$10,000 for a project across three languages sounds reasonable, then you realize your editor spent 200 hours instead of 80. Or the opposite — a client says “if MTPE is that fast, give me 80% off” and you close the project at a loss.
The problem isn’t MTPE itself. It’s that there’s no universal formula. MT quality is unpredictable, editors have different speeds, content ranges from “fix a typo” to “rewrite it completely”, and clients often don’t understand why MTPE doesn’t cost 90% less.
So today we’ll break down how to build a realistic MTPE rate card that protects your margin and is fair to editors.
Why MTPE pricing is so complex¶
Regular translation pricing is simple: the client sends 5,000 words, you need 25 hours at 200 words/hour, done.
MTPE is different because the work depends on a machine that can be brilliant or, honestly, useless. An Artlangs specialist reports that one 80,000-word project cost $0.05/word for light PE ($4,000), yet the same content with financial documents required $0.12/word for full PE ($9,600). However, laboratory research with poor MT output cost $0.14 — more than full translation.
This happens because MTPE has no “minimum effort” like human translation does. A human has rhythm, pace, experience — relatively predictable quantities. MT might be brilliant or translate “tractor” as “budget” and the editor rewrites three paragraphs.
Thus Bayan Tech confirms that 69.6% of agencies use per-word pricing, but it’s not universal — it’s just the least administratively burdensome.
Factors that affect your pricing¶
Before you set rates, determine five things:
1. Which MT system are you using? GPT-4, DeepL, Google, legacy system? If GPT-4 on your domain-tuned model, light PE will be fast. If budget-tier, editors will rewrite.
2. Language pair. EN→DE, FR→ES, most European pairs are well-trained. EN→ZH, EN→JA — editors spend twice as long. EN→LT, EN→HU, African languages — rare, so premium needed to offset low volume.
3. Content type. User guides, microcopy — easy to edit. Legal documents, medical records, financial reports — editors can’t just fix typos, they must worry about meaning shift.
4. Complexity of tables, tags, formatting. Plain .txt without markup — quick. .docx with 50 tables, formulas, fonts — editors spend extra time on formatting, not just editing.
5. Team size needed. One editor handles 3,000-5,600 words/day (full PE). A 500,000-word project needs a team and there’s competition for available editors.
Model 1: Per-word pricing (most common)¶
Most agencies bill like this: X words × $Y per word = invoice.
Simple and theoretically fair, but it hides reality: an editor on easy material hits 1,000 words/hour; on hard material drops to 400-500.
Standard per-word rates:
| Editing Level | Price / Word | When It Works Best |
|---|---|---|
| Light PE (minimal fixes) | $0.03-$0.08 | Internal docs, user guides, industrial manuals, content for comprehension (not sale) |
| Full PE (publication quality) | $0.08-$0.15 | Marketing materials, technical docs, website content, zero errors acceptable |
| Specialized (legal, medical) | $0.15-$0.25 | Contracts, medical instructions, financial reports, certified translation |
Real example. Client sends 50,000-word SaaS manual DE→EN, high-quality DeepL MT. Light PE takes 2-3 days (one editor, 1,000 words/hour). Your invoice: 50,000 × $0.05 = $2,500. Editor cost ($40/hour × ~45 hours) = $1,800. Margin: $700 (28%).
But if the content has technical errors, broken tables, or rare language pair — the editor might spend 60-70 hours. Margin drops to 5% or negative.
Model 2: Hourly rates (for short jobs)¶
If a project is under 5,000 words or MT quality is unpredictable, hourly is fairer.
Standard rates: $35-$85/hour for direct client work, though experienced editors charge $50-$90+.
When to choose it: - This is your first project with them; you don’t know MT quality - Specialized content needs deeper review - Very short jobs (<2,000 words) - Client wants iterative feedback, not a one-shot delivery
Real example. A startup needs a US legal letter translated EN→FR (4,000 words), complex terms. Per-word at $0.15 = $600, but the editor knows it’ll take 15-20 hours checking logic and comparing to original. Instead of per-word, offer “$60/hour, expect 12-18 hours” — the editor isn’t underpaid for complex work.
Model 3: Percentage-based (discount from your full rate)¶
Many agencies set MTPE as a % of their full translation rate. Example:
- Full human translation (baseline): $0.20/word
- Light PE: 50% of baseline = $0.10
- Full PE: 75% of baseline = $0.15
Weglot’s take: if your baseline is $0.20, a 30-70% discount is reasonable, so $0.06-$0.14 depending on MT quality.
Pros: you control your margin (you know you’ll get at least 30% of full translation income).
Cons: can be unfair if MT quality varies. Poor MT input + 70% discount = loss.
