A client emails asking for a quote on 40,000 words with MTPE. “We already ran it through DeepL - just needs a light polish.” You quote 50% of your standard rate. They write back: “Still seems high for post-editing.” Your post-editor reviews the sample: “This output is rough. I basically need to retranslate half of it. I can’t do this at LPE rates.”
You’re stuck in the middle. The client undervalues the work. The translator correctly identifies the effort. The agency absorbs the gap.
This happens because MTPE pricing hasn’t developed the clear shared standards that hourly legal billing or per-word human translation have. Every agency figures it out on their own - sometimes badly. Here’s what the agencies that handle it well actually do.
The three pricing models - and when each makes sense¶
There are three ways to price MTPE work, and each fits different situations.
Per-source-word is the default. You charge by word count, applying a discount to your standard human rate. If your human rate is $0.20/word, your LPE rate might be $0.07-$0.10/word, and your FPE rate $0.12-$0.16/word. According to Nimdzi’s industry survey, 69.6% of translators prefer to be paid on a per-source-word basis. Clients understand it, it scales with volume, and it’s easy to compare quotes. The downside: a source word is a source word whether the MT gave you clean output or garbage. Your margin can swing wildly project to project.
Hourly works when MT quality is unpredictable. LPE editors handle roughly 5,000 words/day at $20-$35/hour (translator-side) or $40-$60/hour (client-facing). FPE editors work at about half that pace - 2,000-2,500 words/day. Hourly billing protects you when a “light edit” project turns into a near-complete rewrite. The problem: clients hate open-ended costs. If you quote hourly without a cap, most clients will push back. Always pair hourly rates with a time estimate and a ceiling.
Effort-based (TER/editing distance) is the most sophisticated model. You track how much each post-editor actually changes the MT output using Translation Edit Rate (TER) - the ratio of edits made to total word count. Segments with minimal changes get a reduced rate; heavily rewritten segments get full rate. In theory this is the fairest approach. In practice it’s complex to track, requires specialized tooling, and is often contentious because TER doesn’t capture cognitive effort - fixing a subtle mistranslation that looks fluent might require only 2 word changes but 10 minutes of research and judgment.
| Model | Best for | Main risk | Client appetite |
|---|---|---|---|
| Per-source-word | Known MT engine, consistent quality | Margin swings if quality varies | High - familiar, easy to compare |
| Hourly | Unknown MT quality, short projects | Clients resist open costs | Low - needs estimate + cap |
| Effort-based (TER) | Long-term agency-translator relationships | Complex tracking, hard to explain | Medium - requires trust |
Most agencies default to per-source-word and switch to hourly only for projects where MT quality is a genuine unknown before they start.
Light vs. full post-editing: what the difference actually costs¶
ISO 18587 - the international standard for machine translation post-editing - defines two levels of post-editing. The level you agree on with the client determines your rate, your timeline, and what liability you’re taking on.
Light post-editing (LPE) targets adequacy: the text must be accurate and understandable, but it doesn’t need to be stylistically polished or publication-ready. You fix factual errors, mistranslations, and anything that could cause misunderstanding. Awkward phrasing, flow issues, and stylistic inconsistencies are out of scope.
LPE makes sense for: internal documentation, product catalogs not intended for external audiences, content with a short lifecycle, situations where the reader knows the content is machine-generated and just needs the information.
Rate range: $0.03-$0.08 per source word. Throughput: ~5,000 words/editor/day.
Full post-editing (FPE) targets publishable quality - the output must be indistinguishable from a professionally translated text. Every sentence gets reviewed for accuracy, terminology, grammar, style, cultural appropriateness, and consistency with previous translations.
ISO 18587 primarily covers FPE. When a client requires “ISO-compliant MTPE,” they mean full post-editing. If that’s on your service menu, your QA process, documentation, and editor qualifications need to match what the standard requires.
FPE makes sense for: public-facing marketing content, legal and compliance documents, technical documentation that ships with products, anything the client will publish under their brand.
