Yes, I want my consultation

Inherent Problems of the Translation Industry: It Is Tempting Not to Pay for 100% Matches

Photo by 401K2012We recently finished a translation with one hundred thousand 100% matches. Our client, a translation agency, asked us not to review them. This is quite a usual request for a translation agency, and until recently, I thought it was more or less okay not to be paid for 100% matches. But as we continue to improve the quality of the translations Velior produces, it has become obvious that skipping 100% matches—however financially attractive it might seem—is a poor decision from the quality standpoint.

Assuming 100% matches are automatically correct is dangerous

Here are three reasons why:

Poor segmentation

To expect that 100% matches are always perfect, one must believe that segmentation in the original translation was perfect, too. But this could not be further from the truth! First, many people in the translation industry either do not understand segmentation, because it is too technical, or do not think it is important. As a result, no one takes the time to look at source files and check whether segmentation is appropriate. Second, agencies tend to prohibit translators from changing (i.e., correcting) segmentation, since any deviation from the automatic segmentation means that in the next project, automatic segmentation will not match the segmentation in the TM. The result of this “garbage in”—incorrectly segmented stuff going into the TM—is “garbage out”; that is, 100% matches in the next project that do not make any sense. Let’s look at a simple example:

First segment: Development

Second segment: Ideas

Imagine that a phrase “Development Ideas” is incorrectly broken into two segments and is being translated into Russian. Because the Russian language commands a different word order than English, a translator will have to put the translation of “Development” into the second segment and vice versa. As a result, “Development” will be translated as “Ideas” and “Ideas” translated as “Development” in the TM. When in the next project, “Development” occurs as a standalone word and is translated as a 100% match, it will be in fact translated as “Ideas,” which is a 100% mistranslation.

Differences in context

Since the meaning of a phrase often depends on the context, the meaning changes when the context changes. For example, “Areas of focus” means one thing when used to describe areas of interest of a public opinion survey, and quite another thing when it means areas of focus for employee development. Even though the English text is the same, translation into many languages will be different. Of course, if no one reviews 100% matches to make sure they fit the new context, some of them won’t.


Some tend to perceive any TM as a source of reliable and approved translations. But this is not always the case. Just because a translation comes from a TM does not mean that it is automatically correct. 100% matches must be reviewed for errors for at least one specific reason: When they were originally created, a translation team might have had less context than is available in the new project, and by looking at this new, additional context, the team might realize that the old translation is inaccurate.

XTranslated units/perfect matches/in-context matches (ICE)/101% matches/guaranteed matches

Despite the different names, these matches mean one and the same thing—a 100% match that occurs in exactly the same context; that is, the same location in a text. A match that also takes into account the context usually eliminates all the issues I described above and does not require revision as a result (while still requiring time for the translation team to read it). But the problem is that some agencies choose to ignore the distinction between the two types of matches. When a project has both in-context matches and simple 100% matches, an agency might request skipping all 100% matches, as if all of them were in-context matches. Of course, while skipping in-context matches is relatively safe, skipping simple 100% matches is too dangerous.

Quality assurance

If a translation team is required to skip 100% matches, it might be difficult to do so when it comes to automatic quality assurance, because QA is done on the entire translation by default, including the 100% matches. Even though many QA and CAT tools do have an option of skipping these matches, they still might present a challenge in terms of QA.

Translators need to know the context

Last but not least, a translation team needs to read 100% matches, even if it is not required to do so by the client. Reading the 100% matches is necessary to see the context, the big picture of the project. Without reading them, the new translations will be inaccurate, especially those that are adjacent to 100% matches. For this reason alone, the translation team must be compensated for 100% matches.

This same reason makes it absolutely unacceptable to remove 100% matches from the files. If they are removed and the translation team cannot see the original context and the existing translations, quality inevitably takes a hit.


Even though I understand why agencies may not want to pay for 100% matches, I do not think this is the right approach from the quality standpoint. If quality does not matter, then not revising and not paying for 100% matches is fine, I guess. But if it does matter, ignoring 100% matches is not an option. A translation team must be compensated for at least reading them, or, better yet, for full-fledged revision. Here is what I recommend:

Translators: Consider not accepting jobs with unpaid matches. If you accept them, you will likely spend more time on those jobs than you will be paid for. And even if you and your client agree that by not paying you for 100% matches, you are not responsible for any problems with them, the client will feel that any errors are partly your fault anyway. If refusing such jobs is not an option for you, at least educate clients as much as possible.

Translation buyers: It is important to pay something for 100% matches. This may be a high rate, if you want your translation vendor to do a full revision of 100% matches. Or this may be a lower rate, if you just want to compensate the vendor for the time spent reading those 100% matches to know the context. Whatever your situation is, some form of compensation is in order.

