Why garbage in your translation memory is so costly

Most of our clients write materials that are updated quite often. An example would be the translation of operator’s manuals or help systems in support of software releases. I can not stress enough how important the quality of the content going into the translation memory database is in this situation.

Yes, I know the old adage of “garbage in, garbage out” is overused. But it really makes sense in this situation. After all, translation updates are really dependent on leveraging previously translated content. The whole idea of choosing a professional translation provider rides on this basic idea. If you can reuse existing content with 100% certainty you can save money, save time and be more consistent from document to document.

Here is how it works. Let’s say the chapter on safety and warnings is identical to your previous release. A properly maintained translation memory will simply place all the content that was previously translated into the new translation. That should mean no charges and no effort for this part of the manual.

So let’s say that in the interim, someone from another department translates a similar document but hires a different vendor, doesn’t share the existing translation memory and does no due diligence on the vendor. The only deciding factor was price so the low cost vendor was chosen.

The new translator completes the project, turns in the document and the translation memory. Unknowingly, the new translation memory entries are checked into the system and overwrite the original entries. Now the next time this content comes up for translation, your vendor (perhaps the original vendor) relies on those 100% matches. If the vendor still thinks it his work he may rely completely on the matches and price the project with no review of the material. After all, why would it be included, that safety section has been translated many times, it has been reviewed and approved, why include it in the scope?

Now subpar translation may propagate into your system and it may spread like wildfire. Here are a few steps you can follow to prevent this catastrophe!

  • Pick a vendor and stick to that vendor. Knowing who is handling your translation over a given period of time is huge. It will give you an audit trail for any issues that arise and will also give you the confidence in understanding how to treat context matches, 100% matches and fuzzy matches. Simply discuss what types of matches you would like locked and not reviewed. Most clients will choose to lock context matches and have the rest of the content reviewed by the editor.
  • If a problem comes up, react with purpose. From time to time you may need to discuss an issue or errors with your translation provider. It is really good to discuss any issues you might have, plan for a solution and give your vendor a chance to adjust the approach. If that doesn’t work, time to change vendors. If your translation memory is properly maintained, switching vendors will not prove to be a huge ordeal.
  • Periodically ask for a backup of your translation memory. It is always a good idea to get backups of your translation memory. Most of our clients ask for the resource quarterly. That is more than reasonable. If your vendor refuses to provide translation memory, switch vendors. You should not be held captive by your translation memory. If you are switching vendors, please be sure to discuss the ownership and the ability to receive backups of the memory periodically.
  • Be very careful in allowing others to add to the translation memory. Be sure you know the quality of the translators or provider adding to your memory. As described in our example above, a few horrible projects can cause deep corruption to the quality of your translation memory and compromise your ability to rely on the database matches. This will cost you time and money to fix.

If you want to learn more about translation memory and other important topics related to proper translation management, contact the team at Argo.