A post-editor and a translator are two completely different professions, even though both work with text. It’s like a designer and a UI designer: both design, but one creates from nothing and one optimizes what exists. Many project managers miss this — they hire a translator, give them a machine-generated draft, and get frustrated when quality drops and timelines don’t improve. Result: projects slip, costs balloon, teams burn out on the wrong person instead of starting fresh.
This guide explains end-to-end: how to pick the right person, where to find them, how to vet them, and how to set them up for success. We’ll cover post-editing vs. translation not just as concepts, but as practical project management — concrete rates, platforms, skills, and onboarding timelines. By the end, you’ll have a hiring roadmap that works: fast, high-quality, and on budget.
What’s post-editing and how is it different from translation?¶
Post-editing is not “faster translation.” It’s a separate process with separate psychology.
A translator starts with a blank page, reads the source across six pages, understands the context, then writes a translation from scratch, applying full linguistic creativity, intuition, and domain knowledge. They can rewrite sentences for naturalness, choose synonyms, rephrase complex constructions. Their job is to create text that sounds like the original, not like a translation.
A post-editor receives an already-completed text from AI or a neural model. Their job is not to retranslate—that’s the biggest mistake untrained people make. Their job is to identify errors in the machine draft and fix them with minimum intervention. They see a segment (usually one sentence)—the machine output, the source alongside—and ask: “Is this already a reasonable translation, or is there an error that needs fixing?”
This sounds simple, but it fundamentally changes the approach. As Yardstick explains, a translator develops “the joy of being an author,” while a post-editor develops “the joy of spotting an error and fixing it fast.” These are psychologically different drives.
Let’s look at three core differences.
1. Direction and logic of work
Translator: blank page → reads source → understands topic → writes translation → finished text.
Post-editor: black page (finished ML translation) → reads source + ML variant side-by-side → asks “is this an error?” → corrects → finished text.
The difference: translator creates, post-editor filters. The cognitive load is vastly different. Translator is 100% attention on creativity. Post-editor is 70% identifying errors, 30% fixing them.
2. Speed — concrete numbers
Post-editors work 30-50% faster than traditional translators on the same content. For example: - Traditional translator: 2,000-2,500 words per day of quality translation - Post-editor with solid ML model: 4,000-6,000 words per day
And an experienced post-editor with a well-trained model can hit 2x productivity—5,000 words per hour—when AI is already 80%+ correct.
Why? Because post-editors don’t write text—they read already-written text. Reading is 2-3x faster than writing.
3. Psychology of thinking
Translator constantly asks: “How does this sound better? Which word should I choose? Would rewording be more natural?”—creative mode.
Post-editor asks: “Is this already OK, or is it an error?”—quality control mode.
These are different skills. A creative translator often makes a poor post-editor, because they’re used to rewriting everything, but post-editing demands minimal intervention. Conversely, an analytical, detail-oriented person often excels as a post-editor more than as a translator.
Core skills a post-editor must have¶
Many managers assume: “If someone’s a good translator, they can post-edit.” Wrong, and it’s expensive. According to elit-asia research, an untrained person drops quality 20-40% when switching from translation to post-editing.
That’s not just theory—it happens routinely. People are used to checking everything, rewriting, polishing. In post-editing, that steals time and worsens quality because they make unjustified changes to already-good text.
Here are four essential skills you need:
1. MT error recognition — systematic error patterns, not old habits
Machines make the same mistakes. Examples: - Context misunderstanding: “The crane was heavy” translates as “the bird was heavy” instead of “the equipment was heavy” - Word omission: “The document must be reviewed and signed” → “The document must be… signed” (missed “reviewed”) - Number errors: “15%” → “1.5%” (wrong punctuation) - Duplication: “The process is simple simple” → text repeats unnecessarily - Wrong dependency tree: “The book about cats written by the author” → grammatically wrong binding
A good post-editor learns to spot these 5-10 core error types in the first week. Someone new to post-editing often either fixes the wrong thing (changes correct text) or misses subtle errors. This improves over time with training.
2. Contextual judgment and terminology awareness — looks hard, but it’s solvable
Post-editors can’t read the full source document like translators do. They see one segment (usually one sentence) and the machine output. They need to infer context from that single sentence.
