What are the translation costs per word? It’s one of the first questions people ask when they start budgeting for a translation project. But it’s not a great question to start with.
Sure, rates matter; nobody’s budget is infinite. But that rate is usually the only number on the quote. The project management hours, the developer time spent pulling strings out of your product and putting them back in, the rework after someone catches a terminology mismatch, none of these things show up in that per-word rate.
This post breaks down what translation costs actually look like in 2026, whether you're pricing out a website, an in-app UI, a documentation library, or a batch of marketing content. You'll get real per-word ranges by method, the hidden line items most quotes leave out, a budgeting formula you can run yourself, and how to reduce translation costs without cutting quality.
What determines translation costs?
Many only consider the number of words and potentially the number of hours that go into translation. In reality, there are five common variables that influence pricing.
- Language pair. Common pairs like English to Spanish or English to French sit at the low end of any rate table. There are many qualified translators, so supply keeps prices down. Pairs with fewer qualified linguists, like English to Norwegian, Japanese, or Finnish, can cost 50% to 100% more per word.
- Word count. More words usually mean a higher cost, but it’s not a linear equation. Many agencies apply minimum project fees of $75 to $150, which makes small jobs disproportionately expensive. Volume discounts typically kick in above 50,000 words.
- Subject-matter complexity. Marketing copy costs less than a medical device manual. That’s because legal, medical, and financial content require specialized translators, and that comes with a premium of 20% to 50% over general content.
- Turnaround time. Compressing a two-week job into three days will have its price. Rush fees usually add 25% to 50% to the final rate.
- Quality tier. This is where translation pricing swings hardest. Sometimes, you notice the same 10,000 words can cost $30 or $3,000. Translation quality is the decisive factor: machine translation with no human input will be the cheapest, but the quality might also be the lowest. MTPE will come with a price increase and better quality, while fully human translation, the type most use for sensitive or highly-specialized content, is the most expensive.
Current translation pricing data from sources like Smartling and Alconost suggests the following rate tiers depending on the translation method you choose. Here's how translation cost per word breaks down by method:
These rates have remained more or less stable for the past two years. What has changed is the quality floor: modern AI translation has made MTPE viable for content that would have required full human translation five years ago, a shift that can and has already had an impact on rates. For a closer look at how AI translation changes the cost equation, see our breakdown of human versus AI translation.
The real cost of translation services: beyond per-word rates
The per-word rate is the easiest number to compare, but it's only one part of the total cost of a translation program. The total cost of translation services includes everything it takes to get translated content live, keep it accurate, and keep it current. For most teams, those hidden line items can add up quickly.
Project management and coordination
Someone has to brief translators, answer their questions, chase deliverables, and route files between systems. Agencies often bill this as a separate line item.
When it’s your own team doing this work, it doesn’t appear on the invoice, but it’s still hours your marketing manager isn’t spending on marketing.
Developer time
This cost probably surprises teams the most. Traditional file-based translation means engineers extract strings from your site or product, package them for translators, then re-inject the translated files and QA the result.
For a mid-sized website, that’s usually estimated at around 20 to 60 developer hours per release cycle. At a loaded cost of $75 to $120 per hour, one translation cycle can easily consume between $2,000 and $7,000 in engineering time before you’ve paid for a single word.
QA and rework
Broken exports, character coding issues, text expansion that wrecks your layouts, and inconsistent terminology across pages. Fixing translation defects after launch costs three to five times more than catching them upstream would, and rework rarely gets budgeted at all.
Minimum fees and tooling
Small, frequent updates are where minimum project fees bite. If your site changes weekly and every change triggers a $100 minimum, you’re paying agency rates for ten-word edits.
Add TMS licensing if you’re running your own translation memory stack, typically $5,000 to $30,000 per year at the mid-market level.
The cost of waiting
A translation cycle that takes eight weeks instead of one is more than just slow. It’s two months of pipeline you’re not capturing in the market you’re trying to enter.
