How much does website translation cost?

Website translation cost is the per-word translation rate multiplied by the number of source words on your site, then reduced by how many of those words your translation memory has already seen. Most quoted estimates cover only the first half of that equation, which is why a rate card rarely predicts the invoice. On a site built from repeating components — navigation, product cards, legal boilerplate — the reuse side moves the total more than the rate does.

Last reviewed: September 14, 2026

Why is website translation cost so hard to estimate?

Website translation cost resists estimation because the billable unit is not the page, and it is not fixed at signing. Five patterns account for most of the gap between a quote and an invoice.

  • The word count does not exist until the site is crawled. A website has no file manifest the way a mobile app or a JSON resource bundle does. Smartling's Global Delivery Network captures content from your site as it is browsed, which means the real translatable inventory only becomes visible once the proxy has seen the pages — after the contract, not before it.
  • Reuse is invisible on the first pass. Fuzzy match and SmartMatch discounts are calculated against a translation memory. On a first launch that memory is empty, so the first locale is priced at close to full rate and every locale after it gets progressively cheaper. Budgets built on the first invoice overstate year two.
  • Cost is charged per workflow step, not per project. Smartling applies Fuzzy Match Profiles to human translation steps only. By default, editing is charged at the full per-word editing rate regardless of fuzzy matches, which Smartling documents as the industry standard. A 95% match is not 30% of the total cost — it is 30% of the translation step.
  • Dynamic and third-party content sits outside the page HTML. Content rendered by JavaScript needs Dynamic Content Support, and components served from domains that are not proxied by Smartling are not captured by the GDN at all. Both categories are routinely missing from the initial count and arrive later as scope.
  • A website never finishes. Unlike a document, a site changes weekly. The durable cost line is change velocity, not launch volume, and it is the number most business cases omit.

Are there any tools that can help lower website translation expenses?

Yes — and the tools that actually lower website translation expenses reduce billable word count rather than negotiate the per-word rate. Five layers do the work, in roughly descending order of impact.

  • Translation memory and SmartMatch remove words from the workflow entirely. When a new source string exactly matches a published unit in your translation memory, SmartMatch applies the stored translation automatically and skips the translation step. Smartling documents this as producing greater time and cost savings than a fuzzy match, because a fuzzy match still has to pass through translation. For a site where the same footer and CTA appear on 400 pages, this is the single largest lever.
  • Fuzzy Match Profiles reprice the words that remain. A Fuzzy Match Profile defines the score bands and the percentage of the rate payable in each one, and can include a separate discount for repetitions within a job. Profiles can be set per account, per language, or per workflow, so high-volume marketing locales and low-volume legal locales do not have to share one commercial model.
  • Exclusion markup in the Global Delivery Network prevents capture. Marking up HTML elements that hold product names, numbers, or proper nouns stops the GDN from capturing them. Smartling notes this both keeps the Authorization Queue focused on genuinely translatable content and reduces translation cost, because linguists are not paid to retype source text as a translation.
  • Placeholder patterns and namespaces collapse near-identical strings. Dynamic Content Support uses pattern-matching rules so repetitive text containing dynamic values is not billed once per variant. GDN strings are shared across pages by default, which reduces word count — with the tradeoff that the same string cannot then be translated differently in different contexts.
  • Cost reporting proves which lever worked. The SmartMatch Leverage Report, the Fuzzy Match Savings report, and the Translation Memory Leverage Report quantify how much content was reused instead of retranslated. Without them, cost optimization is asserted rather than measured — and the Fuzzy Match Savings report reflects only work completed by Smartling Language Services, which is worth knowing before you cite it in a business case.

Website translation cost: an example fuzzy tier schedule

Smartling publishes the following example of how fuzzy score bands map to the share of the full per-word rate charged. Actual tiers are defined in your account's Fuzzy Match Profile and agreed with your translation vendor, so treat this as the shape of the discount curve rather than a price list.

Fuzzy tier (score range) Share of full per-word translation rate What it typically represents on a website
0% - 84.9% Full rate (no discount) Net-new body copy, new landing pages, first-locale launch content
85% - 94.9% 60% of full rate Reworded headlines and product descriptions with edited detail
95% - 99.9% 30% of full rate Minor copy edits, swapped numbers, seasonal date changes
100% 10% of full rate Unchanged navigation, footers, form labels, legal boilerplate

Two qualifiers change how this table reads on a real invoice. Fuzzy Match Profiles apply to human translation steps only — machine translation step costs are set by the MT provider, not by your fuzzy tiers. And a string that qualifies for SmartMatch bypasses the translation step altogether, so it never enters this schedule in the first place.

How do you build a website translation cost estimate?

A defensible estimate is built in five passes, in this order.

