Choosing Language Service Providers (LSPs)

Just as you need to select translation technology to support localization, you’ll also need to select language service providers (LSPs) who will do the majority of the translation, review and post-editing work.

The LSP landscape is huge. It is fundamentally a professional services business, so you don’t see a few giant companies that have effective cornered the market. Instead there are hundreds of companies spread across the world that are specialized by language, region and industry vertical. This is one of the things I like about working in this space. It is an open, competitive market.

What Do LSPs Do

Before evaluating vendors, it helps to understand what LSPs actually do, and equally important, what they don’t do.

LSPs recruit, manage, and quality-assure linguists (translators, editors, language leads)
They provide project management : receiving files, coordinating review cycles, delivering on deadline
They do not provide translation technology : the TMS is your infrastructure; they work within it, and they are generally not developer or engineering led.
Good LSPs bring domain expertise and regional knowledge; they are not interchangeable commodity suppliers
The talent pool overlaps significantly between LSPs : the differentiation is in project management quality, domain specialization, and account service

Factors To Consider In Your Search

Language focus : most LSPs have true depth in 5-10 languages, and some have a regional focus, adequate coverage elsewhere.

Domain expertise : it is common for LSPs to have an industry focus such as: tech/software, medical/pharmaceuticals, automotive, entertainment, etc.

Service focus : LSPs often focus on a specific service. For example, LSPs that focus on entertainment will be more experience with dubbing and subtitling video than a company that focuses on the automotive industry.

Location and time zone coverage : LSPs are scattered all around the globe, and many have a language or regional focus. This is important to consider because of time zone and scheduling factors. When I was at Notion there were two language leads, one in Korea and one in Europe while I was in California. This enabled us to communicate with regional vendors during their business hours, getting faster turnaround and better responsiveness.

Right company size : mid sized LSPs are usually the best match for early stage and scaling companies. The largest LSPs like RWS Moravia and Transperfect focus on large accounts and will not prioritize you. You want them to be big enough that they can meet your needs without getting swamped, small enough that the revenue you bring in is meaningful for them.

Service levels required : smaller LSPs can have trouble with short SLAs or bursty demand while mid-size and larger LSPs can offer a range of service levels and absorb demand spikes.

Comfort level with your tech stack : be sure to evaluate their proficiency in your TMS environment. Beware if they are trying to steer you to a favored TMS because that is often a hint that they prefer a different platform, which may be because of a financial relationship, or lack of experience. LSPs also evaluate TMS platforms on different criteria than client side buyers do.

Translation Costs

A few words about translation cost. The reality is that LSPs draw from an overlapping pool of freelance translators and editors. LSPs that compete on price tend to alienate the good translators. Things are more confusing now that everybody is using AI and machine translation as part of most workflows, which makes direct cost comparisons tricky.

One thing to consider is per hour pricing for AI review and post-editing. While per word metrics are easier to use in estimating, they don’t really map to AI-human flows. For example, one edit might involve just correcting mismatched gender in an adjective, while another might involve jargon that requires time to research, even though both are one word edits.

The other thing to do is to search for the agency in discussions on translator forums like ProZ. If an agency has a history of underpaying or not paying translators, it will be discussed there. Translators keep in touch with each other and do not hesitate to call out bad actors.

Pro Tip : this is a great task for AI. Ask it to compare comments on ProZ about agencies A, B and C and write a report quantifying issues with payment to get a read on how well or poorly they treat people. Caveat: some people love to complain but you should get some signal this way.

So what’s reasonable?

High end, domain specific translation work : expect 30-40 cents per word for all human translation and proofing, or $90-$120 per hour
Transcreation : usually prices out at 30-40 cents per word, or $90-$120 per hour.
Standard translation and post-editing : usually prices out at 10-20 cents per word, or $30 to $60 per hour.
AI or MT with human edits : usually prices out at under 10 cents per word, or under $30 per hour.
Fully automated AI or MT translation : this is where costs really drop, to a fraction of a cent per word (it doesn’t really make sense to think of in terms of an hourly rate here).

Most customers ask about using AI to reduce unit costs. I discuss this at length in Combining AI And Human Translation.

Things NOT to do

Don’t buy translation tech and translation services from the same vendor. This turns into a hostage situation and it is expensive to switch vendors if needed. When you decouple translation tech and services this is just a configuration change.

Assume machine translation will dramatically reduce costs. Its biggest advantage is speed, but it needs to be part of a fix forward workflow, at least for content that meets a visibility threshold.

Search and Evaluation

Vendor Search

Any LSP worth engaging will have bona fides. They mostly draw from overlapping talent pools. Use AI and web service to find candidate firms based on your needs, regional focus and industry sector. If you have access to Nimdzi Insights, they are a localization consulting company that tracks the industry and vendors.

Before you start any work, be sure to develop a comprehensive glossary of terms and a style guide. This makes a huge difference in translation quality whether you are using human or AI translation. The first project the LSP will usually do is to translate the term glossary so that brands and jargon are translated consistently.

Initial Evaluation

Onboard the vendor and monitor work product over an extended period of time. They tend to stack the decks for demos and front-end engagements, so you won’t know how they are really performing for a while. I generally recommend a 2-3 month evaluation period involving real content. I am wary of demo projects because LSPs will often put their best people on a project to close a deal.

This is why it is good to have more than one agency (and let them know so that keeps them honest).

The main criteria for evaluation will be:

Overall quality : this is where language leads can help by spot checking vendor work product to calculate the number of defects. Note vendors will tend to improve as they learn your brand, jargon, etc.
Deliver speed and SLA compliance : measure their delivery speed and SLA compliance. It is not unusual for this to vary by language.
Responsiveness and project management ability : this is more subjective but it’s important because a core function LSPs provide is to recruit and manage translators.

Ongoing Evaluation

You should continue to evaluate your vendors on this criteria and look for trends in these metrics. AI is causing a lot of upheaval in the space and a vendor that’s doing well this quarter may hit headwinds that affect service levels next quarter.

A Few Favorites

Acclaro

I’ve worked in the tech / software space for most of my career, and that’s been a focus of my localization work. Acclaro focuses on SaaS companies, and has done localization work for leading companies such as Netflix and OpenAI.

Entre Les Lignes

ELL is a EU based boutique agency that focuses on European markets and has experience with marketing content. When I was at Notion, we hired them to work on our Dutch and Nordic language release.

Mother Tongue

Mother Tongue focuses on transcreation, which is different than translation. This involves adapting and rewriting content for a region. This is a specialized skill and involves copywriters who are native speakers in the target region. I like to bring them in alongside an LSP to work on high-visibility, high-touch content such as the marketing site landing pages, lifecycle comms and signup funnel.