And many potential customers lost along the way. That’s a number that might not make into a quote, but CFOs usually notice it.
Sticker price versus total cost is the frame that makes every downstream decision easier. Keep it in view as you compare methods.
Translation method cost comparison: DIY, freelancers, agencies, and TMS software
Different translation approaches come with different trade-offs, whether you're pricing out a website, an app, or a folder of documents. The cheapest option upfront isn't always the most economical over time, and the cost of translation services looks different once you count what each model adds on top of the per-word rate.
The pattern to notice: as you move down the table, the sticker price stops falling, but the hidden cost load does.
Speed and scalability follow the same curve. In-house and freelance work move at the pace of whoever you can reach that week. Agencies commit to a turnaround in the contract. TMS platforms publish on a rolling basis once configured. Scalability tracks the same way: doubling your language count roughly doubles PM and developer hours under the first three models, and mostly amounts to a configuration change under the fourth.
There isn’t a single best choice for every organization.
If you're translating a legal contract once a year, hiring a specialist translator or agency may make perfect sense.
If you're updating a SaaS product every week, launching new landing pages every month, or maintaining multilingual documentation, manual coordination quickly becomes the more expensive part of the process. In those cases, reducing operational overhead often matters more than negotiating a slightly lower translation rate.
How much should you budget for translation?
Translation budgets often exceed expectations for one simple reason: teams underestimate how quickly content multiplies across languages. The formula below works the same whether you're pricing a website, a product UI, an app, or a stack of documents — only the word count and rate change.
Here’s a formula you can use in a spreadsheet before you talk to any vendor:
(Source word count × rate per word × number of target languages) + PM overhead (10–15%) + tooling + developer hours
Now let’s check out an example, because there is a multiplication trap inside the formula that leads to mistakes and can blow up most first-time budgets.
Say you have 10,000 source words and five target markets. Right here, you already have the first trap. You may think you’re translating 10,000 words. In reality, you’re translating 50,000. Every new language multiplies your entire word count.
At an MTPE translation cost per word of $0.10, that means:
- Translation: 50,000 words x $0.10 = $5,000
- PM overhead at 12%: $600
- Developer extraction and injection, 40 hours at $90: $3,600
- Tooling (entry-level TMP, annualized share): $1,500.
Total: roughly $10,700. More than double the $5,000 per-word math suggested at first, and that’s for one translation cycle. A website, a product UI, or a set of app strings isn’t static. It changes. So you’ll need to budget for ongoing cycles, not just the launch.
Two takeaways to consider here. First, languages multiply everything, so market prioritization goes beyond strategy; it’s a true cost decision. Second, the biggest non-translation line is developer time, which is exactly the line automation can remove.
If you’re building the full plan around this budget, the full playbook walks through market selection, content prioritization, and the operational setup step by step.
How to reduce translation costs without sacrificing quality
Lower translation costs don’t have to mean lower-quality translations. In many cases, organizations spend more than necessary because their workflow is inefficient and not because translators charge too much.
Here are some of the most effective ways to reduce costs while maintaining quality.
Use Translation Memory and stop translating the same sentence twice
Translation Memory stores every translated segment and reuses it automatically. On a typical website, 30% to 60% of content repeats or recurs across pages and updates.
These savings compound: your second year of translation costs will be meaningfully lower than your first year, and your fifth costs less than your second.
According to Smartling’s 2024 State of Translation Report, a 40% Translation Memory reuse rate is considered an industry benchmark, while the average Smartling customer exceeds 60%. In practice, that means more than half of the content submitted for translation already exists in the Translation Memory and can be reused rather than translated from scratch.
If a vendor doesn’t offer Translation Memory, or charges you a full rate for matched segments, that’s a point where you can and should negotiate.
Go hybrid
Use AI translation alongside human review. Very few types of content still need a level of control that requires 100% human translation. AI-first with human post-editing delivers near-human quality for most content at 40% to 60% lower costs.