Dear readers, what do you think? Under what circumstances will it be okay for you to accept or send a job without paying for 100% matches?

For more information about the inherent problems of the translation industry, read the previous articles in this series:

Replacing a Translator? Couldn’t Be Easier

Agency-Driven Deadlines


  • Hello, Roman,

    Thank you for another good article!

    Our company sometimes sends out documents to a translation agency that our in-house translation team cannot deliver on time. Although we did agree with that agency that 100% or higher matches are 25% of the regular rate, the agency always delivers them free of charge. I suspect the reason for this is that:
    a) there are not many of them (10% of the overall scope, max);
    b) they are usually superior to the agency’s work in terms of quality;
    c) they provide terminology and stylistic information that the agency would otherwise take time to obtain by themselves.

    Overall, I agree with your statements. Better tools, such as those that clearly differentiate between ICE/101% matches and 100% matches, or allow locking of segments to exclude them from QA operations, would certainly minimize the agency’s time and help in setting more optimum pricing for in-context matches versus regular 101% matches. Still, the translator will still need to use consistent style and terminology throughout the document, and this is impossible without reviewing existing translation, which should not be free.

    Best regards,

    • Dear Stanislav,
      Thank you for your comment.
      > c) they provide terminology and stylistic information that the agency would otherwise take time to obtain by themselves.
      Right! This is one of the main things I like about projects with TMs. A good TM does save time and frees from potential hours of terminology research.

  • I often say that the best tool for translators, since long before anyone ever thought of CAT, is COMMON SENSE. It helps us in both the simplest and the toughest decisions, as well as those in-between.

    There are some cases where repeated segments justify the translator giving them for free. Here are two examples of mine, and there are possibly more.

    1. Parts-list-type translations

    I had this situation in my second full-time job. This was long before the microcomputer (and consequently CAT tools) age. Also, as I was earning a salary, it was a matter of cost (my time), not price. Of course, these translations were done from hard copy, using a typewriter (later someone would re-type them on a typesetter).

    I received “international” parts lists in SV/EN/FR/DE; had to deliver them in PT only. As the company was Swedish, the parts names were translated, however only the SV had the part specifications (e.g. Hose 2-1/2″ – 20 m). So I had to translate from EN and get the specs from SV. After a while, I had memorized the part names in SV, so that whenever I got a new product’s “domestic” parts list in SV, I could translate it directly SV > PT without a hitch. For the record, I wouldn’t know how to say “Good morning!” in Swedish.

    I had a similar job – as a freelancer – just a few years ago, when I had to translate an immense list of trade description labels and their subdivisions. Repeated segments were 60%+ of the job. This time I had a CAT tool to do most of the work for me.

    2) Training programs courseware (my #1 specialty in translation)

    In such cases, this is absolutely normal: The very same segment gets repeated on a) the course leader’s guide; b) participants’ workbooks; c) handouts; d) PPT slides; etc. Otherwise the entire course will be a mess, compromising retention.

    In the old days when clients were wallowing in cash, and CAT tools weren’t developed nor popular, clients hired me to go the full nine yards, translate and DTP them all.

    When cutting costs became a forceful trend, they had me translate only the leader’s guide, which supposedly contained most of the phrases. Then they had a staff member of theirs painstakingly copy & paste phrases from my translation onto the other pubs, as well as attempt translating anything left behind. Of course, this led to somewhat lower quality (attempted translations), and higher costs (the staff member’s time on their payroll) than “my” solution.

    My solution was to offer them all the repeated segments for free, on jobs larger than 5,000 words (a threshold I arbitrarily set). They’d get the complete program uniformly translated, and I’d get the extra translation of any text not included in the course leader’s guide.

    3) Some market surveys, standard financial reports

    I’ll leave it to the reader to figure out the details on these, which should be pretty obvious.

    Therefore, out of common sense, in SOME cases it is fully justifiable not to charge for repeated segments, if they should definitely represent – at most – one keystroke each for the translator (or using auto-translate until fuzzy). However this certainly doesn’t apply everywhere.

    P.S.: I don’t give ANY discount for fuzzy matches, ever!

    • Hello Jose,
      Wow, this is an impressive comment, thank you.
      You are talking not about 100% matches, but repetitions. You are making a very good point anyway, though. Repetitions lend themselves to hefty discounts much more than 100% matches. The main problem about 100% matches is that no one wants to accept responsibility for, and deal with the challenges created by, 100% matches translated by someone else. But this is not the case with repetitions, since it is always you who made the translation. You just need to check whether it fits the possibly new context.

