Translation vs Localization When You Buy Content

The cheapest way into a new market is to translate the content you already have. It is also the fastest way to sound slightly wrong, slightly foreign, or slightly ridiculous in a language you cannot read and cannot check. Translation moves the words across. Localization moves the meaning across. Buyers who treat them as the same thing ship content that is technically accurate and commercially dead.

If you are buying content for more than one market, the real decision is not which translator quotes the lowest rate. It is how much of the original you are willing to let change so the piece lands the same way somewhere else. Answer that first, and the vendor, the price, and the process all follow from it. Get it backward, start from the cheapest quote, and you end up paying twice, once for the translation and again for the reader who bounced because it read like a machine.

Translation, Localization, Transcreation

These three words get used interchangeably, and they describe three different products at three different prices.

Translation is the literal job. Take the meaning of the source sentence and render it accurately in the target language. A good translation is faithful and readable, and for a technical document or a legal notice that is often exactly what you want, nothing added, nothing adapted.

Localization goes further. It adapts the content so it feels native to the market, not just decoded into the language. That means converting currencies and units, swapping examples that only make sense in the source country, adjusting tone to match how the market expects to be spoken to, fixing date formats, and catching idioms that translate word-for-word into nonsense. The words change more, because the goal is a reader who never senses the piece started somewhere else.

Transcreation is the top tier and the least understood. It rewrites the piece for the new market while keeping the intent, and it is closest to commissioning fresh content. You use it for the pieces where the exact wording carries the weight, taglines, campaign copy, a hero headline, anything where a faithful translation would be accurate and completely flat. Transcreation is priced like writing because it mostly is writing.

Most buyers need a mix. Documentation and help content lean toward translation, marketing pages toward localization, and the handful of pieces that define the brand toward transcreation. Deciding which bucket a piece belongs in is the single most useful thing you can do before you ask for a quote.

Why Machine Translation Alone Keeps Failing You

Machine translation has genuinely gotten good. Tools like DeepL and Google Translate produce output that is grammatically clean and, for getting the gist of something, entirely fine. The trap is assuming clean grammar means market-ready copy.

Raw machine output misses the things that make content persuasive rather than merely correct. It flattens tone, so your carefully casual brand voice comes out neutral and corporate. It takes idioms literally. It has no idea which of your product terms are proper nouns that should never be translated and which are ordinary words. And it cannot tell that a joke, a cultural reference, or a regional example landed in the source and will die in the target. These are the same reasons machine-generated text struggles to hold a brand voice in its original language, amplified by a second language on top.

The workable middle ground is machine translation plus a human editor, a process the industry calls post-editing. The machine does the first pass, a native speaker fixes tone, idiom, terminology, and the cultural misses, and you get most of the speed at a fraction of full human-translation cost. It is a legitimate tier, and for a lot of content it is the right one. What is not legitimate is publishing the raw machine pass to customers and hoping, which is what “we’ll just run it through DeepL” quietly means.

Keyword Research Does Not Translate

Here is the mistake that costs the most and shows up the least, which is assuming your keywords translate. They do not.

People in a new market search for their problem in their own words, and those words are frequently not the literal translation of the term you rank for at home. A direct translation of your best-performing English keyword can land on a phrase almost nobody actually types, while the phrase they do type sits untargeted in a piece that was otherwise well translated. You end up with content that reads beautifully and ranks for nothing.

Real localized content starts from fresh keyword research in the target language and market, not from translating the keywords you already have. Sometimes the intent behind a query is different in another country, sometimes a single English term splits into two local ones, sometimes English itself is what the market searches in even though the market speaks another language. You only learn which by researching the market directly, the same way you would research search intent at home, in the language of the people you are trying to reach.

This is the strongest argument for commissioning a native writer over translating an existing piece. A translator renders your English intent. A native writer targeting local search can start from what the market actually looks for, which is a better foundation than any translation of the wrong target could be.

The Cost Ladder

Every tier of adaptation costs more than the one below it, and knowing the ladder keeps you from overpaying for documentation or underpaying for a campaign.

