Translation memory discounts: yes, no, maybe?

The issue of translation memory discounts, whereby a translator charges a lower rate for words that appear as repetitions or fuzzy matches in a translation environment/CAT/TM tool, is a contentious one. On the one hand, a client might reasonably argue that changing “press the green button” to “press the red button” doesn’t involve translating four words, but rather one word. On the other hand, a translator might reasonably argue that reading that sentence, finding the difference and changing it takes about as much time or maybe even longer than just typing the translation from scratch.

Translators and clients have a few options to choose from: no TM discount at all; a TM discount only for high level matches such as 100% matches and repetitions; or a graduated pricing structure where fuzzy matches down to a certain percentage, say 75% or even 50%, are priced at a certain percentage of the translator’s per word rate.

At a past ATA conference, I heard a translator argue that there are two translation markets out there: the TM market and the non-TM market, and you have to decide which one you’re going to compete in. While I’m not sure that the choice is as binary as that, I think there’s some truth to the idea. Many legal and financial translators I know are rarely if ever asked to give discounts for repetitions, and use a TM tool mostly for their own consistency and productivity. On the other hand, many technical translators I know would be out of business if they didn’t use TM and give discounts for repetitions.

For some projects, such as product manuals that are regularly updated, it seems to me to border on ridiculous not to give TM discounts since the majority of the text will be recycled. However, I understand translators’ frustrations with clients who nickel and dime over every repeated word simply out of a desire to pay as little as possible for the translation. Your thoughts?