Bad example: if ML said “Interface” → “Интерфейс,” that’s correct for IT text but wrong for medical text (where “interface” = “contact surface”). The post-editor must feel this from one sentence of context.
This requires domain expertise. For contracts, you know legal terminology and exceptions. For medical text, you understand anatomy, chemistry, pharmacology. Without it, you have questions every third line, and productivity drops in half.
Solution: hire post-editors with specialization. Request “English-German legal post-editor,” not just “English-German.” Pay 15-20% more for specialization—it pays for itself through quality and speed.
3. CAT tool proficiency — critical for speed
Post-editors rarely work in Word alone. They work in a CAT system (Computer-Assisted Translation). Tools like Trados, memoQ, Memsource, Wordfast have specialized interfaces for post-editing: source on the left, ML output in the center, the post-editor in a special mode navigating segment by segment with keyboard shortcuts.
If someone doesn’t know these tools, all time savings vanish. Instead of 5,000 words/day, they do 2,000, manually copying and pasting, hunting for segments, losing context.
Non-negotiable. When you hire a post-editor, ask: What CAT tools do you use? If the answer is “I don’t know,” don’t hire.
4. Efficiency and fatigue management — mental stamina
Post-editing is more monotonous than translation. Your brain adapts to context, and you start missing errors. Classic problem: sharp focus at the document start, but by middle/end, fatigue kicks in and errors slip through.
Called “post-editor fatigue” or “editing fatigue.” A good post-editor knows how often to take breaks (usually every 1-2 hours), how to double-check output (second pass through), and how to switch between projects (different context keeps the brain engaged).
Translators know this too, but for post-editors it’s critical because errors are harder to spot. A translator hears wrong rhythm. A post-editor reads someone else’s text—errors can hide.
Real adaptation math:
If you hire someone without MTPE training: - Days 1-3: understanding workflow, tool setup. Unproductive yet. - Days 4-7: first project, quality 60-70%, speed 50% expected. You spend 2 hours daily giving feedback. - Week 2: quality 75-80%, speed 70-75%. - Weeks 3-4: quality 85-90%, speed 90%+. Often “clicks” here.
Average: 3-5 days real adaptation time. 15-25% quality loss in month one is normal.
Where to find post-editors and how to vet them¶
Three main platforms:
ProZ.com — The most popular site for translators and post-editors. Over 500K professionals. Filter for “MT Post-Editing” among job types. Mostly freelancers with ratings and portfolios.
How to use: - Enter “MT Post-Editing” in search - Filter by language pair (e.g., “English to German”) - Sort by rating (min 4.8/5) - Review last 10-20 projects: project types, client feedback, timing, volume
Cost: $0.05 to $0.15 per word. Experienced post-editors with solid ratings ask $0.10-0.12. Junior or rare languages: $0.05-0.08.
Pros: large pool, easy filtering, transparent ratings. Cons: competition is fierce, good people often fully booked.
TranslatorsCafe.com — An alternative to ProZ, more budget-friendly. Often junior post-editors building experience at lower rates ($0.04-0.10/word). Quality varies, but for non-urgent work, you find good deals.
How to use: Similar to ProZ, fewer filters. Requires more portfolio review time. Ratings less organized, but people respond faster. Typical response: 2-4 hours.
Pros: cheaper, often newer people with potential. Cons: quality less predictable, less MTPE specialization.
Smartcat Marketplace — For large-scale hiring. If you need 50+ post-editors simultaneously (high volume), they auto-match by language, experience, rate, and reviews. Platform distributes work automatically. Cheaper than hiring individual freelancers locally.
Pros: auto-scaling, built-in QA. Cons: less control over individuals, platform fees 20-30%.
Which to pick? Rule of thumb: - < 5,000 words/month: ProZ or TranslatorsCafe, 1-2 people - 5K-20K words/month: ProZ or Smartcat depending on volume stability - > 20K words/month: Smartcat or local full-time hire
How to vet a candidate — checklist¶
1. MTPE certification — Ideal if they’ve completed formal training. Look for: - RWS Linguistic AI Certification (free, 3-4 hours, most popular) - Trados Post-Editing Certification (Trados-specific) - ProZ Self-paced MTPE Training ($50-100, self-paced) - Meridian Linguistics MTPE course ($200-400, most comprehensive)
Any of these = +20 likelihood points. No cert doesn’t disqualify—just means more QA on early projects.
