Ask your finance team one question. How much does the company spend on translation and interpreting, across every department, and who are all the vendors?
Most companies cannot answer it. Not because the number is secret, but because it lives in fifteen places at once. Legal expenses a contract of translation last quarter. Marketing agencies handle campaign localization. HR used a freelancer somebody found online. Support pays a per-minute interpreting line. Product ships strings through whatever connects to its localization tool. Each decision made sense on its own. Stacked together, they turned language into one of the least governed line items in the business.
That is the moment the centralized-versus-department-based question stops being theoretical.
The global language services market hit roughly $72.6 billion in 2025, and it is famously fragmented. The hundred largest providers account for under 20% of it. The rest is a long tail of small agencies and freelancers. Your internal sourcing usually mirrors that same sprawl. So, the question is not whether you have language work happening. It is whether anyone is steering it.
This piece breaks down the two models, what each costs, and how to decide which one fits a company your size. I will be direct about where centralization wins and where it does not.
Where this perspective comes from
At Day Translations, we see the fragmented model up close, every week. Companies come to us after years of buying translation and interpreting team by team, and the first thing we usually do is help them see the whole picture: every vendor, every spend line, every language asset scattered across the business. For more than two decades we have consolidated that sprawl into single, ISO-certified programs that span translation, localization, and interpreting, for everyone from startups entering their first market to enterprises operating in dozens of languages. This guide is the thinking we walk operations leaders through before any of that work begins. Read it as a framework first and a sales pitch never. If it helps you make a sharper decision even without us, it did its job.
Nobody chose the fragmented model. It just grew.
Here is the part everyone gets wrong about vendor sprawl. It is not a mistake anyone made. It is the absence of a decision.
Language needs do not arrive on a schedule you can plan around. A regulator wants a filing translated by Friday. A sales rep needs an interpreter on a call in twenty minutes. A launch in a new market cannot slip. In every case, the fastest fix is for the team to solve its own problem with its own vendor. Speed wins. A new thread of spending quietly opens.
And no single instance looks expensive. A few thousand dollars here. A per-minute bill there. An agency retainer folded into a marketing budget. Each one slides under the threshold that would trigger a procurement review. Only when you add it all up, which almost nobody does because the data sits in five systems, does the real picture appear. Fifteen vendors. Sometimes twenty. None of them share context, terminology, or any memory of what has already been translated.
So, the problem lands on the operations desk, not on one department’s head. No team feels the full cost. No team is motivated to fix it. That is the definition of a structural problem, and structural problems need someone with a company-wide view to solve them.
Two models, defined
Quick definitions, because most companies run a version of one without ever naming it.
Department-based (decentralized): means each team sources and pays for its own language work. No shared vendor lists. No shared budget line. No common quality bar. No single owner. Procurement often cannot even see most of the spend, because it shows up as scattered invoices under different cost codes. This is the default for almost every growing company.
Centralized: means the company pulls that demand under one point of ownership: a shared strategy, a managed vendor relationship or a small deliberate set of them, common standards, shared linguistic assets, and unified reporting. It does not mean one person does all the work. It does not mean teams lose control of their content. It means the governance, sourcing, and infrastructure live in one place, even when actual translation is spread across the organization.
The two models break in completely different ways. The department-based optimizes for speed and local control. The centralized optimizes for leverage, consistency, and risk control across the whole company. As you scale, the cost of the first compounds while the value of the second grows. Knowing where those lines cross for your business is the whole game.
What department-based language services cost
Scattered language work does not just mean bigger invoices. The real damage shows up in five places, and most of it never appears on an invoice at all.
You forfeit your biggest lever: Volume. Fifteen teams each buying small quantities at retail rates will always pay more than one organization consolidating that volume into a negotiated rate. You also pay twice for the same words. The same product description, the same legal clause, the same support macro gets re-translated by different vendors because nobody knows it already exists. Translation memory, a database of your previously approved translations, is what prevents that. Reusing it can cut the cost of repetitive content by up to 30%, according to analysts at CSA Research. But only if the memory is shared. In a fragmented setup every vendor keeps their own, and that saving evaporates.
Your brand voice fractures: Different vendors mean different translators, different style calls, different readings of your terminology. A customer meets your brand described one way on the website, another way in the app, a third way in support. That happens in every language you operate in. For a company pouring money into a coherent global brand, this is quiet sabotage. And it is un-auditable. With no shared standard, “good enough” gets defined separately by each team, and nobody can say whether your German or your Japanese content clears a consistent bar.
Your language assets walk out the door: Translation memories and term-bases are real corporate assets. They get more valuable the more you use them. In a department-based model, they sit scattered across vendors who have no obligation to hand them over. Switch agencies and the memory often leaves with them. You end up renting your own accumulated knowledge and losing it on every transition.
