You post a job for a post-editor. A dozen translators apply - experienced, with solid portfolios and good references. You hire the best one. Three days later, the client is asking why the MT output came back completely rewritten. The translator is polishing every sentence from scratch, producing beautiful text - but at 200 words per hour instead of the 700 you needed.
That’s the most expensive mistake in MTPE staffing: treating post-editing as a subset of translation. It’s not. Post-editing is its own profession, with skills that overlap with translation but also skills that directly contradict how a good translator thinks.
Why Good Translators Often Make Bad Post-Editors¶
A translator’s core job is to produce the best possible text. Their entire professional identity is built around finding the exact right word, crafting natural sentences, expressing nuance precisely. When you hand them a piece of MT output, their instinct is to improve it - because that’s what they’re trained to do.
Post-editing requires the opposite instinct: leave it alone if it works.
This isn’t a small adjustment. The ISO 18587:2017 standard - the international benchmark for machine translation post-editing - defines “full post-editing” as producing text that meets publication quality, but it explicitly requires post-editors to avoid unnecessary changes and accept text that is “fit for purpose” even if it isn’t the translation they’d have written from scratch.
The problem is that most translators can’t do this. As the ATA has noted, editors often “misunderstand their task to be aligning the translation to what they would have done as translators.” Editing becomes a way to prove they’re better - not a way to reach the quality bar efficiently.
This phenomenon has a name: over-editing. It’s the single biggest hiring risk in MTPE.
The Four Skills Post-Editors Need That Translators Don’t¶
Both roles require bilingual competence, domain knowledge, and research skills - that’s table stakes. What separates a skilled post-editor from a skilled translator is the additional layer on top.
1. Error classification, not error correction¶
A translator encounters a problem and fixes it. A post-editor encounters a problem and first asks: does this actually need fixing?
Post-editing operates at two quality levels:
- Light post-editing (LPE): fix only errors that make the text unintelligible, factually wrong, or offensive. Style and word choice are not your concern.
- Full post-editing (FPE): fix everything that falls below publication quality - but only what falls below that bar, not what you’d do differently.
Good post-editors classify errors before correcting them. Is this an accuracy error (wrong meaning)? A fluency error (grammatically wrong but meaning is intact)? A terminology error (wrong term from the glossary)? A style issue (acceptable but not preferred)? Each type warrants a different response depending on the quality level specified in the brief.
The interview question that separates post-editors from translators: “The MT output is correct in meaning but phrased awkwardly. The brief says light post-editing. What do you do?” A translator says they’d fix it. A post-editor says they’d leave it.
2. MT pattern recognition¶
Every MT engine makes predictable mistakes. A post-editor who knows the engine’s failure patterns can spot errors in a fraction of the time it takes someone reading the text purely for quality.
Common MT failure patterns to test for:
| Error type | Example | Impact |
|---|---|---|
| Dropped negation | “not required” → “required” | Critical - meaning reversal |
| Number/unit conversion errors | “5 km” → “5 miles” | Critical |
| False cognates | “sensible” in source translated as “sensible” in target (wrong meaning) | High |
| Inconsistent terminology | Same term translated 3 different ways | High |
| Literal translation of idioms | Meaning lost when idiom is translated word-for-word | Medium |
| Pronoun ambiguity | “it” resolved to the wrong gender in inflected languages | Medium |
| Missing articles or prepositions | Common in pairs with structurally different grammar | Medium |
An experienced post-editor doesn’t read MT output the way they read human text - they run a pattern-matching scan, checking known failure zones before deep reading. This is a learned skill, not a natural one. You can test for it by asking candidates to describe the specific MT failure types they watch for and whether they adjust their approach based on the MT engine used.
3. Speed under quality constraints¶
A professional translator averages 2,000-2,500 words per day - roughly 250-350 words per hour when you account for research, formatting, and review. Post-editing benchmarks are different:
| Type | Expected speed | Notes |
|---|---|---|
| Light PE (good MT quality) | 600-1,000 words/hour | Intelligibility fixes only |
| Full PE (good MT quality) | 500-700 words/hour | Publication-ready output |
| Full PE (poor MT quality) | 250-400 words/hour | May not be worth it vs. retranslation |
| Human translation (no MT) | 250-350 words/hour | Baseline for comparison |
According to research on neural MT post-editing, individual productivity gains from MT assistance range from 6% to 56% depending on the post-editor. The top gains go to people who’ve fully internalized the PE mindset. The 6% gains typically go to over-editors who redo the MT from scratch.
