Value-Based Translation Pricing: How Agencies Are Moving Beyond Per-Word Rates
Value-based translation pricing replaces per-word rates with outcomes. How agencies build service tiers, quote fixed prices, and defend margin in 2026.
Value-based translation pricing comes up in almost every conversation we have with agency owners lately, usually about ten minutes after they finish explaining why their per-word rate hasn't moved since 2021. The shape of the complaint is always the same. Volume is steady or up, revenue is flat, and the client who used to pay $0.14 a word now wants $0.09 because "the AI does most of it anyway." Raising the rate is off the table. Matching the number kills the margin. So the real question turns out to be whether the word is still the right unit to sell.
We don't think it is, for a growing share of what agencies do. But we also don't think per-word pricing is dead, and most of the advice floating around on this topic skips the part where you actually have to make the numbers work. Here is what we've seen agencies do that holds up.
Why the per-word rate stopped describing the work
The per-word rate was never a description of value. It was a proxy for effort, and it worked because effort scaled fairly predictably with volume. Twice the words meant roughly twice the hours. That relationship has broken.
Take two projects we've watched go through the same agency in the same month. The first: a 12,000-word equipment maintenance manual, clean source, a mature translation memory for that client, glossary already built from three years of similar files. AI pre-translation plus post-editing came in at under six hours of linguist time. The second: a 2,200-word product launch deck, marketing register, four rounds of client feedback on tone, terminology decisions that had to be escalated to the client's regional office. That one took eleven hours.
At a flat per-word rate, the agency billed roughly five times more for the project that took half the effort. Nobody planned that. It's just what happens when you price an input that no longer tracks the work.
There's a second problem that hurts more over time. Under per-word pricing, every efficiency you build gets handed to the client automatically. You spend six months tuning a workflow, cutting post-editing time by 30%, and your invoice for the same job is identical. The client captures the entire gain. Slator and Nimdzi have both tracked steady downward pressure on per-word rates across the language services market since the arrival of production-grade neural MT, and the mechanism is exactly this: buyers can see that the work changed, and the per-word rate is the only lever they have to price that change into their contracts.
What value-based translation pricing actually means
Value-based pricing is not "charging more." It's setting the price from what the deliverable is worth to the buyer and how much risk you absorb on their behalf, rather than from how many units of input went in.
The clearest way to see the difference is to imagine two documents of identical length. One is a regulatory submission where a rejected filing delays a product launch by a quarter. The other is an internal newsletter. Same word count, same language pair, wildly different consequences of getting it wrong. A per-word rate prices them the same. A value-based approach prices the first one to reflect the fact that your client is buying certainty, and the second one to reflect the fact that they are buying comprehension.
In practice this means the price is built from a few things: what outcome the client needs, how much review and accountability you're putting behind it, what happens to them if it's wrong, and what your actual delivery cost is. Word count is still an input to your internal cost model. It just stops being the thing you quote.
This works best when you have some understanding of the client's business and a direct relationship with the person who feels the consequences. It works badly when you're a subcontractor to another language service provider whose entire procurement system runs on rate cards, or when you're bidding into a public tender that requires a per-word number in a fixed field. We'll come back to those cases, because they're common and pretending otherwise is not useful.
The three structures agencies are actually using
Almost everything we see falls into one of three shapes.
The first is the fixed project price. You quote one number for the whole job, based on scope rather than volume. This is the easiest place to start because it requires no change to how you sell, only to what appears on the quote. The client sees "translation of the 2026 supplier handbook into German, Polish, and Czech, delivered March 14, including terminology alignment and a QA report: €4,800." No arithmetic they can attack line by line.
Second, service tiers. You define two or three named levels of service and price each one, then let the client pick. A common shape: a fast tier with AI translation and an automated QA pass, suitable for internal comprehension; a standard tier adding full human post-editing against a glossary; and a top tier adding subject-matter review, terminology sign-off, and a documented quality score. The tiers are the product. The word count sits behind them in your cost model.
Third, retainers and capacity subscriptions. The client buys a monthly allocation of translation capacity rather than individual jobs. This suits accounts with recurring documents: monthly compliance updates, quarterly reports, a product catalog that changes constantly. We've covered how retainer and per-project models compare in more detail, but the pricing logic is the one that matters here. You're selling availability and continuity, and both of those are worth something the buyer can't get from a per-word transaction.
Most agencies that make this shift end up running two of the three. Tiers for new and mid-size accounts, retainers for the top five clients, per-word quotes still living quietly in the corner for the subcontract work.
