In-house translation – outside the box

Category: Outcome-based translation

  • Beware of sharks? Translation agencies

    Beware of sharks? Translation agencies

    Last month, I returned from a family holiday in the Cote d’Azur with my family and our luggage. As I waited to drop off our cases, I opened my e-mail on my phone. My checking mails is not an earthshattering experience for the ages. The mailbox in question is used for articles that I write and submissions to calls for papers and conference admin. Standing in line, a particular mail caught my eye.

    Normally, I would have deleted it without a second glance. However, I was standing underneath a beautiful shark installation suspended from the ceiling. And there was a certain degree of sharkish practice about the mail. It requested me to make an offer for a translation job. Clearly they’d not visited my website. The mail’s siren song was how it showed “old school” some language services still are.

    It also radiated a tone and professionalism that was vastly different to the slick “cold calling” mails LSPs send me at work. More often than not, I receive those messages only if someone forwards them to me. Some LSPs are very persistent, or lecture me on what they *think* I/we need. Some take any response as permission to add me to their mailing list. And then usually inform me about their pseudo-cutting edge use of technology, followed by offering discounts for immediate jobs (in quiet periods).

    The mail

    Now back to that mail I received in Nice. the mail body below is redacted to remove as much information that might identify the agency. However, it provides clear insights that suggest that some agency practices haven’t changed since my freelancer days. It makes me wonder how agencies survive when their practices are so firmly entrenched in the “good old days”.

    The descriptive redactions of numbers are deliberate to also convey the output-based focus of the job in question. But at the same time the redactions are necessary to ensure the job is not identifiable.

    Mail sent with a mail header “Admin @” admin@[REDACTED MAIL DOMAIN] around the end of working hours in the local time zone. Sent as bcc: with no special reply to instruction.

    Subject: Quotes for translation

    Dear Translators,

    We have a [MEDIUM DOUBLE DIGIT NUMBER] page German historical document containing [LOW FIVE DIGIT NUMBER] words. The client is anxious that it should be translated correctly and wants us to translate a few pages before proceeding with the whole document, can we get a quotes for [LOW SINGLE DIGIT NUMBER] pages approx [MEDIUM-HIGH THREE DIGIT NUMBER] words.  Please state your best price and turnaround time for the sample [LOW SINGLE DIGIT NUMBER] pages and the whole [MEDIUM DOUBLE DIGIT NUMBER] page document separately. 

    Kind Regards,

    Administrator

    Mail received from an agency about a project

    And now for the appraisal about the content of the mail.

    Who do you think you are talking to?

    The partial mail header speaks volumes: a half-hearted admin e-mail, sent out close to end of the working day. It fishes for offers from (potentially desperate) freelancers. “Dear Translators” is an unwittingly honest indication of the half-heartedness. The agency makes no attempt to personalise the mail, but admits to carpet bombing its list of potential leads. Full disclosure: the mail included the agency’s mail signature, but no indication of who anyone was dealing with. Similarly the website felt like an “acme” WordPress site, with no real human personalisation.

    Red flags

    There were a few choices of wording that rang further alarm bells within the mail.

    “historical document”

    The mail mentioned a “German historical document”. The label “historical document” could be anything from a few months old through to a few centuries. The latter might indicate that the text is in either a non-machine readable form requiring transcription. It may even be in a different (handwritten?) script, e.g. Sütterlin or Fraktur. Some premium translators do offer such specialist value-added services. An agency choosing to omit this kind of information impedes someone from being able to submit a sensible offer.

    Translation from a well-transcribed document is a different game from having to decipher the written text prior to translating. In this case it is unclear whether the translator will only translate, or may also have to transcribe. Similarly, there is no indication about permissible use of AI for OCR (as the vague subject matter provides no hints).

    “translate a few pages”

    “The client is anxious that it should be translated correctly and wants us to translate a few pages before proceeding with the whole document […]”

    Understandably, clients don’t want to fork out for a full translation that quickly emerges to be unusable. However, this should set an alarm bell ringing for potential translators: a short initial request may not lead to a full document translation. On a positive note, at least we are not quite in “unpaid sample translation” territory. However, in contacting many translators, the possibility remains that the agency unscrupulously “slices and dices” to the lowest bidders. Of course, I can’t rule out that the client’s intention is simply to avoid unnecessary translation and cost. It is entirely possible that only a small section really needs translating.