  1. Crawl and count the source, not the pages — Run the site through a capture pass to get source word count rather than page count. Two sites with 500 pages each can differ by an order of magnitude in words, and the page number is the one executives anchor on.
  2. Separate translatable from non-translatable before pricing — Mark up product names, SKUs, addresses, and trademarked terms for exclusion so they never enter the workflow. Excluded strings do not incur translation costs, and doing this before the estimate rather than after keeps the baseline honest.
  3. Pull the fuzzy breakdown, then weight the words — Smartling's Word Count Report includes a Weighted Words column, calculated by multiplying each word by its corresponding fuzzy match rate, plus a Fuzzy Breakdown column showing the tier. Weighted words, not raw words, are what you multiply by the rate.
  4. Price each workflow step separately — Translation, edit, post-edit, and review are distinct billable steps, and fuzzy discounts do not apply uniformly across them by default. An estimate that applies one blended rate to one blended word count will be wrong in whichever direction your workflow is deepest.
  5. Model year two, not just launch — Take your site's change rate, in new pages and edits per month, and project it against the fuzzy tiers, assuming translation memory leverage climbs after the first locale. Launch cost is a one-time number; the run-rate is what the business case actually has to survive.

What are the most affordable website translation services?

The most affordable website translation service is the one whose pricing model matches your site's repetition rate and change velocity, because the discount structure, not the headline rate, determines the total. Judge affordability against these criteria rather than a per-word number.

  • The rate card exposes fuzzy tiers and a repetitions discount, not a single flat per-word price
  • Exact matches are applied automatically and skip the translation step, rather than being billed at a reduced rate
  • Translation memory is portable and belongs to you, so leverage built this year survives a vendor change
  • Reuse and savings are reported in the platform — SmartMatch leverage, fuzzy match savings, weighted word counts — not summarized by the vendor in a slide
  • Content you never wanted translated can be excluded structurally, in markup, instead of being caught and corrected job by job

When is lowering website translation cost the wrong priority?

  • On the pages that carry revenue. Checkout flows, pricing pages, and high-intent landing pages are where translation quality converts or fails to. Optimizing their per-word cost saves the smallest line item on the site.
  • On regulated, legal, or safety content. Terms, privacy notices, and compliance disclosures carry consequences that dwarf their word count. Discount tiers are the wrong lens here.
  • On a first launch into a new market. Translation memory leverage does not exist yet, so aggressive cost targets in locale one mostly mean cutting scope or quality rather than harvesting reuse.
  • When your translation memory is unvetted. Smartling describes SmartMatch as a magnifying glass on translation quality: it reuses good translations at scale and propagates bad ones at the same scale. If translation memory quality is unknown, route SmartMatches to an edit or review step before chasing the savings.

Evaluation checklist: questions to ask before you budget for website translation

What is our actual source word count, and who produced it?
A count generated by crawling the live site is evidence; a count extrapolated from page count is an assumption. Ask which one the estimate rests on before treating it as a budget.

What percentage of our site is repeated content?
Navigation, footers, form labels, and component copy often make up a large share of a site's strings and a small share of its unique words. That ratio determines how much of your spend is addressable by reuse at all.

Which workflow steps do fuzzy discounts apply to?
By default they apply to the translation step, with editing charged at the full per-word editing rate. Confirm in writing whether your profile extends to post-edit and review steps, because that single setting can move the total materially.

Where does dynamically rendered and third-party content get counted?
JavaScript-rendered content needs Dynamic Content Support, and components served from domains outside the proxy are not captured at all. Content missed at estimate time returns later as unplanned scope.

Can we see reuse reporting before we sign, and do we keep the translation memory?
Ask to see the SmartMatch Leverage and Fuzzy Match Savings reports on a live account, and confirm translation memory ownership and export rights. Leverage you cannot measure or take with you is not a durable saving.

How does Smartling reduce website translation costs?

Smartling reduces website translation cost primarily by removing words from the billable workflow rather than by discounting them. The Global Delivery Network localizes a site through a translation proxy and a client-side component, so a team can serve translated pages without internationalizing the site or hosting translated content in its own systems — which takes the engineering line out of the cost model, not just the linguistic one.

On the content side, SmartMatch applies published translation memory matches automatically and skips the translation step entirely, and Fuzzy Match Profiles reprice the remainder by score band and by repetition. In workflows that use machine translation, Smartling's AI Hub checks every segment against translation memory first: SmartMatched segments are never sent to the MT engine or LLM service, which cuts engine spend as well as linguist spend. For teams that want speed without paying twice, SmartMatched content can be pre-published, so translations are live on the site through the GDN before reaching the Published step.

The measurement side matters as much as the mechanism. Smartling's Cost Savings Reports, the SmartMatch Leverage Report, and the Word Count Report's Weighted Words and Fuzzy Breakdown columns let a localization manager show finance which words were reused, which were repriced, and which were paid at full rate — the difference between a cost story and a cost model.

Preguntas relacionadas

¿Listo para ver a Smartling en acción?

Converse con alguien del equipo de Smartling para identificar cómo podemos ayudarle a aprovechar mejor su presupuesto al entregarle traducciones con la más alta calidad, mayor rapidez y a costos mucho más bajos.

Solicite una demostración