Per-word vs per-hour vs per-project: which pricing model fits your agency

You’re translating 5,000 words of technical documentation. One client pays you $0.12 per word - that’s $600. Another offers a flat rate of $500 “for the project.” A third is willing to pay $50/hour, and you know you’ll finish in 10 hours - also $500. The difference seems minimal. But if the document has 40% repetitions and you’re using TM - the per-word model is literally stealing from you. And if the text turns out harder than expected - the flat rate eats your entire profit.

Choosing a pricing model isn’t just about “how to invoice.” It’s a strategic decision that shapes your income, client relationships, and business scalability. According to Slator, the global language services market is valued at $31.7 billion in 2025 - and 91% of LSPs still use per-word as their primary model. But does that make per-word the right choice for everyone?

Let’s break down each model - with real numbers, formulas, and specific situations where each one works (and where it doesn’t).

Three pricing models: quick comparison¶

Before we dig deeper - here’s the big picture:

Now let’s break each one down.

Per-word: the industry standard that isn’t always fair¶

Per-word pricing is like a dollar-per-mile taxi meter. Simple, transparent, easy to calculate. The client knows that 3,000 words at $0.12 = $360 before work even begins. The translator knows exactly what they’ll earn.

That’s why per-word remains the dominant model. According to ATA, most US-based freelancers work this way. Agencies too - it’s easy to scale: get a 50,000-word order, split it among 5 translators, pay each per-word, invoice the client per-word with a margin.

Real per-word rates in 2026¶

Data based on Translated.com Pricing Trends 2026 and Smartling Rate Guide.

When per-word works well¶

• Standard texts with predictable complexity - manuals, product descriptions, standard contracts. You know your productivity (say, 2,000-2,500 words/day), and every project brings expected income
• Working with agencies - 90%+ of agencies work per-word, and if you’re a freelancer depending on agency flow - you’ll have to play by their rules
• Large volumes with no repetitions - if the text is unique from first to last word, per-word honestly reflects your work

When per-word works against you¶

Here’s where problems start.

Texts with high repetition rates. You’re translating a 20,000-word product catalog where 8,000 words are repetitions or fuzzy matches from TM. The client pays for 20,000 words, but you’re really only translating 12,000. Sounds like a win? Nope - most agencies apply TM discounts: 100% match = 0% pay, fuzzy 75-99% = 30-60% pay. The result: for 20,000 “external” words you get paid for ~14,000 “weighted” ones.

As ATA notes, translators must account for all costs - not just translation time, but also research, QA checks, formatting, and client communication. Per-word accounts for none of that.

Complex, research-intensive texts. A 2,000-word legal document can take as long as a 5,000-word general text - because of terminology, verification, and formatting. But per-word pays you 2.5x less.

Very short texts. Translating 200 words at $0.12 = $24. But you spent 15 minutes opening the file, setting up TM, formatting, and sending. That’s $96/hour - not bad, but if another 20 minutes went to client communication - it’s already $41/hour. For short texts, per-word doesn’t cover overhead.

The formula: converting per-word to hourly rate¶

This is the key metric every translator should know:

Effective hourly rate = (Word count x Per-word rate) / Actual time (hours)

Example: 3,000 words x $0.12 = $360. If you finished in 6 hours - $60/hr. In 8 hours (complex text) - $45/hr. In 4 hours (half from TM) - $90/hr.

I’d recommend tracking this metric for every project for at least a month. You’ll learn a lot about which text types are actually profitable, and which aren’t. More on rate calculation in our article on how to properly calculate your translation rate.

Per-hour: fair but opaque¶

Per-hour pricing is the opposite approach. You’re selling time, not words. However many hours you worked - that’s what you get paid.

This model is traditionally used by interpreters. But it works for written translation too - just in specific situations.

Real hourly rates in 2026¶

Data from Circle Translations and KL Translations.

When per-hour is the right choice¶

Transcreation and creative translation. You’re adapting a 5-word ad slogan. Per-word = $0.75. In reality you spent 2 hours finding the right adaptation, testing variants, coordinating with the client. Per-hour at $60 = $120 - much fairer.

As noted on Locate Translate, project-based or hourly pricing works for non-standard tasks where words aren’t a measure of work volume.

Editing and proofreading. You’re reviewing someone else’s translation. If it’s good - you move fast. If it’s bad - you’ll rewrite half of it. Per-word is unfair because the rate stays the same regardless of source quality. Per-hour automatically compensates for the difference.

DTP, layout, formatting. These tasks have no word count at all. How long it takes to fix a PDF - that’s what it costs.

Complex texts with unpredictable research time. Patent documentation, scientific papers, regulatory texts - documents where you might spend an hour finding the right equivalent for a single term.

