ProZ.com Translation Forums

Tricks with analysis and wordcounts? | Another trick

MoonMartin — Fri, 04 Sep 2026 09:08:00 +0000

Forum: Localization
Topic: Tricks with analysis and wordcounts?
Poster: MoonMartin
Post title: Another trick

Another trick to lower invoices is to use MT to cover for a lower fuzzy price, but without previous consent deciding that high fuzzies are not based on the regular price for new words, but on the discounted MTPE price.
Example:
a) 100 words of 95-99% x 0.1 euros per word x 20% of full price = 2 euros
Agreed MTPE discount is 50% of full price (0.1 euros) = 0.05 euros
b) 100 words of 95-99% x 0.05 euros per word x 20% of full price = 1 euro

I discovered this after several months (trust) and a quick estimation was a loss of several thousand euros.
What is strange is that some other accounts at this reputable bureau used a), not b).

The explanation for b) was that the fuzzy match originated from MTPE and should therefore be considered as MT.
Which it definititely is not. When MT is edited to become a "human translation" the benefit of MTPE discount is already exhausted and the edited segment is considered as a "human translation".

[Redigerad 2026-09-04 09:13 GMT]

Tricks with analysis and wordcounts? | How can Phrase MT decide fuzzy percentage?

MoonMartin — Fri, 04 Sep 2026 08:57:00 +0000

Forum: Localization
Topic: Tricks with analysis and wordcounts?
Poster: MoonMartin
Post title: How can Phrase MT decide fuzzy percentage?

Posted elsewhere here:

I have seen several dubious uses of MT and analysis in Phrase.

Example 1:
Previous translations of segments have a lower fuzzy rate than Phrase MT.
These often have higher fuzzy rates although the "translation" isn't as good as the previous human translation.
Consequently the analysis shows higher fuzzy rates than if it would just use previous TMs.
It generates smaller sums on invoices and often more work for the translators.
It avoids the general consensus that MTPE word price is an agreed percentage of the full price for human translation of new words.
This can amount to several hundreds of thousands of euros less being invoiced, depending on the source text and workload.

Example 2:
A job with a few segments has one segment of 100 words translated with MT and the fuzzy rate is 80%.
How does the MT evaluate that the rate is 80%. It doesn't refer to any segments in the TM.
It creates an invoice of 100 words x 0,1 euros per word x 50 % MT discount = 5 euros,
instead it creates an invoice of 100 words x 0,1 euros per word x 35% fuzzy discount = 3.5 euros.

Tricks with analysis and wordcounts? | yes

Gregor Trebec — Fri, 04 Sep 2026 07:22:00 +0000

Forum: Localization
Topic: Tricks with analysis and wordcounts?
Poster: Gregor Trebec
Post title: yes

From my experience agencies try to cheat (also) on the wordcount, yes. I've had an agency that stated in the job ad even 500 words less than actual. And that is no mistake. It is deliberate.

Tricks with analysis and wordcounts? | Added experiences

MoonMartin — Thu, 03 Sep 2026 13:37:00 +0000

Forum: Localization
Topic: Tricks with analysis and wordcounts?
Poster: MoonMartin
Post title: Added experiences

Nobody?
It's hard to notice this, but when a job takes more time than your experience and gut feeling tells you, you should look into it.
The other irregularity I've noticed is to apply a lower agreed MPTE rate on high fuzzies.
If the gut feeling tells you smth is wrong, check the PO. All fuzzies are regular, not MTPE.
Myself I discovered it about half a year later and the loss was staggering. I trusted them, but at a point in time they changed the payment without any notice.
Beware.

[Redigerad 2026-09-03 13:37 GMT]

Tricks with analysis and wordcounts?

MoonMartin — Wed, 02 Sep 2026 06:58:00 +0000

Forum: Localization
Topic: Tricks with analysis and wordcounts?
Poster: MoonMartin

Hello!
Have any of you experienced that MT translations are moved in to another TM and then presented as high fuzzies in the analysis to avoid paying the going rate for editing MT?