Model 4: Effort-based pricing (rare but fair)¶
Run a pilot on a small sample (first 1,000-2,000 words), measure editor time, scale to full volume.
Example: 1. Client: 50,000 words, FR→EL (rare), technical content. 2. Pilot: 2,000 words of raw MT. 3. Editor spent 8 hours = 250 words/hour (hard). 4. Calculation: 50,000 ÷ 250 words/hour × $60/hour = $12,000. 5. Client knows why; editor is protected.
More accurate than per-word, but administratively heavier (need pilot, time tracking).
Comparison table of all models¶
| Model | Per-word | Hourly | Percentage | Effort-based |
|---|---|---|---|---|
| Ease of setup | ⭐⭐⭐⭐⭐ Simplest | ⭐⭐⭐ Medium | ⭐⭐⭐⭐ Easy if you have baseline | ⭐⭐ Hard, needs pilot |
| Income predictability | ⭐⭐⭐ Depends on MT | ⭐⭐⭐⭐ Well-known | ⭐⭐⭐⭐ Controlled margin | ⭐⭐⭐⭐⭐ Most accurate |
| Fairness to editors | ⭐⭐ Can be fair or not | ⭐⭐⭐⭐⭐ Fair, pay for time | ⭐⭐⭐ Depends on MT | ⭐⭐⭐⭐⭐ Fair, depends on difficulty |
| When to pick | Standard work, good MT quality | Short jobs or unpredictable quality | When you need quick rates without pilots | Large, hard projects with unknown difficulty |
How to build your rate card: step-by-step¶
Step 1: Define your baseline full translation rate¶
Start with what you (or an editor) charge for regular translation in your language pair:
- English / European languages: $0.15-$0.22/word
- Rare language pairs: $0.25-$0.35/word
- Specialized / technical: $0.30-$0.50/word
Step 2: Assess your MT system quality¶
Run 1,000-2,000 words through your MT, review the output, log errors. Determine:
- How many sentences need complete rewriting? (full PE)
- How many need minor fixes? (light PE)
- What % requires specialized knowledge?
Step 3: Measure editor productivity on your material¶
Have 3-5 editors tackle the same task (3,000-5,000 words), record time each spent.
Typical result: - Light PE: 800-1,200 words/hour - Full PE: 500-700 words/hour
If your content is harder, expect lower numbers.
Step 4: Calculate cost per word¶
What’s the editor’s hourly cost to you (salary, benefits, commission)? Say $45/hour.
- Light PE: (1,000 words/hour) ÷ ($45/hour) = $0.045 cost per word
- Full PE: (600 words/hour) ÷ ($45/hour) = $0.075 cost per word
Step 5: Set your margin¶
Typical margin: 20-40%. Lower margin = fast sales, simple clients. Higher = risk if MT quality drops.
Recommendation: - Light PE: cost × 1.35 = $0.045 × 1.35 = $0.061 (round to $0.06) - Full PE: cost × 1.35 = $0.075 × 1.35 = $0.10
Step 6: Add volume discounts¶
Clients bringing 50,000+ words upfront deserve a break (10-20%). Makes sense logistically, but apply it to your margin, not the editor’s pay.
Example card:
| Volume | Light PE | Full PE |
|---|---|---|
| Up to 10,000 words | $0.06 | $0.10 |
| 10,000-50,000 words | $0.055 | $0.095 |
| 50,000+ words | $0.05 | $0.09 |
Step 7: Set specialized rates¶
Legal, medical, technical — raise by 15-25%:
- Light PE (Technical): $0.06 × 1.2 = $0.072
- Full PE (Legal): $0.10 × 1.25 = $0.125
Just list it separately on your price sheet.
Real case studies: how it works in practice¶
Case 1: SaaS company, English → Ukrainian¶
Setup: - Project: 120,000 words, manual + in-app text - MT: DeepL, high quality - Timeline: 4 weeks
Calculation: - Light PE: 120,000 words × $0.055 (volume discount) = $6,600 - Editor hours: 120,000 ÷ 1,000 words/hour = 120 hours - Editor cost: $45/hour × 120 = $5,400 - Your margin: $6,600 - $5,400 = $1,200 (18%)
Assessment: Fine — large volume, low risk (DeepL quality).
Case 2: Law firm, English → German¶
Setup: - Project: 8,000 words, contracts + NDA - MT: Google Translate (older model), poor quality on legal terms - Deadline: 3 days
Problem: Per-word at $0.12 × 8,000 = $960 looks okay, but the editor’s first time on legal content will spend 20 hours (not the 8-10 expected).
Solution: Go hourly. - Offer: $65/hour × 20 hours = $1,300 - Editor gets: $45/hour × 20 = $900 - Your margin: $400 (31%), editor is protected
Assessment: Better than per-word here.