Reserve full human translation and transcreation for content where nuance genuinely drives revenue: your homepage, your pricing page, and your ad campaigns, or sensitive content like legal and medical, where data privacy is critical.
Automate extraction and injection
This is the single biggest structural saving for web, product, and app translation. A platform that detects content changes automatically, sends new strings for translation, and publishes the results, removes the 20 to 60 developer hours per cycle entirely.
In the example we looked at above, that’s $3,600 or the $10,700 gone. Localize was built exactly around this problem: content detection, translation, and publishing run automatically, so your engineers stay on your product roadmap instead of your translation pipeline.
Consolidate vendors
Three agencies, two freelance pools, and a review vendor means six sets of minimums, six coordination threads, and no shared Translation Memory.
Consolidating onto one platform or one primary vendor cuts coordination overhead and lets your Translation Memory compound in one place.
Prioritize markets by revenue signal
Don’t translate everything into eight languages on day one. Launch in the two or three markets where your traffic, trial signups, or sales conversations already show demand.
Expand when the data says so. Every language you defer is your entire word count you’re not paying for yet.
Prefer usage-based pricing over per-seat licensing
Per-seat TMP pricing punishes you for involving more people in review and grows whether or not your content does.
Pricing tied to translated volume scales with actual usage, which makes next year’s budget a forecast instead of a guess.
Implementing the first three levers can cut total translation spending by around 50% without reducing quality.
Translation software vs. translation agencies: which costs less long-term?
The answer depends on the shape of your translation work: a one-off project or an ongoing program. Sticker price and the total cost point in opposite directions here. For a deeper look at how this decision plays out beyond translation costs alone, see the build vs. buy comparison.
When an agency costs less: single, self-contained projects
For a one-time job, an agency quote often wins on sticker price. One document, one language, one deadline: pay the per-word rate, get a quality deliverable, done. There’s no subscription to amortize and no platform to configure for a project that ends in three weeks.
The agency model is perfect for this type of work. It usually includes quality control in the rate; they also handle project management, and you’re not paying for something you won’t use again. If you have a single translation project with a clear start and end date, an agency is a good answer, and probably the cheapest option.
When software costs less: continuous, multi-market programs
Most companies need translation for more than a one-time project. Because websites update weekly, product releases happen monthly, marketing launches new campaigns with at least every new launch. At this point, the agency cost structure starts working against you:
- Every cycle re-triggers PM fees and minimums. A ten-word pricing update can carry a $100 minimum fee. Weekly updates mean paying that toll every single week.
- File handoffs keep consuming developer hours. The extraction-and-injection tax we saw earlier repeats every cycle, indefinitely.
- Your Translation Memory lives in the agency’s system. That makes switching costly and keeps the negotiating power on their side, not yours.
Software flips this structure. The subscription is the fixed cost, and everything else gets cheaper over time. Translation Memory compounds every cycle and automation holds the developer line at zero. Adding a language becomes a simple configuration change.
The three-year comparison: deleted costs beat discounted rates
If you want to build a successful localization workflow, you need to stop asking, “Which quote is lower this quarter” and ask instead, “Which model costs less over the three years we’ll actually be running this program?”
Software wins structurally because it eliminates cost categories instead of discounting them. An agency can lower its per-word rate, but it can’t remove the PM overhead, the file handoffs, or the developer hours that go into its delivery model.
Localize’s no-code, automated model was built around this idea: continuous content, multiple languages, and no engineering tax on any of it.
The number that matters isn’t on the quote
The per-word rate is the easiest number to find and the least useful one to optimize. What determines whether your budget holds is the total cost of keeping the program running: the coordination, the engineering hours, the rework, and the compounding effect of translation memory and automation working for you instead of against you.
Run the formula. Price the hidden lines. Then compare vendors on total cost, not sticker price. That's how to reduce translation costs for good, not just negotiate a better rate this quarter.
Ready to build the budget properly? Download the full playbook for the full framework, from market selection to launch. Or if you'd rather see what the automated model removes from your cost structure firsthand, book a demo.