When hiring, set productivity expectations explicitly in the brief and the assessment. A candidate who can’t hit 500 words/hour on full PE with decent MT quality isn’t a post-editor - they’re a translator who happens to have MT in front of them.
4. Psychological comfort with “good enough”¶
This is the hardest skill to screen for because it looks like lowered standards. It isn’t.
Post-editors need to internalize “fit for purpose” as a genuine quality goal, not a compromise. The MT output doesn’t have to be the text they’d have written. A product description for a secondary market doesn’t need the same standard as a legal agreement. A light PE pass on internal documentation is done when the meaning is clear - not when every sentence is graceful.
Translators who struggle with MTPE often describe it as frustrating or demoralizing: they can see ways to improve the text but aren’t supposed to. That discomfort is a signal they’re fighting their training. It’s not a character flaw, but it is a genuine mismatch with high-volume PE work.
As one translator noted in a 2025 industry survey on MTPE:
“MTPE is not replacing human translators, but it’s changing the way they work. The hardest part isn’t the language - it’s accepting that ‘good enough’ actually is good enough.”
That sentence describes the core mindset shift precisely. When you’re interviewing, ask about this directly: “Describe a time when you had to leave a translation you could have improved.” The answer tells you a lot.
What to Look For on a CV and Profile¶
You can identify likely post-editors before the test stage. Markers that matter:
Positive indicators: - MTPE-specific experience mentioned explicitly (not just “familiar with MT” but worked on PE projects with volume and deadlines) - CAT tool proficiency with MT integration: memoQ with DeepL, Trados with Language Weaver, Phrase, Smartcat - Stated experience with light vs. full PE distinction - candidates who know the difference have actually done this work - Subject-matter specialization in content that MT handles well: technical documentation, e-commerce, legal boilerplate - Any mention of quality metrics or frameworks: MQM, DQF-TQA, error classification - High-volume, deadline-driven project experience
Red flags: - Portfolio focused on literary translation, creative adaptation, transcreation - these people are trained for exactly the opposite mindset - CV emphasizes “meticulous” and “thorough” as core traits with no mention of speed or volume - No CAT tool experience at all - Only worked on sworn, certified, or legally consequential translations where over-editing is appropriate and expected
Don’t penalize candidates for not having extensive PE experience - MTPE adoption only crossed 46% industry-wide in 2024. Many capable professionals are new to the workflow. Do check for the right attitude and technical setup.
How to Design a Post-Editing Assessment¶
A standard translation test doesn’t work for post-editors. If you give candidates a source text and ask them to translate it, you’ll learn about their translation skills - not their post-editing ones.
The right assessment structure:
Step 1: Give them raw MT output, not a source text to translate
Provide the MT output already generated from the source. Include the source for reference, but make clear the task is to post-edit the MT - not retranslate it.
Step 2: Specify the quality level explicitly
“This is a light post-edit. Fix only errors that affect intelligibility or factual accuracy. Do not make preferential style changes.” For full PE: “Fix everything that falls below publication quality based on the attached style guide.”
Step 3: Seed the MT output with specific errors
Drop one negation, insert one number error, add two or three false cognates and formulaic translations - and also include several awkward but acceptable phrases that don’t technically need fixing. See which errors they catch, and critically, whether they change the acceptable-but-imperfect phrasings unnecessarily.
Step 4: Time the task
600-800 words of MT output is enough for an assessment. Full PE of 700 words should take 60-90 minutes. If a candidate spends 3 hours, they’re over-editing.
Step 5: Review the diff, not just the output
Don’t evaluate only the finished text. Track every change the candidate made and classify each as: necessary correction, defensible improvement, or over-edit. An over-edit rate above 20-30% is a warning sign for high-volume PE work.
On ProZ.com assessment discussions, experienced project managers consistently note that candidates who ask about the quality level before starting almost always outperform those who just begin editing without clarifying the brief.