2. Portfolio or test project — Most important.
Request 500-1,000 words of test post-editing in your language pair. Deliberately include 3-5 obvious errors (wrong word, omission, number). Pay $20-40—it’s cheap insurance against a bad hire. Wrong hire costs 100x more over time.
What to check in results: - Did they catch the errors correctly? (primary) - Did they retranslate unnecessarily? (check % of changes—30%+ is a red flag) - Did they ask for context via email? (plus sign—shows caution) - Can you see they used a CAT tool? (ask which one) - What error types did they miss? (numbers? all critical ones?)
Record findings. You’ll see the same patterns in real work later.
3. Interview or project history — Ask 5 minutes about their last MTPE project: - “How many words was the project?” - “What language pair?” - “Which CAT tool did you use?” - “What quality did you achieve? (How did you measure it?)” - “What was most challenging?”
Good answers sound confident, specific, metrics-based. Bad answers sound vague: “I translated well,” “I worked fast,” “I freelanced long ago.” Shows self-awareness.
4. Ratings on ProZ/TranslatorsCafe — Min 4.8/5 on last 20 projects. Lower = risky.
Read client comments, especially on MTPE projects. Good: “Fast, high quality, understood context,” “First submission needed no changes,” “Asked clarifying questions when ambiguous.” Bad: “Required many revisions,” “Didn’t understand context,” “Slow,” “Mechanical approach.”
Training and certification of post-editors — best courses 2025¶
If you have a translator wanting to transition to post-editing, or a new hire needs training, several standard courses exist. Not all are equal.
Formal courses with certification¶
RWS Linguistic AI Certification (free) — For linguists and PMs. Official RWS course.
Covers: - MT history and why machines error - MT error type recognition - Quality assessment (how to know 60% vs 80%?) - MTPE quality levels (light vs full vs refined) - Maximizing productivity - Post-editing psychology (fatigue, attention, focus)
Time: 3-4 hours instruction + 1 hour test = 4-5 hours total. Can complete in a weekend.
Certificate: valid 3 years, industry-recognized. Respected on resumes.
Cost: $0. Completely free.
Best for: newcomers to post-editing, translators transitioning, managers overseeing post-editors.
Trados Post-Editing Certification — More applied, tool-specific.
Covers: - Using Trados for post-editing - Workflow in Trados - Reading QA reports (when Trados flags errors) - Productivity shortcuts - Translation memory integration
Time: 6-8 hours over 2-3 days. Better done gradually.
Cost: $100-150 (through Trados).
Best for: people already familiar with Trados, or companies using Trados.
ProZ Self-Paced MTPE Training — Flexible course on ProZ. Not super authoritative, but practical.
Covers: - MT and post-editing basics - Error recognition - Various CAT tools (not just Trados) - Language-specific tips (English, German, Spanish) - Hands-on exercises with real texts
Time: 5-7 hours. Very flexible—30 min/day possible.
Cost: $50-100.
Best for: people unsure about post-editing, or budget-conscious options.
Comparison: post-editor vs. translator vs. raw AI — when to use each¶
| Factor | Raw AI | AI + post-editor | Traditional translator |
|---|---|---|---|
| Speed | 1x (baseline) | 0.5-1.5x (depends on AI) | 2.5-3x slower |
| Quality | 50-75% | 85-95% | 95-98% |
| Cost per word | $0.002 | $0.04-0.12 | $0.10-0.30 |
| Critical documents | Don’t use | Full QA only | Yes |
| High volume | Fast, risky | Optimal | Practical |
| Rare languages | Poor | Needs qualified editor | Hard to find |
| Onboarding | None | 1-2 weeks | None |
| Oversight | Minimal | Minimal | Moderate |
When to pick each¶
Pick raw AI if: - Volume < 1,000 words once or twice monthly - Low-priority text (internal comms, rough notes) - You need time to think before deciding - Budget is extremely tight
Understand: quality is 50-60%, still needs substantial manual fixing. Service savings don’t justify poor quality.
