You've used two or three translation vendors over the past decade. Maybe more. Somewhere in your SharePoint site, a shared drive, or a project manager's email archive, there are folders of old translations: Word documents, PDFs, InDesign files, and spreadsheets in six languages. Now you're evaluating a new translation provider, or asking your current one to take a harder look at what you've accumulated. And someone on your team asks the obvious question: are those legacy translation files worth anything?
The answer depends almost entirely on what your provider does with those files before putting them to work. Some legacy translations are clean, consistent, and ready to build on. Others are outdated, inconsistent across vendors, or full of terminology that no longer matches your products or policies. The difference isn't always obvious from the outside, which is why the first thing a good provider does with your legacy translations is ask questions, not upload files. Below are the scenarios a competent provider should be prepared to handle, and what you should expect from each one.
Your Legacy Translations Are Either an Asset or a Liability, and a Good Provider Will Tell You Which
Legacy translations become a liability when incorrect or outdated content enters a translation memory unchecked. A translation memory (TM) is a database that stores matched pairs of source text and translated text. When a similar sentence appears in a future project, the system automatically suggests the stored translation. That translation memory reuse is what drives cost savings and consistency over time.
But the same mechanism that makes reuse powerful also makes bad content dangerous. Translation memory systems are known to propagate errors: when an incorrect translation enters the TM, the system reuses it whenever similar source content appears in future work. One bad entry doesn’t just sit there. It multiplies across every project that matches against it.
Why Legacy TMs Degrade Over Time
Microsoft's globalization documentation identifies two primary causes of TM quality decline:
- Product updates that make existing translations inaccurate
- Multi-vendor use that introduces inconsistent terminology for the same concepts
Our project management team sees this regularly with new clients, where product names, safety language, and regulatory terminology change with every product cycle. A client who used one agency for packaging, another for user manuals, and a third for marketing materials often arrives with three different Spanish translations of the same product name. That kind of inconsistency doesn't resolve itself. Someone has to identify the conflicts, choose the correct term, and clean the TM before new work begins. A passive provider uploads whatever you send and starts matching against it. A provider with a real process asks about file age, content type, quality history, and how many vendors touched the content before deciding what belongs in your TM and what doesn't.
The Best-Case Scenario: You Have a Translation Memory File Ready to Transfer
When you switch translation vendors, the fastest path to continuity is to request your translation memory as a .tmx file. TMX is the standard format for transferring TM data between systems. If your current or previous vendor maintained a translation memory on your behalf, you can ask them to export and deliver that file directly.
Key point: You should own your translation memory. You paid for the translations, and the TM is built from your content. But not every vendor sees it that way, and some contracts don't address TM ownership at all. Get TM ownership and file access written into your contract before signing. If a provider refuses to release your TMX file on request, treat that refusal as a red flag.
Even with a clean TMX file in hand, a good provider doesn’t just import it and start translating. Industry analyses and vendor benchmarks consistently show that mature translation memory programs can significantly reduce translation costs over time, often in the range of 20% to 40%, though this range depends on consistent terminology management and regular TM maintenance throughout the program’s lifecycle. A TM file that hasn’t been reviewed in five years, or that reflects products and policies that have since changed, needs evaluation before it earns that kind of leverage.
To put the upside in perspective: one Argo Translation manufacturing client built more than 300,000 human-verified TM segments per language across a 30-year relationship. That corpus became valuable enough to train a custom AI translation engine that cut editorial intervention by more than 80%. That kind of compounding value is what a well-maintained TM produces over time, and it's what you're protecting when you insist on an audit before a new provider starts building on your existing file.
When You Don't Have a TMX: How to Reuse Existing Translations Through Alignment
A lot of translation buyers don't have a TMX file. They have finished deliverables: translated PDFs from a previous vendor, Word documents returned by email, or InDesign files sitting in a creative team's archive. No one ever mentioned a TMX, and no one thought to ask.
Those finished files aren't useless. A process called translation alignment takes your original source documents and the corresponding translated versions, matches the content at the sentence level, and stores the matched pairs as reusable TM entries. The result is a working translation memory built from work you've already paid for.
Which File Formats Work Best for Alignment
Alignment works best with editable file formats. Word, Excel, PowerPoint, and InDesign files all align cleanly. PDFs can sometimes be aligned, but formatting-heavy or scanned PDFs may need extra preparation or manual cleanup before the content is usable.
One practical detail that often surprises buyers: alignment quality depends on how closely the translated document mirrors the source document's structure. If a previous vendor heavily reformatted during desktop publishing, the alignment process may yield fewer clean matches. A good provider will tell you upfront what percentage of the content is likely recoverable, so you can decide whether alignment is worth the investment for a given set of files.
