Here’s a number most translation quotes won’t show you: the same 500,000 words can cost 25% less with machine translation and human review, or slightly more than a fully human workflow. The difference isn’t the engine. It’s what content you send to it, what assets sit behind it, and how many hours a post-editor spends cleaning up after it.
Machine translation is fast and nearly free. Human review isn’t. So when a vendor promises “30–60% savings,” the right question is: net of what? This guide answers that with the math: typical price ranges, a simple MTPE ROI formula, a worked example with a break-even point, and the hidden costs that quietly erase machine translation post-editing cost savings.
At Localizera, we run MTPE translation programs for documentation, support, and e-commerce teams, and the honest answer to “how much will we save?” is always “it depends, and here’s how to find out.”
TL;DR
- What it is: MTPE is a machine translation draft refined by a human post-editor, at a light (accurate, understandable) or full (publication-quality) level. You pay for editing effort, not the MT output.
- Typical savings: Vendor-reported MTPE pricing runs about 40–70% of human rates, and real-world throughput is closer to 2x than the 5–10x sometimes advertised.
- The formula: Net savings = human cost − (MTPE fees + setup + QA + tooling); ROI % = net savings ÷ total MTPE cost × 100.
- The catch: An unadapted engine, missing glossaries or translation memory, rework, and over-editing can push savings to zero.
- Is MTPE worth it? Usually yes for high-volume, repetitive, technical content; usually not for small batches, brand copy, or regulated documents without expert review.
What Is MTPE? Light vs. Full Post-Editing and What You Pay For
MTPE (machine translation post-editing) pairs an MT engine’s first draft with a professional linguist who corrects it. The ISO 18587 standard defines two service levels. Light post-editing fixes meaning errors and makes the text understandable without polishing style. Full post-editing brings the output to a standard comparable to human translation: accurate, fluent, and terminologically consistent.
The light vs. full post-editing decision shapes your price more than anything else, because what you’re really buying is human effort. The raw MT draft costs fractions of a cent per word; the post-editor’s time is the invoice.
How Much Does MTPE Really Save? MTPE vs. Human Translation Cost
MTPE Pricing and Cost Per Word
Vendor-published ranges give a starting point. ICS Translate reports light post-editing priced around 40–50% of the human per-word rate and full post-editing around 50–70%. Against a hypothetical U.S. benchmark of $0.15 per word for human translation, that puts MTPE cost per word at roughly $0.06–$0.11.
Treat these figures as vendor-reported, not guaranteed. Language pair, MT engine quality, domain adaptation, and source quality all move the number. For instance, Japanese MTPE often needs heavier editing than German MTPE, because English and Japanese differ far more in script, word order, and formality levels. Some providers also bill by the hour, and per-word rate vs. hourly pricing shifts risk: hourly protects the linguist when MT output is poor, while per-word protects your budget when it’s good.
MTPE Turnaround Time
Speed is a separate benefit from cost. ICS Translate cites post-editing productivity of roughly 3,000–5,000 words per day, versus about 2,000 words per day for translation from scratch. Ulatus cautions that in practice, light MTPE tends to deliver closer to 2x throughput, well short of the 5–10x some vendors claim. Even so, a 2x gain can halve MTPE turnaround time on a large product launch, which may be worth more than the per-word discount.
Content type | Typical workflow | Savings potential |
|---|---|---|
Help center, FAQs, internal docs, product listings | MT + light post-editing | Higher |
Technical manuals, software UI (with TM/glossary) | MT + full post-editing | Moderate to high |
Marketing, brand, and conversion copy | Human translation or transcreation | Low (savings often negated by rework) |
Regulated content (medical, legal, financial) | Human translation or full MTPE + SME review | Low; risk comes first |
What the Research Says: Insights From Nimdzi and Translation Studies
Independent analysis paints a more nuanced picture than vendor brochures do.
- Adoption is mainstream; efficiency isn’t. According to Laszlo K. Varga’s analysis for Nimdzi Insights, The MTPE Efficiency Gap (May 2025), average MTPE adoption among language service providers rose from 26% in 2022 to nearly 46% in 2024, and the share of LSPs using MTPE on at least half their projects jumped from 7.8% to 45.2%. Yet Nimdzi argues that much of the industry still operates well below optimal efficiency.
- Over-editing is a cost leak. Nimdzi finds that many teams over-edit MT output, polishing segments that already meet quality standards. This slows delivery and reduces MTPE cost savings.
- Automation multiplies the savings. Nimdzi reports that combining MTPE with quality estimation and automated post-editing can cut human involvement and costs by 50% or more. However, final validation increases linguists’ cognitive load, while legal, IT, and data-privacy constraints limit adoption.
- MT solves problems and creates new ones. Academic research adds a human-factors lens. As Dominika Cholewska explains in her review, An Ultimate Guide to MTPE (2021), Jean Nitzke’s multi-method study comparing post-editing with translation from scratch starts from the premise that MT can resolve some translation problems while introducing others, so post-editors need different problem-solving strategies.