Pick AI + post-editor if: - Volume 5,000+ words/month in one language - Important but non-critical text (marketing, product descriptions, internal docs) - You can wait 1-2 weeks for onboarding - You already have an ML model at 75%+ quality (if worse, post-editor can’t help much)
Advantages: 30-40% time savings, 85-95% quality, easy to scale volume.
Pick traditional translator if: - Text is critical: legal docs, medical texts, contracts - Language pair is rare or specialized (Esperanto, Occitan) - Volume < 2,000 words/month (onboarding doesn’t pay off) - You need creative adaptation (marketing, brand guidance)
Advantages: 95-98% quality guaranteed, no onboarding time, creative approach.
Concrete selection calculator¶
Example 1: Technical docs, EN-DE, 10K words/month - Traditional translator: 10K × $0.15 = $1,500/month - Post-editor: 10K × $0.08 = $800/month + 40 hours onboarding month one - Savings: $700/month, breakeven by month 3.5 - Verdict: post-editor wins
Example 2: Legal contract, DE-UK, 500 words, urgent - Traditional translator: 500 × $0.20 = $100 - Post-editor: 500 × $0.10 + onboarding (no time) = impossible - Verdict: traditional translator only option
Example 3: Website marketing, EN-ES, 3K words/month - Traditional translator: 3K × $0.18 = $540/month - Post-editor: 3K × $0.09 = $270/month + 40 hours onboarding = $570 month one - Savings: $270/month from month two, but month one costs more - Verdict: depends on project length. < 6 months = translator. > 12 months = post-editor.
General rule: - Project > 6 months at 5K+ words/month in one language → invest in post-editor - Project < 6 months or rare language → pick translator - Critical text → always translator
Step-by-step hiring guide — A to Z¶
Here’s a real workflow that works. Times are conservative estimates.
Day 1: Planning¶
Task: Define scope, languages, budget.
Questions for yourself: - How many words/month need post-editing? (Be honest. If 4K, is it really worth a post-editor’s onboarding?) - Which languages? (Rare EN-LT or common EN-DE?) - Which domains? (Technical, medical, legal, marketing?) - What’s your MT quality starting point? (70%? 50%? Unknown?) - What rate can you afford? ($0.05? $0.15?)
Write this down. It’s your search criterion.
Time: 30 minutes.
Days 2-3: Candidate search¶
Task: Find 3-5 prospects.
Algorithm: 1. Go to ProZ.com (highest likelihood of finding) 2. Enter language pair in search 3. Set filter “MT Post-Editing” or just “Editing” 4. Sort by rating (nothing below 4.7) 5. Review last 10-20 projects: - Quality of work? (standard, premium, budget?) - Client feedback? (look for “perfect,” “high quality”) - Recency? (6+ months ago = possibly inactive) 6. Write briefly to 5 people: “Hi, I have a technical project (5K words, EN-DE, light MTPE). Available for MTPE work? Rate and timeline?”
Time: 2-3 hours.
Days 4-5: Negotiation and test project¶
Task: Agree terms, deliver test.
When they respond (usually 2-6 hours for active people): 1. Ask: “What CAT tools do you use? Do you have MTPE certification?” 2. Offer test: “Before starting, I’d like a quick test: 500 words from our actual project, light MTPE, in your CAT tool. I’ll pay €15 as a good-faith check.” 3. Wait for result (usually 24-48 hours)
When results come back: - Review each correction (important!) - Ask yourself: did they correct errors or retranslate? - Is it reasonable? - Can you see CAT tool use?
If good, propose first project: “Great! Let’s start with 2K words. If quality holds, we continue. Rate: €0.10/word, light MTPE, 3-day deadline.”
Time: 3-4 hours active work.
Days 6-10: First real project (only 2-3K words)¶
Task: Validate quality in real conditions, catch early issues.
What to do: 1. Give text. Say: “I’ll give feedback on every segment where I find issues.” 2. When results return, review 100% (not spot-check, 100%). Flag errors. 3. For each error, write briefly: “Line 23: ‘interface’ should be ‘surface contact’, not ‘интерфейс’. Medical text, so…”. Give context. 4. After 1K words, email: “Good progress. Few corrections. See attached. Also note: in medical texts, always check terminology against glossary I attached.” 5. Their response to feedback tells you a lot: accept criticism? Ask clarification? Ignore?