    “best price”

    In the Internet age, “best price” has become a euphemism for a low (dumping) rate. In stating this, the agency unwittingly says “we’ll take the cheapest translator we can”. The agency in question appears to have scraped my e-mail address from LinkedIn. They clearly didn’t read anything other than “German > English Translator” (why would *they* contact an in-house translator?). The “best price” approach encourages teaser rates from desperate translators. Translators offering teaser rates only increase agency margins.

    What they don’t say.

    Almost refreshingly, there is no mention of technological requirements. Not that proprietary TMS that the translator has to learn on the or a request for MTPE. But it also doesn’t say anything about the kind of (internal) review processes they have in place.

    OK cleverclogs, how would *you* have worded the e-mail?

    Here’s the approach I would have taken. The second numbered list is illustrative of the information that would be useful.

    Subject: German to English Translation – [Description of document type]

    Dear [Forename] [Surname],

    A client requires a German to English translation of a historical document. Subject to a satisfactory translation of an initial section of X pages (XXX words), the client will request the translation of the full document (XX pages / XX,XXX words).

    Please provide two separate quotations, including expected turnaround times for:

    1. Initial translation of X pages (XXX words)
    2. Translation of full document of XX pages (XX,XXX words).

    To assist you in drawing up your quotation, we are able to advise that:

    1. (No) transcription of the historical document will be required. (It should be noted that the document is written in Sütterlin/Fraktur).
    2. The document is (not) available in a machine-readable translation format.
    3. Due to the sensitive nature of the document, clearance must be obtained from the client about any potential use of AI (e.g. for transcription, OCR, pre-translation) 
    4. [… further information that might assist comparability of bids, as well as further information about potential value-added services that may need to be provided – this information actually helps reduce questions being asked by return e-mail]

    Kind Regards,

    [Forename] [Surname] – Project Manager [Agency Name]

    Better call an expert in the lead…

    The mail is a typical tale from the uberised / over-commoditised world of translation. A client understandably wants quality, but not wanting to pay a premium rate and not get it. Maybe they have had their fingers burned before. But maybe the opted for an agency to simplify a task (possibly their customer didn’t know what they needed). They probably chose the first agency they found in a search engine. An approach that doubtless meant they missed out on working directly with excellent freelancers nearby. So perhaps I ought to give the agency some credit, as the agency’s website was findable.

    The agency’s mail contains a gymnastic (over-)stretch: delivering quality translation cheaply and quickly (based on Burns’ Iron Triangle of expectations high quality and cheap means slowly). However, the agency gets its run-up all wrong: the incomplete information in the mail fails to inform potential translators whether this is the job for them. If the client approaches expert translators directly, there would be a short dialogue with the translator by phone/e-mail/etc. From that dialogue, the translator would establish the information required. And would be in a position to say whether they could in fact deliver the quality translation.

    Expert translators take the time to establish information from clients. This is part of finding the right fit. I’d be put off by the lack of context that would prevent me from being able to make a proposal. By that doesn’t seem to be the agency’s intention. After all, the mail is devoid of Information about the historical context (year / decade / significance / political structure), provenance (country/region, and therefore a linguistic bearing). This is in addition to the type of document (e.g. official or personal), what the translation is needed for (in turn clarifying the need for a partial or full translation), and background information about the person(s) involved.

    Wakey-wakey, rise and shine!

    Freelancers: read this!

    If you’re a freelancer and chasing agency leads for low rates like this, here’s a wake-up call. Do you really want to be that generalist fighting for any scrap that comes your way? It’s time to reconsider how you operate. Choosing to deliver translations at low rates, also blinds you to that shark coming for you. You need to rethink your strategy. Even if you outswim other fish in your shoal, or remain at its centre, your future prospects are not looking great as you too tire.

    I know these are difficult times, and the “Sommerloch” periods can be notoriously slow. But they can serve as an opportunity to really work on your marketing. Look critically at that tired website, which looked good ten years ago. The cutting edge of pre-NMT days of 2017, is not where the work is. We’re in the data-driven age, and you have two options. Either find customers who aren’t after the data-driven approach, or find ones you can demonstrate how you can harness their data in translation, while not blinding them with technology.