Tier	What you get	Rough relative cost
Raw machine translation	Fast gist, no human touch. Internal use only.	Lowest, effectively per-token cheap
Machine plus human post-edit	Machine first pass, native editor fixes tone and terms	Low to moderate
Human translation	A professional translates faithfully from scratch	Moderate, priced per word
Localization	Human translation plus cultural and market adaptation	Higher
Transcreation	Rewritten for the market, priced like original writing	Highest, close to a fresh commission

The point of the ladder is to match the tier to the stakes, not to always buy the top or always buy the bottom. Spending transcreation money on a help-center article is waste. Running your homepage hero through raw machine translation is a different kind of waste, the expensive kind that costs you conversions instead of budget. Sort your content by how much each piece has to persuade, and let that decide the tier.

Who Actually Does This Well

The vendor landscape splits into a few recognizable shapes, and the right one depends on volume and stakes.

Localization platforms like Smartling, Lokalise, and Phrase are built for volume and consistency across many languages. They pair translation memory, glossaries, and workflow with human linguists, and they earn their keep when you are maintaining a lot of content across several markets and need it to stay consistent as the source changes. For a company localizing a whole product and its content, this is usually the shape that fits.

Translation marketplaces and services like Gengo sit lower on the effort curve, good for straightforward translation at volume without the platform overhead. Bilingual freelancers are the other end, a single native writer or translator who knows your market, which is often the best fit for a small number of high-stakes pieces where a relationship and deep context beat scale.

The trade-off mirrors the one in agency versus freelancer versus marketplace for original content. Platforms buy you consistency and scale, individuals buy you depth and voice, and the right answer depends on whether your problem is many languages at volume or a few pieces that have to be perfect.

Building the Kit That Makes It Cheap

The single best way to lower localization cost is not a cheaper rate. It is giving every translator the same reference material, so they stop guessing and stop diverging.

A glossary is the foundation. List your product names, feature names, and key terms, and specify for each one whether it gets translated, stays in English, or has a fixed approved translation in each language. Without it, five translators produce five different words for your core feature, and you pay to reconcile them later. A do-not-translate list, brand names and product names that should never change, prevents the most embarrassing errors before they happen.

On top of that sits a per-language style guide, covering tone, formality, how you address the reader, and market-specific conventions. Formality alone is a real decision in many languages, and getting it wrong makes you sound either stiff or presumptuous to every native reader. The kit costs a little to build once and pays back on every piece after it, which is the opposite of chasing a lower per-word rate that saves you a few percent and buys you inconsistency forever.

How to QA Work You Cannot Read

Reviewing content in a language you do not speak feels impossible, and doing it badly is worse than not doing it. There is a workable process.

The core rule is that the reviewer is not the translator. A second native speaker reads the delivered work fresh, because the person who wrote it is the last person likely to catch its blind spots, exactly as you would never let a writer be their own fact-checker. That second pair of eyes is the actual quality gate.

• Have a native speaker who is not the translator review anything customer-facing before it goes live.
• Review in context, on the page or in the layout, not as a bare spreadsheet of strings. Meaning shifts once text sits in its real place.
• Check the non-word elements. Currency, dates, units, phone formats, and images with embedded text are where machine and human passes both slip.
• Use back-translation sparingly, only on the highest-stakes lines, since it catches gross errors but not tone.
• Confirm the local keywords actually made it in, so the piece is targeting how the market searches, not a translation of your home terms.

None of this requires you to learn the language. It requires you to budget for a second native reader and to refuse to publish work no one on your side can vouch for.

What to Do With This

The honest summary is that translation is a commodity and localization is a craft, and most content buyers pay commodity prices while expecting craft results. The fix is to sort your content by stakes, buy the right tier for each piece, build the glossary and style guide once, and never ship customer-facing work in a language nobody on your side has reviewed.

For the pieces that carry your brand into a new market, the cheapest-looking path, raw machine translation of an English original, is usually the most expensive once you count the readers it loses. The stronger move on those pieces is often to commission a native writer to write for that market, targeting how it actually searches, rather than translating a piece that was built for a different one.

That is the same standard Postdex holds in any language. Every article in the catalog is finished and readable in full before you buy, an edition of one that leaves the shelf permanently once sold, so your judgment happens at the point of reading rather than after a translation you could not check. When you need a piece written for a specific market to a spec you define, the commissioning desk takes the brief and a native writer produces work built for that audience, with a named editor making sure what reaches you reads like it was written there, not decoded into it.