Time: 5-8 hours your time on QA.
Red flags for continuing: - Quality < 75% → don’t continue with this person - Quality 75-85% → continue, but more feedback needed - Quality > 85% first submission → good person, can scale - If they ask clarifying questions → excellent sign, safe bet
Days 10-14: Finalize first project and decide¶
Task: Wrap up, make go/no-go decision.
When 2K words are done: 1. Do a spot 5% review of random segments (10-15 lines). Quality solid? 2. If quality 80%+ and responsive — offer: “Let’s expand to 5K/month retainer. Can you commit to 2-3 week availability?” 3. If quality 60-75% — offer more training or replace 4. If quality < 60% — stop now. Better to cut losses than suffer 6 months.
Time: 2-3 hours.
Common hiring mistakes and how to avoid them¶
Mistake #1: “Any good translator can post-edit” — Classic error. They retranslate by habit instead of correcting. Result: worse quality, higher cost, obviously over-edited text.
Solution: hire people with MTPE certification or validate them on a test. Spend an hour on testing now instead of a month redoing work.
Mistake #2: Not checking baseline AI quality — You hire a post-editor, then discover the AI model for that language pair is terrible. No fix can help.
Solution: before hiring, run 1,000 words through AI, assess quality. If 50-60%, post-editing maxes out at 70%. If 75-80%, post-editing hits 90%+. Know this upfront.
Mistake #3: Paying too little and expecting excellence — Pay $0.03/word, expect 60-70% quality. Pay fairly ($0.08-0.12 for mid-tier), get 85-90%.
This isn’t generosity, it’s economics. Good people demand fair rates because they have options.
Mistake #4: No clear requirements checklist — Post-editor doesn’t know what you want: check numbers? names? idioms? Without clarity, they guess, and output is uneven.
Solution: send a checklist day one: “Check: 1) Numbers and dates, 2) Company names (use this glossary), 3) Verb tense (present vs past), 4) Punctuation in complex sentences. Don’t touch until you understand context.”
Mistake #5: Hire someone “ready for post-editing,” but they don’t actually know a CAT tool — You get slow work because they copy-paste into Word instead of working in Trados or memoQ.
Solution: always ask at hire: “Which CAT tool do you use?” If “I don’t know” or “MS Word”—pass.
Week 3+: Scale¶
If quality is solid: - Gradually increase volume (toward 5-10K words/week) - Every 2 weeks, give constructive feedback even if good: “Excellent terminology work. One thing: double-check numbers before submitting.” - Monthly spot-check 5-10% of output. If quality dips, talk about it. - Keep a backup post-editor in reserve for coverage
Time: 2-3 hours/month on QA if all goes well.
Complete hiring timeline and scaling benchmarks¶
Here’s what unfolds week by week following this process:
| Phase | Days | Task | Your time | Output |
|---|---|---|---|---|
| Planning | 1 | Define scope, languages, budget | 0.5 h | Search criteria |
| Candidate search | 2-3 | Contact 5 ProZ prospects | 2-3 h | 2-3 responses |
| Negotiation | 4-5 | Test project, rate, deadline | 3-4 h | 1 selected candidate |
| First project | 6-10 | 2K words, 100% QA, feedback | 5-8 h | Quality assessment |
| Decision | 10-14 | Go/no-go verdict | 2-3 h | Scale or replace |
| Full onboarding | 14 days | Complete cycle to production | 12-20 h | Ready post-editor |
| Scaling | 15+ | Regular work, QA, feedback | 2-3 h/month | 5-10K words/week |
Realistic quality trajectory first month: - Days 1-3: tool setup, workflow learning. Unproductive yet. - Days 4-7: first projects, quality 60-70%, you give active feedback. - Week 2: quality climbs to 75-80%, speed 70%. - Week 3: quality 85-90%, speed 90%+. - Week 4+: quality 90%+ (if solid), full speed 100%.
This is normal. Expected. Don’t expect 95% quality day one.
Bottom line: post-editors aren’t faster translators. They’re a different profession with different methods and skills. If you’re using machine translation at scale, hiring a qualified post-editor saves you 30-50% time vs. traditional translation. But only if you hire the right person and train them clearly and patiently. Don’t rush—vet portfolios, run a test project, then scale.