How Translation Memory Saves You Money & Improves Quality

Category: Technology

When you’re expanding your company’s reach into international markets, quality translation is a sound investment for making a good first impression, protecting your brand identity and preventing gaffes that might give your competitors an inadvertent advantage. Most managers tasked with finding global profits find themselves (or their bosses) asking: Is there anything we can do to control our costs without sacrificing quality?

The answer is yes. Translation memory (TM) is a tool that can not only reduce expenses, but also improve consistency (read: quality) with future translation projects.

Once you understand how translation memory works, you’ll see how partnering with a professional translation agency that leverages it will save you money and give you a competitive advantage in your target markets.

Translation memory: Haven’t we met before?

In a sense, translation memory turns human translation into superhuman translation. Translation memory software is a kind of toolkit for professional human translators, enhancing their skill sets through a variety of features.

The core feature of translation memory software is a database of every English sentence and its translated equivalent in any number of languages. Translation memory makes stored translations available within a project and can be leveraged for updates to the project and cohesion across projects. With this database available, a human translator using translation memory software can quickly “pre-process” new English text against stored translations and reuse similar phrases. Given that the cost of translation projects is tied to the volume of content to translate, you can see why having these “pre-translated” phrases in translation memory is like money in the bank.

In addition to saving the cost associated with translating content from scratch, there’s also a secondary benefit: Consistency. At times, your project might involve multiple translators. Translation memory can guarantee that crucial phrases are not translated differently. This can be a huge issue when moving between product information online and printed customer documentation, for example.

Don’t confuse translation memory (TM) with machine translation (MT), by the way. Raw machine translation is produced without the guiding hand of a human translator, and though often used in large-volume projects with post-translation editing, it is not a sound solution for your brand’s high-value, customer-facing content. Translation memory augments human translation power, and as we’ll see next, it still relies on human judgment to produce the best translation.

Refresh me, would you? My memory is a little fuzzy.

Translation memory’s ability to save you money depends in part on the quality of the matches it makes between its database of stored phrasing and the new content you’re translating. It’s a kind of sliding scale: When there’s a 100% match, the translator can quickly approve a translation and move on. But what happens when the match is less than perfect? “Less than perfect” means a “fuzzy” match. We break fuzzies down along two major categories: High fuzzies and low fuzzies. In both cases, the TM tool will highlight the difference between the old and new text so the translator can translate any new text and/or edit the changes from the previous translation.

With high fuzzy matches, a small amount of the text and/or formatting is changed and it requires a small amount of editing.

In a low fuzzy match, there’s less reusable content in the translation memory and thus more of the translator’s time will be required to complete the translation.

Without a fuzzy match, the translator will need to provide a fresh translation. This is typically the case when the new phrasing has less than 75% in common with the translation memory’s source. Fortunately, if you’re using translation memory, this fresh translation will be available for future translators, so you’re actually building potential “translation equity” with each 100% human translation.

Even if your project produces a high number of low fuzzy matches, the savings compared to a brand new translation can be considerable. And let’s not forget: all TM matches require less time than new translation, meaning your project’s time line can be shortened, providing you with a speed-to-market advantage.

Déjà vu for any project

Almost any project can benefit from translation memory. Technical documentation, where simple sentence constructions and syntactic consistency abound, is ideal. In software localization, translation memory helps keep terminology consistent and can leverage exact matches based on string IDs across new versions. Website localization projects are excellent candidates as well, especially in rolling out updates based on changes made in your translation memory database. Translation memory tools can look at previously translated material and compare it to these updates, making corrections almost automatically.

Beyond the raw volume cost savings of translation memory, there are secondary benefits that can contribute to your bottom line. Numbers in charts, tables, and specifications are passed through without any opportunity for human error. (The transposition of a few numbers in an industrial manufacturing process could be an expensive, litigious disaster, for example.) Also, the amount of post-translation formatting can be reduced, especially if the electronic files have been handed off in such a way that the content can be extracted from the formatting (as in XML documents).

It’s good business sense to make your translation dollars do double duty by making sure that your translation partner is investing in translation memory. As an aid to professional human translation teams, it produces results with higher linguistic consistency on tighter deadlines. And when you’re eager to make your mark in new regions, there’s precious little time to say the same things twice.