Field Notes

The Real Price of Localization Quality

Stephanie Episode 7

Use Left/Right to seek, Home/End to jump to start or end. Hold shift to jump forward or backward.

0:00 | 13:54

“Quality” in translation sounds like a simple promise until you ask one uncomfortable question: what happens if it’s wrong? Stephanie and Erik unpack why so many localization pricing conversations get stuck on a single, vague standard, and how that confusion leads teams to either overbuy heavy workflows or quietly push risk onto vendors without paying for it. 

We walk through three distinct ways to define translation quality. First is output quality: the measurable correctness of the words on the page, often evaluated with MQM-style scoring or other quality assessment methods. Second is process quality: the workflow signals you can verify and price, like translator qualifications, translation memory health, terminology, and classic TEP steps. Third is the one that changes everything: fit-for-use or outcome-based quality, where “quality” includes warranties, liability, and who owns the consequences of failure across human translation, MT, light post-editing, and full post-editing tiers. 

Then we get practical about risk-based pricing. If your organization maps content on a probability-versus-severity grid, you can build multiple service tracks that match real business exposure, from cost-optimized low-risk content to high-assurance, high-liability work that deserves different safeguards and a different price. If you want smarter RFPs, cleaner vendor relationships, and fewer surprises when errors happen, this is the framework to start with. Subscribe for more Field Notes, share this with your localization or procurement team, and leave a review with the toughest “quality” question your org still avoids.

Three Ways To Define Quality

Stephanie Harris-Yee

Hello, I'm Stephanie, and here for another Field Notes episode with Eric. And today we're going to be talking about quality and pricing in a lot of different ways. So let's go ahead then, Eric, and can you tell us a little bit about what you're thinking around this topic?

Erik Vogt

Yeah, so I I think there's different ways of thinking about quality. And I think often we tend to think about quality as a measurable characteristic of the output. And it turns out you only can get that if you do like an MQM score or some kind of output test of some kind. So quality estimation is often intended to estimate what that outcome quality is. But we're looking at the word itself or the segment itself. Is this right or is this wrong? That's kind of category one of when we think about quality. And it's usually where we go to, like as buyers or as providers. There's another kind of quality, a second type of quality, which uh industry experts would talk about as process quality. So that's what was the machinery that led to a particular output happening. So that would be the qualifications of the translator, for example, or how many eyeballs saw it, or what was the quality of the TM, or you know, the workflow characteristics are intrinsic to the outcome. So if you have a very good translator, all you need to know about the quality is that you trust the translator. They're just that good. And we see that with certain industries where there's essentially a translator who certifies so the translation is correct. There's kind of a formal process for that. In that case, the name and reputation of that individual is your quality. Like that is knowledgeable enough. So it's still a process quality question. Same thing goes into automated translations as well, where certain process is what defines what that output is. But the simplest way to think about it is like one translator versus two, is it reviewed or not? And that sort of implicitly implies some quality characteristic. Well enough. But there's this probably like the most common that we see these days, especially for companies with big projects. So they don't know the exact translator because they have way too much content to rely on that one translator. So they're more relying on their language service partners, et cetera, to have kind of those structured steps on this is how we make sure the TM is good. We have the TEP translation, editing, proofreading. And we put more layers and layers on top, depending on, you know, the type of content. So I would say, and you can agree with me or disagree with me, that this is the standard for today on how at least the mechanics of having a quality process is put together. But sometimes people will throw like that first one that you mentioned, output quality at the very end of the process. So that is a component, but this layer process is how people have been thinking about it. Is that correct? Yeah, I think it's right. But I think one of the reasons why the process quality is so popular is because it's so easy. Because you know whether or not you did something. It's it's aligned with costs. So you can easily price things in a cost plus model. You pay this much for the translator, this much for the reviewer, then you get a price. It's relatively straightforward.