Rate range: $0.08-$0.15 per source word. Throughput: ~2,000 words/editor/day.
| Light post-editing | Full post-editing | |
|---|---|---|
| Quality goal | Accurate, understandable | Indistinguishable from human translation |
| Words/editor/day | ~5,000 | ~2,000 |
| Client rate range (2026) | $0.03-$0.08/word | $0.08-$0.15/word |
| Editor pay range | $0.02-$0.05/word | $0.05-$0.10/word |
| Hourly client equivalent | $20-$40/hr | $40-$75/hr |
| ISO 18587 scope | Mentioned, not primary focus | Yes, primary scope |
One counterintuitive point: the daily earnings for editors at both levels often converge. An LPE editor doing 5,000 words at $0.04/word earns $200/day. An FPE editor doing 2,000 words at $0.09/word earns $180/day. The difference in rate per word compensates for the difference in throughput. This means if you’re paying significantly less per word for FPE than for LPE, something in your model is wrong.
Building your MTPE quote: a practical walkthrough¶
Let’s use a real example: a technical manual, 40,000 source words, English to German, your standard human rate is $0.20/word.
Step 1: Assess MT quality. Before quoting, run 500 words through your engine and have an editor estimate effort. This takes 15 minutes and can save you a project that loses money. If the sample shows 85%+ fluency with mainly terminology issues, you’re in LPE territory. If the editor says “I need to rewrite entire paragraphs,” that’s FPE minimum.
Step 2: Choose the level. Technical manual for external distribution = FPE is the right choice. Even if the client says “light edit is fine,” publishable technical content with your agency’s name on it should be FPE or explicitly agreed as LPE with the client acknowledging the quality limitations.
Step 3: Calculate the client rate. FPE for EN-DE technical: $0.12/word × 40,000 = $4,800. Compare to full human translation: $0.20 × 40,000 = $8,000. The client saves $3,200 (40%) - which is a real benefit worth selling.
Step 4: Calculate editor pay. At 2,000 words/day, 40,000 words = 20 editor-days. At $0.08/word editor rate, that’s $3,200 for the editor. Your gross margin before PM time: $4,800 - $3,200 = $1,600 (33%).
Step 5: Account for overhead. PM coordination time on a 20-day project: 3-5 hours. QA review pass: 2-3 hours. Tools/platform costs: minimal. Total overhead: ~$400-600. Net margin: $1,000-1,200 (21-25%).
That’s a healthy project if your PM is efficient and your editor is reliable. It falls apart if the MT quality is worse than the sample suggested (editor takes 30 days instead of 20) or if the client requests revisions you didn’t budget for.
The lesson: always charge at least $0.10/word for FPE on professional language pairs. Below that, after paying the editor and covering overhead, your net margin is less than 15% - not worth the operational complexity.
What to pay post-editors - and why getting this wrong breaks your workflow¶
The most common agency mistake with MTPE is trying to pay post-editors at translation discount rates while charging clients at realistic rates. The math doesn’t work for the editor, and when good editors stop accepting your MTPE projects, your capability collapses.
Research published in The Journal of Specialised Translation that surveyed LSPs on their MTPE pricing practices found:
Many LSPs acknowledged setting rates upfront without assessing MT quality, creating situations where editors were underpaid relative to their actual effort - particularly in language pairs where MT engines perform inconsistently.
Practical benchmarks for post-editor pay in 2026:
| Content type | LPE (word rate) | FPE (word rate) | Hourly equivalent |
|---|---|---|---|
| Standard content (EN-DE, EN-FR, EN-ES) | $0.03-$0.05 | $0.07-$0.10 | $20-$40/hr |
| Technical/specialized | $0.04-$0.07 | $0.09-$0.12 | $30-$50/hr |
| Legal/financial | $0.05-$0.08 | $0.10-$0.14 | $40-$60/hr |
| Rare language pairs (+20-40% premium) | $0.05-$0.10 | $0.10-$0.15 | $35-$65/hr |
The margin you should target: pay editors 65-70% of your client-facing rate. If you charge $0.12/word and pay $0.04/word, the editor is doing 80% of the value creation while you earn 67% of the revenue. That relationship doesn’t hold long-term.
Three things that keep good post-editors on your roster:
Assess MT quality before sending projects. Send a 300-500 word sample and ask the editor to give you a 5-minute effort rating before you commit them to the whole project. It costs you 15 minutes of email. It costs them 10 minutes of reading. It prevents a situation where they’re doing FPE work at LPE rates and feeling cheated.
Pay on actual effort when quality varies. For your high-volume regular editors, consider a model where you share a portion of projects that go over estimated effort - either by adjusting the rate retrospectively or by giving them next priority on the next project. This signals that you’re not just off-loading risk onto them.
