Every translation you’ve commissioned in the last decade is an asset. Most businesses don’t know they own it, let alone how to use it. The translation memory sitting in your supplier’s systems could be saving you money, improving consistency, and speeding up delivery – if you knew how to access it.
Translation memory isn’t a new technology. It’s been standard practice in professional translation for over twenty years. Yet many businesses that have spent six figures on translation over the years have never thought about what happens to that investment beyond the immediate deliverable.
What translation memory actually does
At its simplest, translation memory is a database. Every time a translator works on your content, the original text and its translation are stored as paired segments – usually sentences, sometimes smaller units. When similar text appears in future work, the system finds the match and offers the previous translation.
The benefits are obvious. If your product documentation contains the same safety warning across fifty documents, that warning only needs translating once. Every subsequent occurrence comes from memory – faster, cheaper, and guaranteed consistent.
The matches can be exact (identical text) or fuzzy (similar but not identical). Exact matches typically cost nothing or near-nothing to reuse. Fuzzy matches need translator review and adjustment, but still save time compared to translating from scratch. New content with no matches goes through full translation.
This isn’t machine translation. A human translator created every segment in the memory. The system simply retrieves and suggests their previous work, maintaining the quality and consistency of human translation while reducing repetitive effort.
How it reduces costs over time
The commercial impact of translation memory accumulates. Year one, you’re building the memory – most content is new, costs are standard. Year two, repetitive content starts matching. Year three and beyond, mature memories can reduce costs substantially.
The exact savings depend on your content. Technical documentation with standardised sections and repeated terminology benefits most. Marketing content with unique creative language benefits less. Legal documents with boilerplate clauses benefit enormously.
A manufacturer with consistent product documentation might see 40-60% of content matching from memory in mature projects. That’s not a 40-60% cost reduction – fuzzy matches still require work – but it’s significant. More importantly, the consistency benefit often exceeds the cost benefit.
Consistency benefits beyond cost
Translation memory doesn’t just save money. It enforces consistency across time and across documents.
Without memory, every translation project starts fresh. Different translators might choose different terms for the same concept. Documents translated years apart might handle the same content differently. The result is a body of translated material that reads like it was produced by multiple different companies.
With memory, previous decisions carry forward. If your German translations use “Sicherheitsventil” for safety valve, that choice persists across every future document. Product names, technical terms, and standard phrases stay consistent automatically.
This matters for brand perception. It matters for technical accuracy. And it matters for compliance, where consistent terminology across documentation sets can be audited and demonstrated.
The ownership question
Here’s where many businesses stumble: who owns the translation memory?
If you’ve been commissioning translation for years without thinking about this, the answer is probably your translation supplier. They’ve built the memory in their systems, maintained it, and used it to benefit your projects. But you funded its creation through the translations you paid for.
This becomes important in three scenarios.
First, if you change suppliers. A new translation partner without access to your historical memory starts from zero. All that accumulated consistency and cost benefit evaporates. You pay to rebuild knowledge that already exists.
Second, if you want to use the memory differently. Perhaps you’re bringing translation in-house, or splitting work between multiple suppliers, or implementing your own translation technology. Without ownership, you can’t.
Third, if you’re acquired or merge with another company. Translation memories are business assets. They have value. But only if you own them.
The good news is that translation memories are portable. Standard exchange formats (TMX files) mean memories can move between systems. The question is contractual, not technical: do you have the right to your data?
Getting value from existing memories
If you’ve been working with translation suppliers for years, you almost certainly have translation memory assets – even if you’ve never asked for them.
Start by asking your current supplier what memory they hold for your account. How many segments? Which language pairs? How far back does it go? A good supplier will be happy to discuss this and should be able to provide statistics on memory leverage – what percentage of recent work matched from memory.
Ask about export. Can you get a copy of the memory? In what format? Is it contractually yours, or do you need to negotiate? Some suppliers include memory ownership in standard terms. Others retain ownership unless explicitly agreed otherwise.
Once you have the memory, you need to decide what to do with it. At minimum, ensure it travels with you if you change suppliers. Better, use it actively in quality processes – checking new translations against established terminology, auditing consistency, identifying drift.
Building memory strategically
For businesses starting fresh or wanting to maximise memory value, the building process matters.
Prioritise consistency over volume. A smaller memory with well-curated, high-quality translations is more valuable than a large memory full of inconsistent or questionable content. Bad translations in memory propagate forward, creating ongoing quality problems.
Invest in terminology upfront. If you define key terms before translation begins, those terms enter the memory correctly from the start. Retrofitting terminology after years of inconsistent usage is possible but painful.
Consider which content builds memory best. Technical documentation with repeating elements builds useful memory faster than creative content. Some businesses strategically prioritise technical translation partly to build memory assets that benefit later work.
When memory isn’t enough
Translation memory has limits. It captures what was translated, not why translation choices were made. It doesn’t replace human judgment for new content. And it can’t handle significant changes in terminology strategy without careful management.
If you decide to rebrand, change product names, or update technical terminology, memory can become a problem rather than a solution. Old translations keep surfacing, requiring manual intervention to override. Memory maintenance becomes necessary – reviewing and updating stored translations to reflect current standards.
Memory also doesn’t substitute for expertise. A memory full of translations from unqualified translators captures their errors alongside their good work. Quality in, quality out – memory amplifies whatever standards created it.
The conversation to have
Most businesses have never had a strategic conversation about translation memory. If that includes you, it’s worth starting one.
Ask your supplier to explain your memory status. Ask what it would take to export and own your historical memories. Ask how memory leverage is tracked and reported. Ask what happens to memory if the relationship ends.
At Bubbles, we’re transparent about translation memory. Our clients own the memories we build for them. We report on leverage and can export memories in standard formats at any time. Because translation memory is your asset – the accumulated value of every translation investment you’ve made. It should work for you, not just for your supplier.