  • Olga says:

    This is a great and very insightful post, Roman! While understanding all the advantages that a CAT tool gives, I hate receiving the “part of the project” in a server based remote CAT environment, where I basically just translate the segment, assigned to me. I always need the full picture – and you are of course right, I want to be paid for looking at the full picture, especially when it comes to translating complicated technical texts, which is my case usually. I thought I’m the only one not really understanding how you can save on matches and just let them transfer into a new text from previous projects. Even matches from the same text are not safe, and your segmentation example is a very strong prove to that! Working in Wordfast I found that often what is marked as a 100% match has differences with the source text. So my rule is to never EVER overlook the green fields 🙂 And I surely want to be paid for that!

  • I. BRUCHER says:

    Not paying for any kind of fuzzy matches, whether 75% or 100% or context matches, or paying less for them is a theft if your basic rate does not cover your research time, which is often the case, since those rates are based on the earlier rates, which have never been high, which GLOBALLY covered a translator’s time (without making him a millionnaire though).

    So now there are many translation agencies asking for “your best rate” AND the use of a CAT tool: this is totally unacceptable!

    CAT tools have been designed to rob translators. SDL, who design and sell the Trados CAT tools, is ALSO a translation agency. It only takes one second to realize that they design their tools to get rich themselves, at the expense of the translator, who is the one who actually provides the core of the work!

    Also, SDL and the like tend to call themselves “Language Service Providers” (LSPs) when addressing themselves to end-customers, which shows how dishonest they are. They profile themselves as the ONLY language service providers, whereas translators do the actual work! I.e. they are ROBBING OUR CLIENTS!

    And they are robbing us on rates. And robbing our TM files…

    I would recommend all translators to spend some time every month to send mailings to end-customers and charge like double their usual rates, since this is the margin of those “Linguistic Sausage Producers”, as they should be called !

    And if you get too much work, just contract it out to colleagues, functioning as an agency, but UNDERSTANDING THE JOB: a world of difference in interpersonal relations, rates and the whole business.


    And throw away those CAT tools. Using the Autocorrect macros in MS Word is just as efficient and does not cost an extra cent.

    Backup larges entries in a MS Word file (with a reference number for each of the files you will have to create, by language pair + a number).

    Backup your Autocorrect macros regularly. You can find them in a path something like: C:/users/yourname/App(lication)Data/Roaming/Microsoft/Office.

    You can save typing accelerators, templates, everything you want in Autocorrect macros (originally designed to replace usual typing errors by the correct word; here you just replace the source word/text by the target word/text, especially if you always translate into your one and only mother tongue, but you can enter translations in several languages, just giving them the language code between brackets for example).

  • Denise Muir says:

    This is one of the main reasons I left one of my main IT>EN agency clients after more than 10 years of working happily for them. The “race to zero” and drive to offer the lowest possible rates took over and destroyed the agency-translator relationship.
    1. I was fed up being held partly responsible – continually – for errors in the translation resulting from unpaid 100% matches.
    2. I was fed up being asked to proof the entire final translation of texts which often contained more exact match content (unpaid + not mine) than new input. So, I was re-reading and making corrections to thousands of words that I didn’t translate and wasn’t being paid to proof.
    3. I was fed up having to put my name to a final document splattered throughout with mistranslations, out-of-context words and phrases, consistently inconsistent translations (which happens when the same large documents are put out time after time to different translators, with each one using their own – and often different words for a given word/phrase – because there is no time or budget to check exact matches for uniformity.)

    In general, the final product was often a mix of good work and garbage. This gets quite soul-destroying after a while, especially when your working hours are getting longer and longer, and the agency client reproaches you more and more for the garbage part, never recognizing the good part… because it may not be consistent with the garbage!
    I just hope there are some agencies out there who might read this.

    Denise Muir
    Magicamente Translations

    • I was fed up being asked to proof the entire final translation of texts which often contained more exact match content (unpaid + not mine) than new input. So, I was re-reading and making corrections to thousands of words that I didn’t translate and wasn’t being paid to proof.

      Terrific addition, Denise, this is so true. I did not realize this until recently, but after a couple of projects where we had translated a few new and updated pages here and there and then were asked to proofread a 200- or 300-page PDF stuffed with unpaid 100% matches, I have learned the lesson the hard way. Readers, beware.
      Thank you so much for your comment.

Add comment

About the Author

Roman Mironov
Roman Mironov
CEO & Founder

As the founder of Velior, Roman has had the privilege of being able to turn his passion for languages into a business. He has over 15 years of experience in the translation industry. Roman has helped dozens of clients increase sales by making their products appealing for speakers of other languages.