Don’t dress FPE work as LPE to save costs. When editors figure out that you’re calling something “light edit” because you want to pay LPE rates, but the work requires FPE effort, they mark you as a client they’ll decline. Word travels quickly in translator communities.
Rate modifiers: what adjusts the base rate up or down¶
Every project has variables that shift the quote from the base rate. These modifiers compound - a legal document in a rare language pair on a rush deadline can legitimately cost 2× the base FPE rate.
| Modifier | Rate adjustment | Why it matters |
|---|---|---|
| Language pair: major (EN-DE, EN-FR, EN-ES) | Base rate | Good MT engine performance, large editor pool |
| Language pair: mid-tier (EN-PL, EN-NL, EN-IT) | +10-20% | Some MT gaps, smaller editor pool |
| Language pair: rare (EN-UK, EN-HU, EN-FI) | +20-50% | Poor MT training data, few qualified editors |
| Content: standard business | Base rate | Standard terminology, moderate precision required |
| Content: technical/IT | +10-20% | Terminology consistency, specialized knowledge |
| Content: legal/financial | +20-35% | High-stakes errors, certification sometimes required |
| Content: marketing/creative | +10-20% for FPE | Cultural adaptation beyond literal accuracy |
| Volume: below 10,000 words | +10-15% | Setup cost doesn’t amortize, not worth discounting |
| Volume: 50,000-100,000 words | -10-15% | Stable editor workflow, lower PM overhead per word |
| Volume: 100,000+ words | -15-25% | Economy of scale, dedicated editor assignment |
| Turnaround: standard (5+ business days) | Base rate | Normal editor capacity |
| Turnaround: rush (48-72 hours) | +25-40% | Weekend rates, priority scheduling |
| Turnaround: critical (24 hours) | +40-60% | All-hands capacity, potential overnight |
| Client provides TM with high leverage | -5-15% | Reduces actual editing scope |
| Client requires ISO 18587 compliance | +5-10% | Documentation overhead, QA process |
According to Artlangs’ 2026 MTPE pricing guide, a large fintech project of 80,000 words across three languages at LPE averaged $0.05/word - but that masked a range from $0.04/word for English-Spanish to $0.07/word for English-Polish, reflecting the different MT quality and editor availability for each pair.
Language pair is often the dominant variable. EN-ES with a tuned engine might give you 88% BLEU scores on technical content - editors make minimal changes and LPE rates genuinely reflect the effort. EN-UK or EN-FI with the same engine might give 65% BLEU - editing effort doubles, and your LPE rate needs to account for it or you’ll be quoting work that loses money.
When MT quality tanks your margin - and how to protect yourself¶
This is the scenario that breaks MTPE pricing. Client sends files, you run them through your engine, the output is 60% correct. Now what?
If you’ve already quoted LPE rates and your editor spends FPE effort, you either eat the loss, go back to the client, or have a difficult conversation with the editor about why they’re getting paid less than the work warranted.
Three ways to protect yourself:
Pre-assess MT quality before every quote. Don’t assume “we use a good engine, it’ll be fine.” Domain mismatch, non-standard formatting, unusual terminology, or source text structure issues can crash output quality even for strong engines. A 300-word sample review before quoting takes 20 minutes and can prevent a money-losing project.
Build a quality-escalation clause. Include language in your client contracts: “Quote assumes MT output match rate above 75% for the content type. If average editing effort exceeds [X threshold] after processing 10% of content, the project will be renegotiated or redirected to full human translation.” Most enterprise clients will accept this if you frame it as protecting their quality, not protecting your margin.
Use the maximum-charge model. As explained by Hunnect, this works as follows: calculate the maximum project cost at full human translation rates - that’s your ceiling. Begin editing, track actual effort, invoice based on real work done, never exceeding the ceiling. The client knows their maximum exposure upfront. Your editors are compensated fairly. If MT quality turns out to be better than expected, the client pays less than the ceiling and feels they got a good deal.
As Weglot’s MTPE guide notes:
When machine output quality is poor, fixing nonsensical AI errors can be more mentally draining than writing from scratch - which pushes costs to 70-100% of full human rates and eliminates the cost advantage entirely.
If you can’t assess quality upfront and can’t include escalation clauses, the safest default is to quote at FPE rates and treat any efficiency gains as margin rather than passing them to the client through a lower quote. You can always come in under estimate; you can’t always go back for more.
