Guides · Choosing
The best localization platform is the one that fits the work: a decision guide
Documents, a website, app strings, or all three; a one-off delivery or a site that changes every week; your own translators or people you hire. The best localization platform for a team depends on those answers, so this guide asks them in order and shows, for each, what to check in a pilot and what Lionrapid does at that step.
How to choose a localization platform: start with the work, not the feature list
The best localization platform for one team is the wrong one for another, because the work differs. Seven questions sort it out.
- What is the content? Documents go in as files and come back in the same layout: 13 formats. A website is served translated through a plugin, a proxy, a script tag or an API. App strings sync from JSON and seven other formats through a CLI. A mixture is normal, and the memory and termbase are shared across all of it.
- How often does it change? A delivered document is finished; a site or an app is a loop of changed strings, reviewed and published, which is the continuous localization question.
- Who does the translating? Your own people, volunteers, translators you hire by offer, or the machine with a reviewer behind it. The platform fits each of those because the editor, the memory and the checks are the same for all of them.
- Check the complete path. Input, translation, review, output, publication: a platform that handles four of the five leaves you doing the fifth by hand.
- Check ownership. Memories and termbases should import and export in standard formats on every plan, and you should be able to name where the data lives.
- Count the second month. The plan, the machine translation, the review, the route and the updates: five numbers, not one.
- Run the pilot on the awkward file. The one with tables, footnotes and the terms your reviewers argue about.
Run the pilot on the awkward file, not the easy one: the tables, the footnotes and the terms people argue about are where platforms differ.
- document formats in and out in the same layout, checked by structural QA
- 13
- website routes included on every paid plan: plugin, proxy, script tag, API
- 4
- languages, with lemma-aware term matching in about 35 of them
- 345
- plans without translation memory and termbase: they are on every plan, including Free
- 0
A few documents, now and then
Upload, review, download the same file
If the work is a handful of documents a year, the best localization platform is the one you do not have to set up. Sign up, upload a DOCX, XLSX, PPTX or PDF, read the machine draft in the editor with the memory and the live checks, and download the same file with the translation in its layout. The free plan covers 3,000 words of documents for 90 days, which is enough to run the pilot on a real file before any plan is paid for.
- Free: 3,000 words, 90 days
- Same file back, layout checked
- 18 live checks on the draft
- No setup
A website that keeps changing
Serve the languages, then publish only the changes
A website is a publishing problem as much as a translation problem, and the best localization platform for one is the one that serves the languages itself. Check that the platform renders each language on the server with its own URLs and hreflang, that nothing goes live before a person approves it, and that a changed page costs the change rather than the site. Here the WordPress plugin and the edge proxy do the serving, a reviewer marks rows ready per language, and the [continuous localization guide](/guides/continuous-localization) walks through a page changing after launch.
- Server-rendered language URLs
- hreflang and x-default
- Ready rows only
- Changed strings re-translated
An evolving app
Strings in the pipeline, not in a spreadsheet
For software the questions are placeholders, plurals and the build. ICU plural and select branches are translated per branch, placeholders are guarded by a check, the CLI pushes only the rows that changed, validate with the CI flag fails a build on a missing translation, and a signed webhook fires per approved row so your deploy can react. The SDK renders from cache and fetches updates in the background.
- ICU plurals per branch
- validate --ci
- Signed webhooks
- Cache-first SDK
Several teams or vendors
Roles, offers and gates instead of email threads
When more than one person touches a job, the best localization platform is the one that carries the handoffs. An organization has owners, PMs and members; a project has stages, each with a QA gate that warns or blocks; translators are reached by offer, ranked by fit, with the price derived from memory leverage and shown in full before anyone accepts; delivery is a Finalize step that locks the result. The [translation management system](/product/translation-management-system) page shows each of those in the interface.
- Owners, PMs, members
- Stages with warn-or-block gates
- Assignment by offer
- Finalize locks the delivery
Existing language assets
Bring the memory in, use it, and export it again
Years of translations are an asset, and the best localization platform is the one they travel into and out of; the pilot should prove it. Import a TMX memory and a TBX termbase, run a file against them and watch the exact and fuzzy matches appear, then export both and diff them against what you uploaded. Here import and export are on every plan, a Trados or memoQ XLIFF can be attached as pre-translation, and the same memory prices the quote and fills the editor.
- TMX and TBX in and out
- SDLXLIFF / MQXLIFF as pre-translation
- Export never gated by plan
- One memory for quote and editor
The second month
What it costs once the first translation is done
The first month is the plan plus the initial machine translation; the second month is the plan plus the changes. Plans start at EUR 9.90 a month including VAT with 25,000 machine-translation words, packs add EUR 10 per 10,000 words, a source word is counted once however many languages you serve, and a translator you hire sets a per-word rate in the offer. Put those four numbers next to your content volume and the comparison between platforms becomes arithmetic.
- From €9.90 a month
- €10 per 10,000 MT words
- Source words counted once
- Review priced per word, in the offer
FAQ
Choosing a localization platform, answered
A CAT editor is where a person translates with memory and checks; a translation management system runs the people, stages and quotes around it; a website translator serves the languages. If your content is documents, the editor is the core; if people and deadlines are the problem, the management layer; if the site is the content, the serving route. Lionrapid is all three on one login, so the question becomes which parts you switch on, not which product you buy.
Documents need nothing: upload and translate. The WordPress plugin is a ZIP and an API key. The edge proxy is one DNS record on a subdomain you own. The script tag is one line in your page template. Only the CLI and the API assume someone who commits code.
Import a TMX 1.4 or CSV memory and a TBX or CSV termbase, on any plan, and attach them to a project. An XLIFF from Trados or memoQ can be attached to a document as its pre-translation. Everything exports again the same way, whenever you like.
Yes. Invite your translators as organization members, or hire from the marketplace by offer; both work in the same editor. Publishing stays where it is: WordPress keeps its theme and SEO plugin, a site behind the proxy is untouched at the origin, and an app keeps its build pipeline with the CLI inside it.
Upload the awkward file. Read the editor's live checks on the draft, look at the quality-estimation score on each machine-translated row, and download the file: the structural check has already compared the rebuilt document against the source. Then export the memory the pilot created and open it: the pilot, not the brochure, shows which is the best localization platform for your file.
The plan, plus machine translation for whatever changed, plus any review you order. The website translation cost guide works three sites through month one and month two with the real figures.
Export your memories as TMX, your termbases as TBX and your finished translations as XLIFF, per file or as a project archive, on every plan. The data is yours and the formats are the standard ones every other tool reads.
For a team that wants to audit the process, often: MFA that an organization can require, roles per organization and project, a published DPA and a 99.5 % SLA, servers Lionrapid operates in Finland and machine translation on models it hosts. The enterprise page lists each of those as the mechanism behind it, and a pilot on one real workflow is how to find out.
Run the pilot on your awkward file
Every signup starts free with 3,000 words of documents; every paid plan has a 14-day trial.