Translation Agencies, Don’t Penalize Your Valuable Resources

The most obvious measurement of translation performance and, therefore, the basis of determining remuneration of translation services is word count. (Since the smallest meaningful unit of a language is a word, the translator basically translate words (and sets of words called sentences), not characters, I therefore believe character-based pricing is non-sense.) If recycling comes into picture to support re-using legacy translations, vendors elaborate a cost model which applies discounts proportional to the similarity of the sentences (segments, in general) to translate to those translated previously. Similarity is organized into classes such as 85-94% [of similarity]: if the similarity of a sentence to one translated previously falls into this range, a discount of cc. 30-70% is applied against the words constituting that sentence. Count the number in the source segment to translate and multiply it with 100–discount, and you get the weighted word count or WWC: for a 10 words long sentence in an 80% match class for which 30% discount applies, WWC is calculated as 10 times 100-30=70%, so it’s 7 weighted words.

This has a long tradition and works well: if you get a translation assignment with lots of high matches, you are paid less, but the time to spend on the task is proportionally less over the long term, and a great amount of your work is only to edit and adjust the previous translations by replacing or deleting a word or two, adding a few other words and aligning affixes, instead of translating from scratch. Fine!

An example for an 80% match

What many translation agencies seem to fail to realize, however, is that this method falls short for proofreading and review. While translators need to focus on changes between the previous and the current source text to adjust the previous translation to be that of the current source text, proofreaders and reviewers need to check all the segments as if they were completely new. They don’t know what that specific translator has done whose translation they are currently checking. Since the translation memories contain the translations made by the translators, they don’t even get a chance to see what pre-translation the translator was provided with, so they have no way to focus on changes. Just take a look at the screenshot, the translator can clearly identify what to focus on. If this were a screenshot from a review session, there were no blue parts, yet a significant discount were applied for the 80% match.

Yet, many agencies require reviewers and proofreaders to work for a word-based fee, where word counts are calculated using the same figures and cost model which were used for the translator, only the unit price (that of a “no match” word) is less, say half o the translator’s unit price.

This is a very bad practice which pushes proofreaders and reviewers, which should be of higher quality resources than translators once they review translators’ work, to a detrimental situation to get paid against discounted words for which they do not receive any benefits in exchange.

Proofreaders and reviewers usually do not mention the problem to their agencies to not risk losing assignments, but agencies should realize they penalize their higher quality and more valuable resources using this method.

And look! The solutions are rather simple: such assignments can be compensated for the time spent on completing the proofreading/review session (rather good capacity approximations can be set so you do not have to be afraid of your proofreader or reviewer overcharging you), or by finding a lower flat unit price for a word, but not applying any discounts. Both would result in a more fair situation.