of it:
- Content fidelity: did the full meaning and structure come through? Watch for reversed meaning, wrong figures, dates or names, dropped points, and things the speaker never said.
- Target-language fluency: are field terms correct, and does the output read as natural target-language usage rather than word-for-word calque?
- Delivery: pace, tone, clarity, and long pauses. Would this sound right in your room?
What a client can do without being an interpreter
You do not need to be an interpreter to run a reasonable check.
Ask for a recording of a short field-specific passage. Request one take, interpreted on first hearing. A polished showreel from an unrelated event is not the same evidence.
Provide two to three minutes of real material from the meeting. An excerpt from a past call or a paragraph of the product pitch, read aloud, works well. Remove anything confidential first, or put an NDA in place before sharing. Real material is the closest thing to the meeting itself.
Check both directions. Send one clip in each language. If no one on your team speaks both, focus on what anyone can verify: numbers, dates, names, product names and the order of the points. Errors there are visible without full fluency.
Check terminology from your own glossary. Pick five to ten terms your team uses daily, including internal product names and acronyms. Hear how the interpreter handles them, and whether they ask about the ones they do not know rather than guessing.
Ask what the interpreter needs before the meeting. A strong answer names specific material: the agenda, slides, participant list, glossary, notes from earlier meetings. An interpreter who asks good questions about the content is showing how they will prepare.
Red flags
- No sample of any kind, or only a general demo with no connection to your field.
- "The field doesn't matter, interpreting is interpreting." It does matter, and an experienced interpreter knows it.
- Only one direction offered, usually into the interpreter's stronger language.
- A sample that was clearly edited or re-recorded, with no way to tell whether it was heard once.
- Vague answers on terminology, or confident guesses at terms they should have asked about.
- No request for preparation material, with the conversation focused entirely on rate and hours.
- No stated process for when things go wrong. Ask the interpreter or the channel what happens if the match falls short, before you book.
None of these alone proves an interpreter is weak. Together, they mean you are booking on trust rather than evidence.
How MetaPret does it
MetaPret runs this check before the match exists, so the client does not have to organise it.
For every request, MetaPret builds a short interpretation test from that engagement's field, in both directions of the language pair. Candidates take it under live conditions: a microphone check first, each passage played once with no replay, and recording that starts automatically when playback ends.
Each recording is scored against a published 15-point rubric with three axes, five points each: content fidelity, target-language fluency, and delivery. An AI monitor drafts itemized deductions, each quoting the source line and what the candidate said, and an expert evaluator listens to both directions before a score is confirmed. The AI does not decide pass or fail.
On a private comparison page, the client sees each confirmed candidate under a pseudonym, with the three axis scores and the total out of 15. Interpreters whose test was confirmed receive written feedback on their own recording. If a match falls short, contact cs@metapret.net within seven days and MetaPret will work with you to resolve it.
The full method is at MetaPret's testing method. To start, submit a request.
FAQ (FAQPage schema)
Q: How long should an interpreter skill test be?
A: Short is fine. Two to three minutes of real material in each direction is usually enough to hear content fidelity, terminology and delivery. Length matters less than the passage being from your field and heard only once.
Q: What if nobody on my team speaks both languages?
A: Check what anyone can verify: numbers, dates, names, product terms and the order of the points. For a fuller judgement, ask a bilingual colleague or partner to listen to the recording against the source.
Q: Is it reasonable to ask an interpreter for a test before booking?
A: Yes, as long as the request is short and specific. Send a brief clip, ask for one take, and be clear about how you will use the recording. An interpreter who is confident in the field has little reason to prefer being judged on a résumé over real work.
Q: Does a certificate or association membership replace a skill check?
A: No. Credentials are a useful signal for general ability. They do not test skill in your specific field, in both directions, on your material. Use them to shortlist, then check the work.
Q: Why test both directions if my side only speaks one language?
A: Because the meeting runs both ways. Your side's words go out in one direction and the other side's answers come back in the other. A weak return direction means you misunderstand what was offered or agreed.