    Your website, even if only intended as an online business card, must show that you mean business. Think about that big fish you want to land, and what image you project to them. I frequently look for suitable freelancers for potential outsourcing of translations. As part of the procurement process, I have to do an appraisal assessing whether they can be considered as a serious provider. In case you are wondering about some of the things I look for in a translator’s website, check out this blog post. And, please don’t blow your professional image with a free e-mail address when you’re dealing with high stakes confidential texts.

    And to the old school agencies…

    The signals given off that you run an agency like it is still 2005 also won’t help. As someone who worked for a number of agencies until the Global Financial Crisis, I know fully about business models relying on the margin/cut from squeezing your astronomical array of freelance translators. The freelancers I worked with back then, if still in translation, long since moved towards working with direct clients. Some of the providers I worked have also exited the market. Those hollow claims about covering every language pair also don’t ring true. Customers are more demanding nowadays. They want to get bang for their buck, without lengthy feedback chains, and dealing with real experts. They want to know who they are dealing and working with (and not the faceless admin account!).

    Recently, some translator communities I lurk in and skim through have had increasingly frequent threads about translators being owed money. The shift is that it is no longer the “here today, gone tomorrow” agencies, but the long established ones. And I don’t mean the mid-sized ones that BigTransla snap up. Or the ones they rescue but doesn’t take on their debts. I’m talking the old school ones, where old school ran out of road. You need to be giving the impression of expanding and rising to the way the market is going, rather than fading away.

  • Outcome-based translation: don’t (just) show me the numbers…

    Outcome-based translation: don’t (just) show me the numbers…

    I recently completed a questionnaire for benchmarking purposes regarding a potential reorganisation of a multi-person language services unit. The wording hinted at a fundamental re-evaluation of the role of human translation in a specific institutional setting. There were a mixture of qualitative and quantitative questions. Quantitative questions sought to benchmark the language service in question’s productivity against that of its peers. They highlight output, but do not necessarily measure the qualitative nature of the translation.

    This ties in with a theory discussed in Sam Illingworth’s “Slow AI” (part of my Summer 2026 Reading List). Output devoid of an outcome is a constituent part of slop. A couple of years ago, I started actively questioning the necessity of specific translations as an outcome. This was partially due to a takeaway that I didn’t need to do so much of the mundane grunt work. It made more sense to deploy me on high stakes jobs, and where the output was reusable rather than single-use. It was a way to critically assess that I wasn’t translating publications unnecessarily. I also argued that I could use it to prove that I was not artificially busy.

    I can always find something that can be translated. It also proved an effective visibility tool. As English versions of publications were released, I began dropping comments on LinkedIn and other channels about English versions of publications. This ensured that English versions of documents were read. In some cases, I used download figures to confirm/question/disprove the need for future translations. Sometimes a translation of the executive summary or localising downloadable data would suffice.

    Translators work best when trusted rather than watched.

    LinkedIn posts often boast about productivity and output levels – expressed in figures. However, this may be an attempt at validation rather than being a meaningful metric. A trainer decided to recently post how someone who had completed his course wanted to mention the size (in words) of a project on social media. I jumped on the coat tails of an esteemed colleague, who highlighted the importance of outcome rather than output in terms of value.

    Initially, when I went in house, I recorded daily/weekly translation volume and jobs. It did serve a purpose, namely to chart peaks and troughs (and help me plan when I go on leave and not return to an overflowing inbox. I no longer log translation output in such detail now, although reports are available “on tap”. In this regular, in my responses to the questionnaire, I mentioned that I comfortably exceeded the initial benchmark (from memory, set at 200 words/hour). Once it was clear that I consistently exceeded the benchmark, I argued to have the benchmark dropped. I argued that the margin by which I exceeded the benchmark and the non-translation tasks I also performed showed the shortcomings of a benchmark based on apparent 100% productivity.

    Benchmarks are changing

    On the subject of benchmarks, the goalposts have moved: GenAI and MT solutions make claims about high volumes that they are able to process. Human translators are not there for “out and out” high volume translation, but for the translation that affects outcome. Occasionally I will still have 30,000 word items to turn around in 4-5 days. However, mature subject-specific translation memories and terminology databases and MT for pre-translation helps significantly.