What post-editors actually report about MTPE rates¶
MTPE adoption among top LSPs has reached 91% in 2025 - it’s no longer optional to have this capability. But the translator-side dynamics are increasingly important to understand.
According to a GTS Translation survey cited in Translators Journal, nearly 49% of translators report that MTPE has significantly lowered pricing expectations in the broader market - not just for MTPE work, but for translation generally. The translator perspective on this is direct:
MTPE is sold as ‘simple post-editing,’ but in reality it often demands the precision, creativity, and effort of a full translation. Machine output looks like a head start, but fixing it often takes just as long.
The deeper issue: MTPE work attracts the worst dynamic from a translator’s standpoint. You’re paying them less because “the machine did the translation,” but you still need their professional judgment, terminology expertise, and quality responsibility. When something gets through a post-edit with an error, it’s the post-editor who is accountable - at the same level of accountability as if they’d translated from scratch, but for a fraction of the fee.
Experienced translators who understand this are selective. They’ll accept your MTPE projects when rates are fair and MT quality is consistently decent. They’ll stop accepting them when rates feel exploitative relative to effort - and they often won’t tell you why, they’ll just start being unavailable.
The practical implication for agencies: don’t optimize MTPE pricing purely for client-facing competitiveness. A pricing model that wins the bid but can’t reliably source competent post-editors is not sustainable. The agencies running MTPE successfully at scale are those who’ve built editor relationships where MTPE work feels worth doing - not just contractually acceptable.
MTPE works as a service line when the economics work for everyone in the chain: client pays less than full translation, agency runs a lean workflow with healthy margins, and editors earn comparable daily rates to translation work. When one of those three breaks, the service line eventually breaks too.
FAQ¶
What are typical MTPE rates per word in 2026?¶
Light post-editing (LPE) runs $0.03-$0.08 per source word; full post-editing (FPE) runs $0.08-$0.15/word. Human translation for comparison is $0.09-$0.35/word depending on language pair and content type. Eastern European agencies often price 20-30% below these North American and Western European benchmarks while maintaining comparable quality for major language pairs.
Should I quote MTPE per word or per hour to clients?¶
Per source word is the industry standard and what clients expect when comparing quotes. Use hourly rates ($35-$85/hr client-facing) only when MT quality is genuinely unknown before starting, or for very short projects where per-word creates unnecessary complexity. Always pair hourly quotes with a time estimate and a cap.
How do I know if a project is LPE or FPE?¶
Ask these three questions: Will the output be published or shared externally? Does the client’s brand reputation depend on the quality? Are there legal, safety, or compliance implications? If any answer is yes, it’s FPE. If the content is internal, temporary, or gisting-only, LPE is acceptable - but confirm with the client in writing.
What margin should agencies target on MTPE projects?¶
Target 20-30% gross margin after paying the post-editor but before PM overhead. Below 15% gross, the operational complexity (briefing editors on MT-specific issues, handling quality surprises, turnaround management) makes the project not worth taking. Factor in 5-10% of project value for PM time on well-run projects.
What is the maximum-charge model for MTPE pricing?¶
The maximum-charge model sets a ceiling at full human translation rates and then invoices based on actual editing effort. The client pays for actual work done, never more than the full translation price would have been. This protects the client from unlimited cost uncertainty, gives editors fair compensation for actual effort, and protects your agency if quality is worse than expected. It’s more complex to administer than simple per-word, but it’s the most defensible model for projects where MT quality is variable.
How do volume discounts typically work for MTPE?¶
Volume discounts usually kick in at 50,000 words (10-15% off base rate) and again at 100,000+ words (15-25% off). The rationale: at high volumes, editors develop familiarity with the content, terminology, and style - reducing per-word effort and allowing the agency to pass some of that efficiency to the client. Never discount below your editor cost + overhead floor.
Джерела¶
- Nimdzi: Machine Translation Post-Editing - How Much Is the Fish?
- Weglot: Beyond Per-Word Translation - A Guide to MTPE Costs
- ISO 18587: Machine Translation Post-Editing Standard
- Hunnect: Maximum Charge - Pricing Strategies for MTPE
- Journal of Specialised Translation: In Search of a Fair MTPE Pricing Model
- Translators Journal: MTPE Rates - A Mirror of the Translator’s Identity Crisis
- Artlangs: How Much Does MTPE Cost - Transparent Pricing Guide 2026
- Polilingua: MTPE Adoption Surges 75% in the Post-Editing Revolution