Case 3: Accounting localization, English → Portuguese (Brazil)¶
Setup: - Project: 200,000 words, stable accounting docs - MT: In-house system, medium quality (rare pair) - Pair: Rare
Logic: Per-word forces you to price higher because Portuguese editors are scarce and MT quality is lower.
- Full PE (Portuguese, rare): $0.12 (baseline) × 1.3 (rare) = $0.156
- Volume discount (200K): 200,000 × $0.14 = $28,000
- Editor hours: 200,000 ÷ 600 words/hour (harder MT) = 334 hours
- Editor cost: $50/hour (Portugal, experienced) × 334 = $16,700
- Your margin: $28,000 - $16,700 = $11,300 (40%)
Assessment: High margin justified — much higher risk, scarce editors, unpredictable quality.
Common MTPE pricing mistakes¶
Mistake #1: Price below 50% of human translation
Some agencies think: MTPE is twice as fast, costs half. Wrong. If you charge $0.20/word for full translation, MTPE can’t be $0.10, because: - Editors still do 75% of the work on hard content - Infrastructure, QA, timeline management costs the same - Underpaid editors leave for competitors
Realistic minimum: 50-75% of full translation rate.
Mistake #2: One rate for all content
If you quote $0.07/word light PE for everything: - Easy content = 40%+ margin - Hard content (legal) = 5% margin or loss - Editors on hard content earn same as on easy
Better: 3-4 tiers by difficulty.
Mistake #3: Ignore MT quality variation
You set $0.07/word when MT was good, then a client brings you worse MT and the editor spends twice as long. Result: loss.
Protect yourself: always pilot new MT systems before quoting per-word.
Mistake #4: Forget about admin time
Per-word doesn’t cover: - File analysis, prep, export - QA on editor errors - Client communication - Encryption, confidentiality handling
That’s 15-25% more time. Build it into your margin.
Mistake #5: Price wars
If a competitor offers $0.03/word for full PE, they don’t have a margin. Don’t copy — when your quality drops, clients won’t come back.
Regional and language pair differences¶
Though MTPE is primarily a Western practice, markets vary:
Northern Europe (Germany, Sweden, Denmark): - Editor rates higher ($55-$70/hour) - Clients pay for quality - Per-word median: $0.08-$0.12
Southern Europe (Spain, Italy, Greece): - Editors cheaper ($35-$50/hour) - Fiercer competition - Per-word: $0.05-$0.08
Eastern Europe (Ukraine, Russia, Poland): - Editors low-cost ($25-$40/hour) - MTPE business potential high (low overhead) - Per-word: $0.03-$0.06
Rare language pairs (Portuguese, Greek, Hungarian): - Editors scarce, charge premium - Per-word: $0.10-$0.20
Bottom line: adjust for local labor market, but don’t go below cost + minimum margin.
Tools to help build your card¶
You don’t need complex software, but useful to have:
- Google Sheet with PE tiers, volume discounts, language pairs
- Toggl or Harvest to track editor productivity on your content
- Excel calculator for cost-per-word (editor rate ÷ productivity = cost)
Simplest: Google Sheet, update quarterly when you have new data.
FAQ on MTPE pricing¶
Q: How do I re-price if the client switches MT systems mid-project?
A: If you quoted per-word based on MT quality (e.g., DeepL) and they suddenly say “use my internal MT”, quality may drop. Options:
- Recharge by actual hours (switch to hourly for remainder)
- Blended adjustment (e.g., 10% complexity fee)
- Decline if risk is too high
Q: Should I list QA as a separate line item?
A: Yes. Many agencies show: - Post-editing cost (editor time) - QA cost (+10-15%)
If you do QA yourself, your margin shrinks. If separate QA person, list it so clients see where money goes.
Q: How do I adjust rates when MT gets better (GPT-4, new DeepL)?
A: 2024-2025 saw big MT jumps (GPT-4, Claude, DeepL 3.0). If you set rates 2 years ago, editor productivity on new systems may jump 20-30%. Options:
- Lower your rate 15-20% — fair to clients, editors still earn well
- Re-pilot on new MT to confirm
Q: How do I keep editors when competitors pay more?
A: It’s not just about money. Give editors: - Stable volume (40-60 hours/week) - Predictable content (no “black hole” projects) - Performance bonuses (quality maintained, volume hit) - Specialization option (legal, technical track)
Sharing your MTPE rates publicly is risky, but knowing how you arrived at them is power. When the next client asks “Why $0.08 and not $0.03?” you have an answer: editor cost, quality, language pair, MT system. That’s a professional conversation, not a bazaar haggle.