You carry compliance risks you cannot see: This is the one that should keep operations and legal up at night. In healthcare, finance, legal, and life sciences, a mistranslation is not a brand problem. It is liability. Picture an interpreting call for a patient over an unvetted line. Or a regulated document translated by a freelancer with no audit trail and no data security guarantee. Standards like ISO 17100 for translation quality, ISO 27001 for information security, and HIPAA for protected health data exist precisely to control this. Centralized sourcing lets you enforce them once, everywhere. Fragmented sourcing means every department is one shortcut away from an exposure nobody approved. In regulated fields like life sciences, specialized linguists already command 30 to 50% higher rates. The cost of getting it wrong runs far higher than the cost of getting it right.
You pay a hidden coordination tax: Someone in every department is doing a job they were not hired for: hunting vendors, chasing quotes, refereeing quality disputes, reconciling invoices. Multiply that across the organization and it is a meaningful chunk of senior staff time bled into amateur vendor management. It never shows up as a cost. It is one.
Here is what that looks like in practice. Picture a 400-person SaaS company expanding into Germany, Japan, and Brazil. Marketing localizes the website through its creative agency. The product team pushes UI strings through a separate tool and vendor. The support team stands up a per-minute interpreting line for inbound calls. Legal sends the new data-processing terms to a freelancer. Four teams, four vendors, four invoices in four budget lines. The German UI says “Konto” while the marketing site says “Nutzerkonto” for the same thing, because no glossary crossed between them. The terms of service get translated twice in six months, by two different people, because nobody knew the first version existed. None of them is anyone’s fault. All of it was avoidable.
The case for centralizing
Centralization is really one decision. Treat language as a managed capability instead of a string of one-off purchases. Done well, the benefits mirror the costs above, point for point.
Real leverage: Consolidating demand turns you into one significant buyer instead of fifteen small ones. That unlocks negotiated rates, volume commitments, and service levels no single team could land alone. It also kills duplicate spending, because a provider working from shared memory will not re-translate what you already paid for. And the upside is not only defensive: CSA Research found that companies investing in localization were 1.8 times more likely to grow revenue and 1.5 times more likely to increase profits than those that treated it as an afterthought.
Consistency by default: One glossary. One style guide. One set of standards applied across every language and channel. Brand voice and terminology stop depending on luck. This matters most when you expand into several markets at once, where the alternative is letting each market drift on its own. And consistency is a growth lever, not just a tidiness one. CSA Research surveyed 8,709 consumers across 29 countries and found 76% prefer to buy in their own language, while 40% will not buy from a site in another language at all. Patchy localization leaves that money on the table.
Assets that compound: When translation memory and terminology live in one company-owned place, every project makes the next one cheaper, faster, and more consistent. The asset appreciates instead of fragmenting. This is the most underrated financial argument for centralizing. You stop paying twice for the same words and start building something that lowers your marginal cost every quarter.
The chart above is the argument in one picture. Department-based sourcing keeps paying near-retail rates for the same content, quarter after quarter. Centralized sourcing starts a little higher, because you invest in setup, then bend as translation memory accumulates and reuse climbs. The gap widens every quarter. The numbers are illustrative, but the shape is real, and it is the reason “we are too busy to consolidate” gets more expensive the longer it holds.
Risk controlled at the source: Set your certification, confidentiality, and data-handling requirements once, then apply them everywhere. For an enterprise that means being able to prove, to auditors and regulators and your own legal team, that every piece of language work met a defined standard. Working with an ISO-certified enterprise language partner replaces a patchwork of unvetted freelancers with one accountable relationship. That is exactly what risk, and compliance functions want to see.
Speed, oddly enough: Centralization can make you faster, not slower. Teams stop restarting the vendor hunt from zero every time and draw on an established relationship with known turnaround and capacity. When something breaks, there is one owner, not a finger-pointing match between an internal manager and a freelancer. A coordinated approach to running multilingual projects means the routing, tracking, and delivery infrastructure already exists. The same logic covers spoken language. Consolidating under one on-demand interpreting service means support, sales, and clinical teams all reach a vetted, certified interpreter through a single channel, with consistent quality and billing, instead of each one improvising under pressure.
Where centralizing falls down, and when department-based is fine
An honest comparison has to admit what centralization costs. Pretending it is pure upside is how these projects fail.
Centralizing takes real upfront work. Mapping current spend. Consolidating vendors. Building shared glossaries and memory. And getting teams to give up some autonomy. That last one is the killer. Teams that have run their own language work for years feel centralization is a loss of control, and if the central function turns into a bottleneck, they are right. A badly designed central model, one that adds approval layers without adding capacity, will slow the business down and breed workarounds. The whole thing collapses the moment people decide it is easier to go around the system than through it.
Some cases genuinely favor department-based sourcing. A small company with light, occasional needs gains little from the overhead of a formal central function. A team with a rare, low-volume, highly specialized need, say a single obscure-language legal matter, may be better off with a direct specialist than a general pipeline. And a company still figuring out what its language needs are can benefit from a stretch of decentralized experimenting before locking anything in.
So, centralization does not always win. But the default, fragmentation by accident, is almost never right past a certain size. The choice should be deliberate. For most growing companies, deliberate points toward consolidation.
A decision framework
Skip the abstract “centralized or decentralized?” debate. Work through the questions that actually decide it. Honest answers usually point one way.
