The most expensive sentence in enterprise localization is "we will get the translation memory when the contract ends." Program owners say it with confidence, and then discover that a decade of approved translations, glossaries, and style decisions lives in a vendor's system, in a format nobody has exported, under a contract that never mentions it. Every language program that has been through a vendor change knows how much of its quality was stored in someone else's database.
That is the misconception worth correcting first: a localization program is not a stream of purchase orders, and its most valuable asset is not the vendor. It is the linguistic data you accumulate: translation memory, terminology, and style rules. Programs that treat those assets as the client's property, governed by the client, get cheaper, faster, and more consistent every year. Programs that leave them with the vendor start from zero every time they change suppliers.
Why Ad-Hoc Localization Gets More Expensive Every Year
Ad-hoc purchasing feels efficient because each request is small. The cost is hidden in the pattern: marketing orders a brochure from one vendor, the legal department sends contracts to another, HR finds a freelancer, and the training team uses whichever agency built the course. Four buyers produce four glossaries, four interpretations of the brand voice, and no shared memory. The same product name appears in three spellings in one language. Nobody can say what the organization spends on language, which means nobody can negotiate it.
The villain here is not any one vendor. It is fragmentation: no owner, no shared assets, and no measurement. A program fixes that with a handful of unglamorous decisions.
The Five Pieces of a Managed Program
1. Centralized intake
Every request, whatever its origin, enters through one front door: a form, a ticket queue, or a TMS connector. Intake captures language pairs, audience, content type, deadline, and risk level. That single record is what allows routing (a patient consent form goes to a different workflow than an internal newsletter), budgeting, and reporting. It does not require a large team. It requires that nobody bypasses it.
2. Translation memory the client owns
Translation memory (TM) stores approved source-and-target segments so that repeated and similar text is not translated, or paid for, from scratch. The practical rule: the client should own the TM, hold a current copy, and be able to export it. Put three things in the contract:
- Ownership language that names the TM, termbase, and style guide as client property, including work created by subcontractors.
- Delivery in open exchange formats, TMX for memories and XLIFF for in-progress files, so assets move between systems without conversion loss.
- A refresh schedule, such as quarterly exports, so you are never dependent on the vendor's goodwill at the end of the term.
Ownership also protects quality. When a vendor changes, the new team inherits the old team's approved decisions instead of relearning them at your expense.
3. Terminology and a style guide per language
A termbase records the approved translation of each key term: product names, regulated terms, and words you deliberately do not translate. A style guide covers register, formality, punctuation, units, and date formats for each locale. Both need a named owner inside your organization, because vendors can propose terms but only you can approve them. Start small: the 50 to 100 terms that appear in nearly every document will do more for consistency than an exhaustive list nobody maintains.
Running localization across departments? Taika builds managed programs around client-owned translation memory, terminology, and one intake process, delivered through GSA Schedule and NASPO ValuePoint vehicles that procurement teams already recognize.
4. In-country review built into the workflow
ISO 17100, the international standard for translation services, requires that a translation be checked by a second linguist, a reviser, who compares it against the source. That step is baseline quality control, not a premium add-on. Beyond it, many programs add an in-country review: a subject-matter reviewer in the target market, often a client employee, who confirms that the text sounds right to local readers and uses approved terminology. The key is to define the review's scope (terminology and tone, not rewriting preferences), give reviewers a deadline, and feed their corrections back into the TM so each fix is made once.
5. Metrics that show whether the program works
Choose a few measures and track them every quarter:
| Metric | What it tells you |
|---|---|
| Turnaround against agreed tiers | Whether the vendor and your intake process are keeping promises |
| TM reuse rate | How much of each project is repeated content, and therefore how much the memory is saving |
| Reviewer corrections per project | Whether terminology and style guidance are working; a falling number means they are |
| Rework and complaint count | Whether defects are reaching the audience, which is the number that matters most |
Where Machine Translation Fits
A program decides in advance which content may use machine translation with human post-editing and which may not. ISO 18587 defines the post-editing process, and the sensible split follows risk: high-volume, low-stakes internal content may suit it, while certified, legal, clinical, and safety content does not. Writing that rule into the program, rather than leaving it to each requester, is what keeps cost savings from turning into liability. Our article on when machine translation post-editing is appropriate goes through the decision in detail.
Consolidate Across Service Lines, Not Just Languages
Most organizations that think about vendor consolidation think about translation alone. The larger gain is consolidating everything that touches a multilingual audience: document translation, interpretation, captioning and transcription, and the accessibility work that has to be redone in every language. A translated PDF that is not tagged for screen readers is still inaccessible, and remediation does not carry over from the English original. When four vendors handle four of those steps, no one is accountable for the whole result, and your terminology fragments again at every handoff.
One managed provider with ISO 17100-conformant translation, interpretation, and accessibility under a single governance model is easier to measure and easier to hold to account. It is also easier to buy: Taika's GSA Schedule and NASPO ValuePoint contracts let public-sector program owners place orders on a vehicle their procurement office already recognizes. For a framework on judging any provider, see how to evaluate a translation vendor.
A Simple 3-Step Plan
- Inventory what you have. List every vendor, every language, and every place translation memory or glossaries currently live. Ask each vendor to export what they hold in TMX or equivalent, and put the exports in one location you control.
- Set the rules. Name a program owner, open one intake channel, adopt a glossary and style guide per language, and write the machine-translation rule and the review workflow into a one-page policy.
- Contract for the program, not the project. Move to a managed provider with asset ownership, delivery formats, turnaround tiers, and quarterly reporting written into the agreement, and review the four metrics every quarter.
What failure looks like is familiar: inconsistent product names across markets, a vendor change that resets quality to zero, a compliance notice published in a language nobody reviewed, and a budget no one can explain. Success looks quieter: reuse rates that climb, review corrections that fall, and a language budget that buys more every year because the memory belongs to you.
Why Work With Taika
Language Access Hub, powered by Taika Translations, is a veteran-owned (VOSB), SAM-registered provider on GSA Schedule and NASPO ValuePoint contracts, working with ATA-certified linguists. We run enterprise localization programs and managed translation services that start with the inventory above and end with assets that belong to you. If you have vendors, languages, and glossaries scattered across departments, a short scoping conversation is the fastest way to see what a managed program would change.