From Cost Plus To Fit For Use

Erik Vogt

But there's this third category, which I think is uh people are talking about it. So some people call it value-based or fit for use uh or outcome-based. I think um outcome-based means that there's some difference in how the consequence of failure is handled. And this is more and more of an issue, right? Because often buyers want different options, different pricing based on some quality metric. One easy one that Taos, I think, introduced originally was light post edit, full post edit, human translation. So it's implicitly three tiers. So tier one, I presume to be the best quality. Tier two, full post edit, better, but not best. And then third is cheapest, is the light post edit, or just raw MT. So that's four tiers for prices. But with outcome-based pricing, you can also make workflows that do something different depending on what happens if there's a mistake. Who owns the consequences of a mistake? That I think is an interesting area, which is a little harder for people to wrap their head around because it's not just a cost plus thing, but it is something that I think the industry demands. I think the market is demanding some more thinking with regards to this.

Stephanie Harris-Yee

Yeah, when you first said fit for use, I thought, okay, so like life science content, very bad consequences if something goes wrong. So then the client would pay more in order to have that work. But you're actually seeing something a little bit different. It's not just the add more layers on top just to make sure the quality is good. The consequences, if the quality is bad, does that go to the company that's paying for the translations? Does that go to the translator? Does that go to the machine somehow if it's MT translated? Is that kind of more of the concept that you're seeing with that use concept?

Erik Vogt

That's right. So very simplistically, we have some notion of this already. MT is raw MT, especially if you're the one who's emptying it yourself. You get what you get. And it's right or wrong, you use it or not. There's no consequence to it being wrong. You just simply are confused, but it it's on you because you're the one who used the MT. In the middle, you have the LSPs who often have a warranty-based quality strategy. And that's more of the long lines of if it's wrong, we'll fix it. If you have reviews, we'll do review changes guaranteed. That's kind of warranty-based, is the consequence of failure, is the vendor is accepting the cost consequences of fixing it if there's a defect. And it's a little bit like when you buy clothes, you can buy one set of clothes as is, right? So you you buy it, it's yours. If it's broken, it's your problem. In the second tier, you have sort of more warranty support. Well, there's a defect in the fabric. I'm gonna get a refund or I'm gonna get a replacement. That's kind of normal. And then there's consequence-based, which is if there's a defect, who incurs the liability of the problem? Now, this is happening in the back end, whether we like it or not. Like lots of contracts are including liability clauses. And sometimes some clients actually ask for unlimited quality like liability, which is no sane person would ever sign that, but they're asking for it. If your bad translation causes a problem, then that problem's on you, and you're going to incur the total cost of that. In a blunt level, I think most LSPs have some liability insurance and they'll sort of cover whatever, a million dollars worth of like liability costs. And that's one way of approaching it. But there's also there's, I think, even more need to think about this as we clarify what that should cost. So I think that right now clients are like, it's all on you. We want liability in case something goes wrong. But if you want it cheap, then do you do you also need the liability insurance? Maybe you don't. Maybe with raw MT, you're like, I'll be okay with this one being wrong. This one I kind of am okay with being wrong, but as long as you fix it, we're cool. And then in this third category is like, if this is wrong, then I'm coming after you because I need you to be accountable if this is wrong. Like I want your head if this is wrong. So, like, that's kind of more outcome-based kind of way of thinking about it. It's more and risk-based. So I think some clients, the more sophisticated ones, are starting to do risk grids where they have probability uh of something happening and the consequences of something happening. And they map all of their content according to

Risk Grids And Who Decides

Erik Vogt

this grid. And I think uh one grid I saw at a recent conference was like a five by five grid, very likely to happen to very rare, and then also very minor or negligible consequences to very severe. And then they'll map all of their content against one of these tracks. Now, what if we had a conversation in our pricing conversations of like let's map all this stuff out and build different tracks for each of these different risk types, which are a combination of our quality-based like measurement versus our process quality versus our consequence quality conversation?

Stephanie Harris-Yee

So like pulling them all in together. So from these grids that you've seen, who's making those decisions? Is that more like the localization managers just seeing all the content they get and then gridding it out themselves? Is that putting them in a position of liability as well? Or is it more like a full company conversation? Like what have you seen, or even wouldn't you recommend for this kind of thing?

