Most Language Service Providers would agree that you could measure translation quality at least to some degree by counting the number of errors. However, not all errors are equally serious of course. A good place to start if you are seeking to measure translation by numbers is to define quality. To do that inevitably involves some means of ‘sizing’ errors.  For example:

High: Where the meaning would be completely wrong with a potentially serious impact e.g., million as unicorn

Medium: May affect the company’s reputation or the reader’s ability to understand

Low: Probably won’t affect either of the above, but there is a mistake

You can then allocate some kind of scoring e.g., 10 for a High error, 5 for a Medium, and 1 for a low. You would then have a simple mathematical formula to calculate the error score in a piece of translation. You would multiply the number of low errors x 1, the number of medium errors x 5 and the number of high errors x 10 and then add it all up. This would then have to be given some kind of weighting, though. Otherwise, it would be unfair on longer passages with more words. The simplest way would probably be a direct percentage of the word count.

In this way, you can achieve a score that will enable you to compare the accuracy rating of two 5,000 word translations for example. The question is, if one has more errors in it, does it necessarily mean that it is not as good as the other? Not necessarily. There is no information here to indicate the level of technical complexity of the pieces, or indeed whether there is any repetition. The piece with the higher accuracy rating might be translating some recurring terms for example.

Access to technical glossaries or terminologies that are industry specific or unique to a company can greatly help in international markets. This isn’t really a reflection of accuracy at all, but of the relationship that is built between the language services provider and the client. In this instance, the translation quality could be heavily impacted by something that the translator themselves would be unlikely to have a direct influence on.

There are of course other types of error as well as meaning and terminology. Grammar errors and spelling errors do both occur and are less ‘excusable’ in that with proper checking software they should all be identified and corrected. Yet, we know that is still not the case. It would also be fair to say that a piece with a higher number of spelling and grammar errors would be generally agreed to be of a lower quality to one with a lower number. This could also be clearly measured, so to that extent quality can be clearly measured and fairly so.

What if we delivered the perfect piece of work, without a single error on it… but after the deadline that the client specified? That has failed to meet the translation quality standards required too, so all the accuracy in the world can only get us so far. As well as the deadline set by the client, it is certainly relevant to look at the number of edits received from a client, and the tone of their overall feedback. There is not much point being ‘right’ if we are not listening to the client’s requirements at the outset.

There will always be an element of personal style when it comes to language, and this makes it hard to measure entirely by numbers. Accuracy can be measured to a large extent by numbers, but unfortunately, it can only get you part of the way.