How to build a tier when you can't count words
This is where the transition usually stalls, so it's worth being concrete. You cannot price a tier until you know what it costs you to deliver, and most agencies don't, because per-word billing let them avoid finding out.
Start by logging actual time on your next ten projects. Not estimated time. Actual, per stage: file preparation, glossary work, translation or post-editing, review, QA, formatting fixes, client correspondence. Two weeks of this is enough to see the pattern. What you're looking for is your effective hourly yield per project type, which is revenue divided by total hours, including the unbilled hours you currently pretend don't exist.
One agency we worked with ran this exercise and found their "simple" DOCX projects for two long-term clients were returning about €68 an hour, while their prestige legal work, which had the highest per-word rate in the price list, was returning €31. The legal jobs carried three rounds of review, endless formatting rework on tables, and a partner who read every delivery personally. Their most expensive product was their least profitable.
Once you have that baseline, price each tier at a target margin against its measured cost rather than against a competitor's rate card. If your standard post-edited tier costs you €340 in delivered labor for a typical 8,000-word manual and you want a 55% gross margin, the tier prices at roughly €755. Whether that reads as €0.094 per word is your business, not the client's.
Two things to watch. First, tier definitions have to differ in a way the buyer can perceive, otherwise everyone buys the cheapest one and you've just cut your prices. The difference has to be visible in the deliverable: a QA report, a named reviewer, a turnaround commitment. Second, your gross margin per tier will drift as your workflow changes, so this needs a recheck every couple of quarters. Our breakdown of realistic agency gross margin targets covers what the numbers usually look like.
Quoting a project when the client expects a per-word number
The awkward moment is predictable. You send a project price, and the buyer replies asking what your per-word rate is. Three responses tend to work better than deflecting.
The first is to answer the question and then reframe it. "It works out to about €0.10 for this file, but that number moves depending on the source quality and how much terminology work the document needs, so we quote the project rather than the word." You've given them the comparison figure they wanted while making clear it's derived, not fixed.
The second is to quote a range in the proposal itself, tied to scope. "Between €4,200 and €5,600 depending on whether you want subject-matter review on the technical annexes." Buyers who are used to rate cards find a range easier to accept than a single opaque number, because it shows them the lever they control.
The third is to move the conversation to turnaround and accountability, where per-word pricing has nothing to say. A procurement manager comparing three vendors on rate alone has no way to price the fact that one of them commits to a 48-hour turnaround on urgent files and stands behind a documented quality score. If you have those commitments, put them in the quote as line items with no price attached. They shift the frame away from unit cost without requiring you to argue about it.
What doesn't work: refusing to give any per-word equivalent at all, especially with a client who has been buying that way for a decade. It reads as evasion. Give them the number, then explain why it isn't the price.
Where per-word pricing still wins
Three situations where we'd tell an agency to leave it alone.
High-volume commoditized content with stable characteristics. If you're processing 400,000 words a month of the same product descriptions in the same format for the same client, effort really does scale with volume, and per-word is simply an accurate model. Don't fix what isn't broken.
Subcontracting to other language service providers. The buyer's entire system runs on rate cards and TM discount grids. A value-based proposal from a subcontractor is a proposal that doesn't get read. Price per word, protect your margin by controlling which jobs you accept.
Formal tenders with mandatory rate fields. Public sector and large enterprise RFPs frequently require a per-word figure in a structured form, and the evaluation matrix scores it. Fighting the format loses you the bid. What you can do is bid the rate and use the qualitative sections to build the value argument, so the rate is not the only thing being scored.
There's a fourth case worth naming: projects where scope is genuinely unknown at quoting time. If the client cannot tell you how many source files exist or what state they're in, a fixed price is a bet, not a price. Quote per word or quote a discovery phase.
What to do in the next thirty days
Don't rebuild your price list. Do this instead.
Pick your next ten projects and log actual hours by stage, including the unbilled ones. At the end, calculate effective hourly yield per project and sort the list. You will almost certainly find that at least one high-rate service is quietly losing you money and at least one low-rate service is your best earner. That single spreadsheet changes more decisions than any pricing framework.
Then take one client, ideally a mid-size one with recurring work and a direct relationship, and convert them to a fixed project price on the next job. Use your measured cost plus your target margin. Say nothing about pricing philosophy. Just send a project price instead of a word count and see what happens. In our experience the objection rate is far lower than agency owners expect, and the conversation you have with the client who does push back tells you exactly which parts of your value are legible to buyers and which parts you've been assuming they can see.
Everything else follows from those two things. You cannot price on value until you can measure cost, and you cannot sell a new pricing model until you've watched one client accept it.