Erik Vogt

Yeah, I think I think because there is a cost quality conversation. So with each of these tracks, there's consequences. If you're looking at the process quality, there's more steps involved, there's more sort of effort for the liability side of it. If I'm gonna take on more responsibility, I'm gonna be bonded for this. That's a cost that I need to pass on to my customers somehow. Need to build that into my pricing somehow. So each of these could have a conversation about the cost to risk category. To answer a question, I certainly hope localization representatives are not put in a position of having picked this decision themselves. Like I think that's a lot of a lot on their heads for that. I think they should facilitate the conversation, though, and they should be actively engaging with their internal stakeholders to create a rubric and to assign things based on that, and then to have some notion of the mechanisms by which you're going to guarantee. So there's some categories of stuff where you need a second vendor to do a full review because the consequences of failure are so high. They really need to outsource that risk. And the in order to do that, it's worth six, seven-figure relationships to make sure that liability is somewhere else. But that's what they're paying for, is that risk mitigation.

Stephanie Harris-Yee

Okay. So yeah, it sounds like there's a lot of these different quality levels. And to just to reiterate, you kind of are using all of the different quality levels. You're really just looking at what components go into that price tag at the bottom. So not just, you know, the process, not just the really good translator. It's also like who takes the risk. If it's low risk, that's adding less to the price tag because you don't have to have extra stuff in there. If it's high risk, you might need to have like the extra certifications or guarantees that will then add to that price tag. So that is like one more component of the pricing structure.

Erik Vogt

Exactly. And I think a lot so many

RFP Pitfalls And Better Buying Models

Erik Vogt

RFPs are pretty one-dimensional on this. They're like, what is your quality process? Period. And or they'll say, What is your what is your return? So they're handling all of our content under one exact same expectation for it is good quality, what is your quality? What do you do if it's wrong? It's all one bucket. What? But I think the differentiation of these could be a conversation that hiring that the procurement managers or localization managers could have with their LSPs of let's craft some different risk quality products that represent a range of needs for their business. And I think having an honest conversation up front about that means they don't have to overpay for quality that isn't adding value. Yeah. And won't underpay for risk that they're asking their vendor to incur for them. So if you're actually not talking about the characteristics of the sentence itself, because that is not actually what you're paying for, right? You're paying for confidence that I'm gonna get my money back or I'm gonna have you fix my problem if there's something wrong. So I think that that's the thing that we're looking for. It's a nuance of not just characteristics of the word itself, but what are you gonna do about it and then building that into your price relationship?

Stephanie Harris-Yee

Okay. Yeah.

Erik Vogt

Yeah.

Stephanie Harris-Yee

Well, I think it's pretty, pretty clear to me. And I know this is probably a discussion that could go on and on as we dig deeper into it, but is there any last points that you think folks who are listening to this should know or think about as they're approaching their own organizations and thinking about pricing and how they're going to be paying for stuff into the future?

Erik Vogt

I think that for those listeners who want to wrap their heads around this, think about your own buying as a consumer. Right. And what are you what would you pay more for? And sometimes people are like, I want the cheapest thing possible. I just need a box of nails. I don't really care if there's a bunch of nails that don't work. I just want them all to be nails. Like if I want it to be cheap, that's a more of a value-based model. Like I just want it cheap, like cost optimized this forward for that. There's others which are quality optimized, like I want the best of the best, and I'm willing to pay for it. I don't care what it costs. I just want to make sure it's the best. And there's value-based, which is kind of a balance between the two. So as we reflect and quantify our price relationship with our suppliers and our personal lives, we can get ideas for what is the appropriate model for buying and selling translation and language services or globalization services in a way that makes sense for the purpose by that you're spending that money on. So I think uh in general, we all can have a better conversation about this if we don't assume that everything needs to fit into one box.

Stephanie Harris-Yee

Yeah, for sure. All right. Thank you, Erik. And yeah, talk more next time.

Erik Vogt

It's great stuff. Talk soon. Thanks. Bye.