How to Calculate MTPE ROI (Formula + Worked Example)
Use three simple formulas:
- Net savings = (words × human rate) − (words × MTPE rate + setup + QA/review + tooling)
- ROI % = net savings ÷ total MTPE cost × 100
- Break-even volume = fixed costs (setup + tooling) ÷ net savings per word
- Worked example (hypothetical, illustrative numbers only, not client results). Assumptions: 500,000 words per year; human rate $0.15/word; full MTPE rate $0.08/word; one-time setup (engine adaptation, glossary, TM cleanup, integration) $8,000; QA/review $0.01/word ($5,000); tooling $3,000/year.
- Human-only cost: 500,000 × $0.15 = $75,000
- MTPE cost: $40,000 + $8,000 + $5,000 + $3,000 = $56,000
- Net savings: $19,000 (25% lower spend)
- ROI: $19,000 ÷ $56,000 = about 34%
- Break-even volume: $11,000 ÷ ($0.15 − $0.08 − $0.01) = about 183,000 words
Now stress-test it:
Scenario (hypothetical) | Total MTPE cost | Net savings | ROI |
|---|---|---|---|
Base case | $56,000 | $19,000 | ~34% |
Rework penalty: 15% of words need an extra $0.10/word | $63,500 | $11,500 | ~18% |
Poor MT quality: 40% of words retranslated at an extra $0.10/word | $76,000 | −$1,000 | ~−1% |
Sensitivity check: if the engine isn’t adapted to your domain and post-editors end up retranslating 40% of segments, the program loses money before you even count the delays.
Hidden Costs That Eat Your Machine Translation Post-Editing Cost Savings
- Wrong content routed to MT: Skipping content triage sends brand and creative copy into a workflow built for manuals.
- Unadapted engine: Generic engines lose quality fast in specialized domains without domain adaptation.
- No glossary or translation memory: Without TM and glossaries, editors fix the same terminology errors over and over.
- Poor source text: Ambiguous, unedited source content multiplies MT errors downstream.
- Rework and downstream defects: Errors found after publication cost far more than errors caught in review.
- QA and review time: Quality estimation and QA steps are real line items, not freebies.
- Setup, integration, and security: Connectors, data-privacy reviews, and vendor onboarding add upfront costs.
- Post-editor pay pressure: Quota-driven, low-rate models push linguists to rush, lowering quality and raising rework.
Is MTPE Worth It? When It Pays Off and When It Doesn’t
MTPE is worth it when you have:
- High volume, comfortably above your break-even point
- Repetitive or technical content such as help centers, manuals, or product catalogs
- A strong translation memory and glossary
- Timelines that leave room for QA
It usually isn’t worth it for:
- Small volumes, where setup costs never pay back
- Creative or brand-led copy that needs transcreation
- Highly regulated content such as medical, legal, or insurance documents, where risk outweighs savings (see Machine Translation for Medical & Insurance Documents)
How to Maximize MTPE ROI and Translation Cost Reduction
- Triage content by risk. Route low-risk, high-volume content to MT and keep high-stakes content with professional human translation.
- Build glossaries and TM first. They improve every downstream step.
- Adapt the engine to your domain and terminology.
- Measure quality with a scorecard such as MQM, so “good enough” is defined, not debated.
- Track post-editing effort (edits per segment, words per hour) and right-size the editing level to avoid over-editing.
- Layer in Language AI workflows such as quality estimation once the basics are stable (explore our Generative AI Language Services page).
Conclusion
MTPE can deliver real savings, but only when the content, the engine, and the process fit together. Vendor ranges of 40–70% of human rates are a useful starting point, not a promise. Your actual MTPE ROI depends on volume, MT quality, the strength of your translation memory and glossaries, and how much editing you truly need. As the Nimdzi data shows, the biggest gains often come from fixing process waste like over-editing, not from switching engines. Run the formula, stress-test it with a rework penalty, and validate it with a short pilot before scaling. If you’d like help modeling your numbers or running that pilot, Localizera’s Machine Translation Post-Editing (MTPE) Services team can build a workflow around your content, not the other way around.
Frequently Asked Questions
How much cheaper is MTPE than human translation?
Vendor-reported ranges put MTPE at roughly 40–70% of the human rate, so the MTPE vs. human translation cost gap is typically 30–60% on fees alone. Net savings are lower once setup, QA, and rework are included.
What is the difference between light vs. full post-editing?
Light post-editing makes MT output accurate and understandable; full post-editing makes it publication-ready and comparable to human translation. ISO 18587 defines both levels.
Is MTPE worth it for small volumes?
Often not. Setup costs such as engine adaptation and glossary creation need enough volume to pay back. Calculate your break-even volume first.
How is MTPE priced: per word or per hour?
Most MTPE pricing is per word, usually as a percentage of the human rate. Hourly pricing is common when MT quality is unpredictable.
Does MTPE speed up turnaround time?
Yes, typically. Post-editors often handle about twice the daily output of translation from scratch, which shortens MTPE turnaround time on large projects.
Is MTPE quality as good as human translation?
Full post-editing can match human quality for suitable content. For creative, brand, or regulated content, human translation or expert review remains the safer choice.