Old Translations Can Still Guide Quality Even When They Can't Feed a TM
If you have an approved terminology list, even a messy spreadsheet of preferred translations, hand it over during onboarding. Even when legacy files can't be aligned into a translation memory, those files hold value as reference material for establishing the terminology and brand voice your new provider needs on day one. Skipping that glossary handoff is where many vendor transitions lose quality.
Translation Memory vs. Glossary: What's the Difference?
Here's the distinction that matters: translation memory operates at the sentence level, reusing full translated segments automatically when it finds a match. A glossary or termbase operates at the term level, ensuring that specific words and phrases stay consistent across every project, regardless of sentence context. Most termbases are exported as simple spreadsheets, typically in .xlsx or .csv format, with columns for source term, target term, and definition. Many buyers mix up TM and glossary, but they serve different purposes, and a provider should be equipped to receive and use both.
Your old translations, even imperfect ones, show a new provider what terminology your audience expects, what tone your brand has historically used, and what past decisions your team made about industry- or product-specific language. A good provider uses reference files to build or validate a glossary before the first project ships, so your established preferences carry forward rather than starting from a blank slate.
Sometimes Starting Fresh Is the Right Call, and a Good Provider Will Say So
Starting over with new translations is the right decision when legacy content is too outdated, inconsistent, or error-prone to clean up without spending more than a fresh start would cost. When your company has changed product names, updated regulatory language, significantly evolved its brand voice, or had multiple vendors translate content without a shared glossary, the honest recommendation is to start from a clean baseline.
In Argo Translation's experience across 30 years of managing translation programs, content tied to product names, compliance language, and regulatory terminology tends to drift fastest, often within 18 to 24 months of a product update or policy change. General corporate communications are more durable, but even those can accumulate inconsistencies over time when multiple vendors are involved.
A provider who tells you "these files aren't worth building on" is being honest, not upselling. A provider who silently ingests everything and bills you for cleanup later is not being straight with you.
That said, a complete restart is rarely necessary across all content types at once. A provider may recommend starting fresh with compliance documents when regulatory language changes, while still aligning with and reusing more stable materials, such as general corporate communications. The decision should be made file by file, not all-or-nothing.
What to Ask When Switching Translation Vendors About Your Legacy Files
The way a translation provider responds to questions about your existing files during evaluation reveals more about their process than any sales presentation. These four questions expose the difference between a provider with a defined intake process and one that's improvising.
- "Do we own our translation memory, and can we get a copy of our TMX file at any time?" The answer should be an unqualified yes. Get TM ownership and access rights in writing before signing.
- "What will you do with the translated files we already have?" You're listening for a specific process: evaluate, align where feasible, build a glossary from reference files, and flag what's outdated. "Just send us whatever you have" is not a process.
- "How do you handle terminology that may be outdated or inconsistent across our old files?" This question tests whether the provider audits legacy content before adding it to a TM. A good answer describes a review step. A vague answer suggests they upload first and discover problems later.
- "If our legacy files aren't worth building on, will you tell us?" The answer to this question tells you whether you're talking to a long-term partner or a transactional vendor.
Argo Translation builds these questions into every new client onboarding, backed by a quality management system registered to ISO 9001:2015 and ISO 17100:2015 for translation services. The evaluation happens before the first project begins, not after problems surface.
Frequently Asked Questions
Who owns my translation memory?
You should. You paid for the translations, and the TM is built from your content. But not every vendor sees it that way, and some contracts are silent on the question. Get TM ownership and TMX file access written into your contract before signing. Any provider who refuses to release a TM file on request should raise serious questions about how that vendor manages client assets.
What if I don't have a TMX file, just old translated documents?
A translation provider can often build a translation memory from existing translated documents through a process called translation alignment. Alignment pairs original source documents with their finished translations and stores matched content as reusable translation memory entries. Word, Excel, PowerPoint, and InDesign files align most cleanly. PDFs can sometimes be aligned, but formatting-heavy or scanned documents may require extra preparation before the content is usable.
How do I know if my old translations are still accurate?
Legacy translations are most likely to be inaccurate when product names, regulatory language, or brand terminology have changed since the translations were completed. Content tied to compliance requirements, product specifications, or industry regulations tends to drift fastest. A competent translation provider will audit legacy content for terminology drift and flag outdated segments before adding them to a translation memory, rather than assuming all existing content is still current.
Do I have to start over when I switch translation providers?
Switching translation providers does not usually require starting from scratch. If you can obtain your TMX file from your current provider, a new provider can import the translation memory and begin leveraging your existing approved translations immediately. Even without a TMX file, translation alignment can often recover reusable content from your translated documents. You only need to start completely over when legacy content is too outdated, inconsistent, or error-prone to build on reliably.
If you're evaluating translation providers and want to understand what your existing files are worth, Argo Translation's team can review your legacy assets and tell you exactly what's reusable, what needs updating, and where to start.