Rates and Output Expectations: Set Them Explicitly¶
MTPE rates are lower than full translation rates because the volume per hour is higher. The current market range for full post-editing is $0.05-$0.15 per word, compared to $0.15-$0.30 per word for full human translation - roughly 50-75% of standard rates.
A practical structure when setting rates:
| Scenario | Rate model | Justification |
|---|---|---|
| High-quality MT, routine content | $0.06-$0.10/word for full PE | Speed advantage is real; discount is fair |
| Medium-quality MT, technical content | $0.10-$0.13/word | More corrections needed, closer to standard rates |
| Poor-quality MT or complex content | Retranslate at standard rate | PE adds no efficiency advantage |
About 50% of translators don’t offer MTPE discounts at all, and they’re not wrong that poor MT quality can make PE as slow as translation. The key is transparency: tell candidates the MT quality before they quote. Agencies that send PE briefs without disclosing the MT is poor burn relationships and get worse output.
Also communicate productivity expectations upfront. If your project needs 4,000 words/day per post-editor, say so before the contract is signed.
When to Use a Post-Editor vs. a Translator¶
Not all content suits post-editing. Staffing correctly from the start means matching the tool to the content type.
| Content type | Best approach | Why |
|---|---|---|
| Technical docs, user manuals, legal boilerplate | Full PE | MT handles structured text well; efficiency gains are real |
| E-commerce product descriptions (high volume) | Light PE | Speed matters more than stylistic polish |
| Marketing copy, taglines, brand voice | Human translation | MT can’t capture tone and cultural resonance |
| Medical instructions, drug labels | Human translation or full PE with subject expert | Accuracy stakes too high for light PE |
| Literary or creative content | Human translation | MT fundamentally misses literary devices |
| Internal communications, informal content | Light PE or raw MT with review | Low stakes; speed dominates |
| Sworn/certified translations | Human translation | Legal requirement for a human certifying the output |
The mistake many agencies make is assigning post-editors to content that fundamentally needs a human translator, then wondering why quality is disappointing. The post-editor isn’t failing - the content type is wrong for the workflow.
FAQ¶
What’s the main difference between a post-editor and a translator?¶
Both need native-level target language fluency and strong bilingual competence. A translator produces text from scratch. A post-editor corrects existing MT output, working at higher speed and within explicit quality constraints (light or full PE). The key mindset difference: translators optimize for the best possible text; post-editors optimize for text that meets the brief as efficiently as possible.
Does a post-editor need to be a qualified translator?¶
Under ISO 18587:2017, a post-editor must meet the same competence level as a professional translator under ISO 17100 - a translation or linguistics degree, or equivalent experience (5+ years of full-time translation practice). The standard adds post-editing-specific competencies on top, not instead of the translator baseline.
How can I tell if a candidate will over-edit?¶
The most reliable method is the seeded MT assessment - giving raw MT with deliberate acceptable imperfections and tracking whether candidates change them. In the interview, listen for how they describe their editing philosophy: “I produce the best text possible” is a red flag. “I work to the quality level in the brief” is what you want to hear.
What CAT tools are standard for MTPE?¶
Most professional post-editing happens in CAT tools with MT integration: memoQ, Trados Studio, Phrase (formerly Memsource), Smartcat, and XTM all support this workflow. The CAT provides a segment-by-segment view with MT suggestions pre-filled; the post-editor accepts, edits, or replaces each segment. Candidates without CAT experience will be significantly slower even if their linguistic instincts are good. For a comparison of major CAT tools, see our Trados vs MemoQ vs Smartcat overview.
What productivity should I expect from a post-editor?¶
For full post-editing with decent MT quality: 500-700 words per hour. For light post-editing: 600-1,000 words per hour. These translate to roughly 3,500-5,000+ words per day for full PE - compared to 2,000-2,500 words per day for human translation. The gains depend heavily on MT quality: poor MT eliminates the speed advantage entirely.
Should I hire in-house post-editors or freelancers?¶
For steady-volume MTPE programs, in-house PE specialists build institutional knowledge of your specific MT engine’s failure patterns over time and get faster. Freelancers give you flexibility for variable volume but require more briefing overhead. Many agencies run a hybrid: a small in-house PE lead who trains, sets quality standards, and reviews output, with a freelancer pool for volume overflow. See our guide to the hybrid AI-human translation workflow for how agencies are structuring this in practice.