MTPE with unpredictable MT output quality. Sometimes machine translation hits 85% - and editing takes 20 minutes per 1,000 words. Sometimes it’s at 40% - and those same 1,000 words become a 2-hour job.

Problems with per-hour¶

Clients don’t trust it. “How do I know you actually worked 8 hours and not 5?” - a typical question. Zero transparency, and that’s the main reason per-hour never became the standard.

The productivity paradox. The more experienced a translator is - the faster they work. Per-hour literally penalizes you for efficiency. A beginner spends 10 hours and earns $500. An expert finishes in 5 and earns $250. Yet the expert’s quality is higher.

Hard to estimate budget upfront. The client asks “how much will translating my document cost?” and hears “I don’t know, depends on how long it takes.” Not what they want to hear.

How to make per-hour acceptable for clients¶

1. Give an estimate with a range - “I estimate 8-10 hours, so $480-600. If I go over, I’ll check with you first”
2. Set a cap - “Maximum $600. If I finish early - you pay less”
3. Track time transparently - use Toggl or Clockify and send the client a report

Per-project: a fixed price for the outcome¶

Per-project (flat rate) is when you name a price for the entire project, and no matter how long it takes - the price stays the same. The client is buying an outcome, not words or hours.

This model is getting more popular - especially for complex projects where translation is just one piece of the puzzle.

When per-project is the best choice¶

Large projects with multiple stages. The client wants to translate a 50-page annual report from Ukrainian to English. That’s not just translation - it’s DTP, chart adaptation, layout, two rounds of review. Per-word covers only word translation. Per-project covers everything.

As recommended by Interprotrans, an agency can offer an all-in price - file conversion, style guide creation, two editorial rounds, and a buffer for last-minute changes.

Regular clients with predictable volume. A law firm sends 5-7 contracts every month. You know the average volume, complexity, terminology is already in TM. Flat rate $800/month for the whole package - the client is happy (predictable budget), and you’re happy (stable income).

Multilingual projects. Translating a website into 8 languages - the client needs one number for budgeting, not 8 separate per-word estimates with different rates.

Projects with unpredictable scope. Localizing a mobile app - few words (2,000), but there’s UI context, character limits, testing. Per-word = $240. Per-project = $800-1,200 accounting for all the work.

How to properly calculate a flat rate¶

Here’s the formula I recommend:

Flat rate = (Estimated words x Per-word rate) + Overhead buffer + Complexity premium

Where:
- Overhead buffer = 15-25% of the base amount
- Complexity premium = 0% (simple text) to 50% (legal/technical with research)

Example: 10,000 words x $0.12 = $1,200. Overhead 20% = $240. Medium complexity (+15%) = $180. Flat rate = $1,620.

Golden rule: if you consistently spend more time on projects than you estimate - your overhead buffer is too low. Track time and adjust.

Risks of per-project¶

Scope creep - the number one enemy. The client sends “just one more small file,” then “a few tweaks,” then “can we add this page too.” If the contract doesn’t fix the scope - you’re working for free.

How to fight it: - State clearly in the contract: “Price includes translation of up to 12,000 words. Additional volume billed at $0.12/word” - Revision cycles: “2 rounds of revisions included. Each additional round - $X” - Scope change = price change. No exceptions

More on contract structure in the article on how to draft a client contract.

Inaccurate estimate. If you quoted $1,000 but the project actually required $1,500 worth of work - the difference comes out of your pocket. The first few projects are always risky - until you build experience in estimation.

Hybrid models: combine the best of each¶

In reality, pure models almost never exist. Experienced translators and agencies combine approaches:

Per-word + minimum fee¶

Standard practice: there’s a per-word rate, but minimum order is $30-50. Translating 100 words at $0.12 = $12, but you’ll charge $35 minimum - and that’s fair, because overhead per order is the same.

Per-word + hourly for DTP/QA¶

The most common hybrid model in agencies: translation is billed per-word, while DTP, layout, and QA checks are billed per-hour. This way the client understands what they’re paying for.

Retainer (subscription)¶

The client pays a fixed amount monthly and gets a set volume of translations. According to research, 67% of enterprise clients prefer predictable pricing models.

Retainer structure: - Base fee: $500-2,000/month (PM, small tasks, priority service) - Included volume: e.g., 10,000 words/month - Overage: at a reduced per-word rate

Retainers are a win-win: the client gets stable service and a reduced rate, you get predictable income and can plan your workload. According to Common Sense Advisory, LSPs that transition clients from transactional to subscription models improve margins by 15-20%.