“Can anyone tell me what we spend?” If you cannot produce one consolidated number for company-wide language spend, that alone makes the case. You cannot manage what you cannot see, and visibility is the first thing centralizing buys you.
“How many vendors, and is the same work being done twice?” “More than a handful” and “probably yes” means you are paying the fragmentation tax in full.
“What is our exposure if a translation is wrong?” The higher the regulatory, legal, or safety stakes, the stronger the case for centralized control over certification and quality. In high-risk industries, this is not efficiency. It is risk management.
“How much does consistent brand and terminology matter?” If you are building a global brand and competing on experience, the consistency that centralizing provides feeds that strategy directly.
“How fast are we scaling, and into how many markets?” Faster and broader means fragmentation becomes unmanageable sooner. The cost of fixing it rises with every vendor and every market you add, so the cheapest moment to act is always now.
Hidden spend, duplicated work, real risk, brand stakes, fast growth. If most of your answers light those up, the case is made. What is left is execution.
The model that actually works: centralized governance, distributed execution
The best enterprises rarely go pure-centralized or pure-decentralized. They run a hybrid that captures the leverage of one and the responsiveness of the other.
The idea is simple to state. Centralize what benefits from scale and standards: vendor relationships, translation memory, glossaries, quality bars, security requirements, reporting. Let departments keep control of their own content, priorities, and timelines. Teams own what gets translated and when. The central function owns how it gets sourced, with what assets, and what standard.
That settles the autonomy fight directly. Marketing keeps creative control and gains a faster, cheaper, more consistent pipeline backed by shared terminology. Legal keeps oversight and gains certified, auditable workflows. Support keeps its responsiveness and gains a reliable interpreting channel that is already vetted. The central function wins not by gatekeeping but by making the central path the easiest path. People choose it because it is better, not because they are forced. That is the difference between centralization that sticks and centralization that gets quietly bypassed.
How to make the move, and how to know it is working
You do not need a disruptive reorganization. This works as a sequence.
Start by mapping the current state. Pull every vendor, every spend line, and every type of language work happening across the company into one view. This audit alone usually shocks leadership and builds the case on its own.
Then consolidate the assets you already have. Translation memories, glossaries, and style guides scattered across vendors, pulled into one company-owned foundation. Recovering those is often where the early savings come from.
Next, pick a primary partner, or a small set of them, that can handle the breadth: translation, localization, certified work, and interpreting, with the certifications and security your industry demands. The goal is a partnership that scales with you, not a vendor you outgrow in a year.
Finally, design the operating model so the central path is genuinely easier than going solo. Clear intake. Fast turnaround. Visible service levels. Centralization earns its keep by being the better option, not the mandated one.
- Measure it, because this is an operational change and deserves operational metrics. Track total language spend and cost per word before and after.
- Watch translation memory leverage, the share of content served from memory instead of translated fresh, because a rising number is recurring savings.
- Watch turnaround time to confirm the central path is faster, not slower.
- Watch vendor count, which should fall and stay low.
- Watch internal adoption, the share of work actually flowing through the central channel.
That last one is the real test. High adoption means teams find the central path better. Low adoption means you built governance that people quietly route around, and the fragmentation you set out to kill is reassembling itself one workaround at a time.
The real question is not whether. It is when.
Department-based language services are what you end up with when nobody chooses. Fast and autonomous in the moment. At scale, they leak budget, fracture the brand, scatter assets you paid to build, and quietly stack up compliance risk nobody signed off on.
Centralized language services are what you choose on purpose, once language gets too expensive and too important to leave ungoverned. Usually that means centralized governance with distributed execution: lower cost through scale, a protected brand through consistency, assets that compound, and real control over risk.
For a growing company, the honest answer to “when” is almost always “sooner than it feels urgent,” because fragmentation only compounds. If you have hit the point where nobody can tell you what the company spends on language, you have already hit the point where it is worth finding out.
So find out. Pull the invoices, count the vendors, and put a real number on what language costs you today. That number is almost always the start of the business case, and it is usually bigger than anyone expected.
And when you are ready to pull those fragmented vendors into one accountable relationship, Day Translations is built for exactly this. We bring translation, localization, and interpreting under one coordinated, ISO-certified roof, with shared memory, consistent terminology, and the security and compliance standards that regulated industries demand. Our enterprise and operations teams handle the consolidation work most companies dread: mapping current spend, recovering the language assets scattered across old vendors, and standing up a single intake that every department will actually use. On-demand interpreting runs inside the same program, so spoken-language needs stop being a separate scramble.
You do not have to rebuild your language operation overnight. You just have to stop letting it run itself. Start with the audit, then talk to us about turning fifteen invoices into one. Language stops being a hidden cost and starts being something you actually run.
Sources: Nimdzi 100, 2026 for market size and fragmentation; CSA Research for native-language purchase preference (“Can’t Read, Won’t Buy,” 8,709 consumers, 29 countries), translation-memory savings, and the localization revenue and profit findings; Mordor Intelligence for life-sciences rate premiums. All sources are independent analysts and industry research firms, not language-service vendors. Cost-per-word figures in the chart are illustrative.