    My involvement centres around ensuring accountability and consistency. I also become the “Most Critical Reader” who asks the difficult questions to the authors. I query numbers that don’t add up (or become skewed) and question ambiguities of the source text. Increasingly I also condense texts, check their coherence, and ensure that the message gets through.

    Suggesting changing benchmarks where necessary

    I also recently responded to a Call for Input on a translation rates survey for a translation association. Previous editions of the survey have helped me regarding procurement for working with freelancers. Clarifying rates as “going rates” in a best bidder procedure has been an important step. In preparation for the finalisation of the 2026 iteration of the survey, I also submitted input in light of the changing climate for freelance translators. One fellow member had made a number of very sensible suggestions. My additional suggestions consisted of:

    • Have you considered changing tariff models, such as no longer charging by line (Normzeile)?
    • How do you bill additional technical services (e.g. extra delivery of bilingual files or alignments)?
    • How do you bill for transcreation?
    • What proportion of your work comes from direct customers and what proportion from agencies?
    • What proportion of jobs can you handle in your preferred CAT tool, and what proportion in a customer-prescribed CAT tool?

    When publication is only the cherry on top of the icing on the cake

    I recently published a 100,000 word translation of an Austrian law that I had worked on and revised for a number of years. Clearance was ultimately only a formality that had always just slipped through. It could have been published with a disclaimer years ago. However, I used subject matter experts signing off my work to establish dialogue with them. Yes, a six figure word count is still a landmark, in the same way I remember past quantitative landmarks. Amazingly it is over 25 years since I remember receiving payment on my first ATS 50,000 translation job (EUR 3663.64). Other weekly, monthly, quarterly and annual landmarks ensued. And yes, I still remember my first “million word year” with pride and exhaustion.

    Usage rather than publication is key

    However, personal pride about the recently published translation’s value didn’t stem from clicking upload in our DMS. And it was not about its monetary value at a normal line rate. Pride already existed from knowing the role played in its various iterations over the last decade. It had been in use internally, with periodic legislative amendments, for years. It contributed to outcomes like the orderly exits of failed banks, or in discussions of our Austrian approach in countless European working groups. Finally, it served as the basis for IMF country inspections and exercises evaluating national transposition of the European legal framework.

    Ultimately, publication on our website was the transparency cherry on top of the icing on the cake. I still get a buzz when people use my translations. And that is why I had always consciously made working translations available internally to ensure their active and effective use. That also generates critical feedback. A translation’s usefulness is also measurable by the frequency of its usage as a reference work.

    There is more to output than just a word count

    Word count, along with character count ultimately resonates more of a charging unit of output than as a unit of outcome. Transcreation is a classic example of this. Nina Sattler-Hovdar’s recently published “Transcreation: The all-in-one handbook and workbook” contains a detailed breakdown about pricing aspects. Just to take one mini-example: slogans become part of a brand identity and cannot be priced by word. For every chosen slogan, there will be numerous others suggested and rejected within the project. Royalty-based remuneration is very rare in transcreation (i.e. effectively licensing the use of a slogan for a specified period of time). With translation, the impact may lie in the impact of the brevity of the translation and whether it communicates the desired message – a case of “more is less” and naturally language lengths vary considerably.

    And as for those qualitative aspects?

    The focus of my presentation in Spiez, and my subsequent blog post, was how translators ensure that they are not only an afterthought. Similarly, when I responded to the questionnaire, I felt it necessary to highlight the value translators bring to an institution. This value increases by involving translators at the core of a publication or procedure, rather than as an afterthought.

    Good and useful translation work is not a one-off expense, but is an investment. Sometimes, the return on the investment is not immediate, but accumulates gradually over years – as with my recently published translation. In environments when translation doesn’t convert directly into sales, its qualitative role is in improving or bolstering institutional reputation. At the same time, for the questionnaire, I highlighted about what is lost without having translators. However good the grasp of subject matter experts, they still do not always have the grasp of the language to handle the nuances required, ensuring there is no ambiguity.

    I’ve encountered people who claim that AI has helped their (non-native) language skills – however, they are the ones most at risk of deceptive fluency – something that Sam Illingworth calls “superficial competence”. At the same time with GenAI, translators are on the receiving end of what Illingworth describes as “asymmetric effort”. I’ve often referred to it as cognitive downstreaming. In this case, translators take on the role of the most critical reader, if the author and/or their editor abdicates it.