How to choose: a decision tree¶

Here’s a simple algorithm:

Step 1: Who’s your client? - Agency -> per-word (their rules, play along or find direct clients) - Direct client -> continue

Step 2: What type of content? - Standard text (manuals, descriptions, general documents) -> per-word - Creative / marketing -> per-hour or per-project - Complex specialized (legal, medical, patent) -> per-project or per-hour with cap - Multimedia / DTP / layout -> per-hour

Step 3: What volume and frequency? - One-time project < 5,000 words -> per-word with minimum fee - One-time project > 10,000 words -> per-project - Regular orders -> retainer

Step 4: Are there repetitions / TM? - High % of repetitions -> per-project (don’t discount 100% matches for free - you still spend time reviewing them) - Few repetitions -> per-word is fine

For a deeper dive into the calculations, check out the article on the real cost breakdown of translation.

How AI is changing pricing¶

We can’t talk about pricing in 2026 without addressing AI. Machine translation has radically changed the economics of the industry.

According to Slator, AI integration cuts overall translation costs by 40-60%. Here’s what that means:

Per-word rates are under pressure. Where the “normal” rate for EN-DE used to be $0.12-0.15, MTPE rates are $0.04-0.10. A 2-3x difference. Clients know this and push translators accordingly.

Per-hour is becoming fairer. MTPE editing is exactly the kind of task where per-hour makes more sense. You’re not translating from scratch - you’re editing MT output. The workload is unpredictable. Per-hour = honest pay for actual work.

As noted in a GTS study, 50% of freelancers do NOT offer discounts for MTPE, arguing that post-editing can take as much time as regular translation. Among those who do offer discounts, the most common range is 10-30%.

Per-project wins. The client doesn’t care who translated - a human or AI + human. They care about the result. Per-project lets you use any tools (AI, TM, glossaries) to boost efficiency - and the entire speed gain stays with you, instead of being passed to the client as a reduced per-word rate.

More on the hybrid AI + human workflow and its impact on translator income in a separate article.

Practical tips: 7 pricing rules¶

1. Track your real hourly rate. Even if you work per-word - know what you earn per hour. It’s the only honest metric
2. Minimum fee is mandatory. Don’t take orders under $30-50 - overhead will eat all your profit
3. Rush fee = +25-50%. Rush orders at the same price are a gift to the client at your expense
4. Count by source, not target. The source text is what you receive as input. The target can be 15-25% longer (e.g., English to German). If you count by target - factor this into your rate
5. TM matches ≠ free work. Even a 100% match needs to be verified in context. Minimum 10-20% of the full rate
6. Review your rates annually. Inflation, experience, specialization - all valid reasons to raise prices. How to do it without losing clients - in the article on raising your rates in 2026
7. Diversify your models. Don’t keep all clients on one model. Per-word for agencies, retainer for regular direct clients, per-project for large orders

FAQ¶

What’s the most popular pricing model among translators?¶

Per-word remains the standard - about 91% of language service providers use it as their primary model (Smartling, 2026). But the trend is shifting toward per-project and hybrid models, especially among agencies and freelancers with direct clients. The more a translator works directly with clients (rather than through agencies) - the more freedom they have in choosing a model.

How many words per hour does an average translator produce?¶

The standard benchmark is 250-350 words/hour for new content. With TM and MT - 500-1,000+ words/hour for MTPE. But this varies greatly by language pair, specialization, and complexity. Legal translation with terminology research might run at 150 words/hour, while simple product descriptions can hit 500+.

How do you calculate MTPE rates?¶

General rule - MTPE rate = 50-70% of the full translation rate. If your per-word rate is $0.12, MTPE would be $0.06-0.08. But this only works if the MT output is decent quality. For poor output, use per-hour - otherwise you risk working for $15/hour.

Should a translation agency publish prices on its website?¶

It’s debatable. Pros: filters out non-target clients, builds trust, simplifies initial communication. Cons: competitors see your prices, some clients need custom quotes, hard to account for all variations. Compromise: publish a range (“from $0.10 per word”) or a calculator, not fixed prices.

How do you transition from per-word to per-project with existing clients?¶

Gradually. Next time a client sends a large project - offer a flat rate with an explanation: “This includes translation, formatting, and two rounds of review. If we billed per-word plus formatting separately - it would cost more.” Show the benefit for the client.

What if an agency offers a per-word rate that’s too low?¶

Calculate your hourly rate at that per-word rate. If it comes out below $25-30/hour - decline. No amount of “steady flow” compensates for working below market rate. Better to invest time in finding direct clients who pay more.

How do you explain to a client why per-hour is better for their project?¶

Point out the per-word risks for the client: “With per-word billing, the translator has an incentive to work fast, not well. With hourly billing - the incentive is to do the best possible job. For a legal contract where an error could cost you thousands - that’s critical.” Always provide a cap or estimate to ease budget anxiety.

What tools help track time and calculate rates?¶

For time tracking - Toggl, Clockify, Harvest. For rate calculation - ProZ Rate Calculator. For invoicing - Invoice Ninja, FreshBooks. For automating your workflow overall - Zapier + a TMS like Protemos or XTRF.