Pages in topic:   [1 2] >
Do you give reduced rates for CAT tool generated fuzzy matches?
Thread poster: sencho1999

sencho1999
United Kingdom
Local time: 21:41
Japanese to English
Nov 19

Hi everyone,

I've had several enquiries this week from LSPs regarding translation work. Which is great! Except I'm lost when it comes to pricing.

I'd really like to get back to one agency in particular fairly soon regarding my rates. Apart from my standard rate(s) for Japanese > English translation per Japanese character, I don't know how to approach the following information request below.

Any help or advice either would be greatly received.

... See more
Hi everyone,

I've had several enquiries this week from LSPs regarding translation work. Which is great! Except I'm lost when it comes to pricing.

I'd really like to get back to one agency in particular fairly soon regarding my rates. Apart from my standard rate(s) for Japanese > English translation per Japanese character, I don't know how to approach the following information request below.

Any help or advice either would be greatly received.

This particular LSP has asked for my rates for the following:

1. GBP per source character (Japanese): This is fine I'm confident on my rates for this.

Next they are requesting (expecting?) discounts based on CAT tool usage. I use SDL Trados Studio 2017. They want me to quote my rate(s) in GBP for the following matches.

A fellow translator told me that they charge full standard rate for any matches that are less that 85% - I don't know if this is reasonable or not?

2. CAT tool fuzzy matches
No Match: Obviously charge my full rate
50-74%: Charge full rate?
75-84%: Charge full rate?
85-94%: Should I give a discount for this? If so how much?
95-99%: As above
100%/Repetitions: As above

So I guess I'm asking do you agree to discounts for matches and if so how do you calculate a fair discount? Should we give discounts at all?

How will they know how many matches MY own TM will generate anyway?

How do I quickly find out, after I've translated a document in SDL Trados Studio 2017, how many matches there were and how many are in each percentage match category? Won't this make working out the final cost of the translation time consuming and hard work?
Collapse


 

Samuel Murray  Identity Verified
Netherlands
Local time: 22:41
Member (2006)
English to Afrikaans
+ ...
@Sencho Nov 19

sencho1999 wrote:
A fellow translator told me that they charge full standard rate for any matches that are less that 85% - I don't know if this is reasonable or not?


Yes, it's reasonable and not unfair to either party. The categories lower than 85% are sometimes useful for informational purposes. Some prefer to set the bar at 75% instead of 85%.

85-94%: Should I give a discount for this? If so how much?
95-99%: As above
100%/Repetitions: As above


One suggestion may be:
75-84 = 0.6 of usual rate
85-99 = 0.3 of usual rate
100/rep = 0.1 of usual rate

Do you agree to discounts for matches ... ? Should we give discounts at all?


It's up to you if you want to give discounts. If you do give discounts, you may get more clients or more jobs.

How will they know how many matches MY own TM will generate anyway?


Don't include your own TM's matches in the calculation. Analyze your files against the client's TM only.

How do I quickly find out, after I've translated a document in SDL Trados Studio 2017, how many matches there were...


Matches are calculated beforehand, not afterwards. You can attempt to calculate them afterwards, but it'll be harder.


Dan Lucas
Angie Garbarino
 

TranslateWithMe
Poland
Local time: 22:41
English to Polish
+ ...
Yes, but... Nov 19

Hello

I would say matches below 75% are pretty useless for translation and must be basically rewritten.

So I usually charge 10% for 100% matches and repetitions, as in my language if something is 100% match it does not mean that it is fine and I must at least have a quick look at it.

Not to complicate the counting later for 75% - 99% range I charge 50% of the basic rate

Everything below is 100% of the basic rate.

K.


toasty
Christine Andersen
 

Bernhard Sulzer  Identity Verified
United States
Local time: 16:41
Member (2006)
English to German
+ ...
Don't let agencies bully you Nov 19

Samuel Murray wrote:


Yes, it's reasonable and not unfair to either party. The categories lower than 85% are sometimes useful for informational purposes. Some prefer to set the bar at 75% instead of 85%.


None of these discount schemes are reasonable, really.
If you get offered a scheme that equates with the rate you would charge "per character" or "word" without giving discounts, then maybe it makes sense in a ridiculous kind of way.

Translators need to stop letting themselves be bullied by agencies.
You decide what you charge for your work, period. And that should be by word, or by character, or by hour. You know how much work will be involved. Matches of words in the original text don't equate to matches in the translation - we're not shifting oranges from one place to another. Every professional knows that. And just because "everybody" (not really) gives in to these schemes, doesn't mean you have to. But if you're already caught in the wheel of falling rates, it won't be long until you hit the bottom. My two Cents.


[Edited at 2019-11-19 16:16 GMT]


philgoddard
Elena Aclasto
Oleksandr Ivanov
Christine Andersen
 

Dan Lucas  Identity Verified
United Kingdom
Local time: 21:41
Member (2014)
Japanese to English
Fairness Nov 19

sencho1999 wrote:
So I guess I'm asking do you agree to discounts for matches and if so how do you calculate a fair discount? Should we give discounts at all?

Samuel has offered useful advice on the matrix, so I shall not repeat what he has said.

I myself do offer fuzzy matches for JP-EN translation, and I don't find that it disadvantages me. That is, the benefit I perceive to extract from my CAT tool makes up for (or outweighs) the discounts given for the matches. Moreover, I have been told explicitly by certain clients that they unambiguously prioritise those freelancers with CAT tools.

My experience is that the impact of discounts for fuzzy matches varies from project to project. On some projects I find I get more leverage than suggested by the fuzzy matches. Let's say for the sake of argument that the software hasn't picked up that two phrases are juxtaposed one way in the TM, and the other way in the current segment, with very little other difference. Basically I can copy and paste them in the right order, change a preposition and I'm done. I get the benefits of the TM without it being discounted. Conversely, sometimes although there's an 82% match it's actually quite convoluted to get it done right. But over time I find it's worth it.

What I find is that some clients nearly always send packages rather than a "raw" file, and some send them only occasionally. So sometimes - if you have created a TM for that client and it turns up a lot of matches - you get the benefits all to yourself. I do advise you to build up such client-specific TMs and termbases if you have good flows of work. I do that for almost every job, so I have hundreds of them.

Bernhard makes his usual point about fairness, which is in itself fair. I would amplify the point about not allowing yourself to be bullied. If you are, like myself, prepared to offer fuzzy matches, then don't hold back from working with the client just because they want them. However, I suspect that, like the rate itself, it is almost impossible to change afterwards, so one needs to think carefully. (You are, obviously, otherwise you would not be here.)

Finally, I would add that the industry background to this is very similar in concept that of Toyota Motor's approach to cost reductions. It expects its parts suppliers to cut costs, and it expects that it will receive some, but not all, of the benefit of those lower costs in the form of lower prices for parts. The point is that the overall cost of the vehicle comes down, the vehicle becomes cheaper (or the vehicle stays the same price but offers additional features), demand increases, and thus the market grows, and the entire supply chain benefits. Far from being unique to translation, pressure on prices is endemic to commerce. The trick is doing it in a way that is sustainable for both parties.

Regards,
Dan


MollyRose
Angie Garbarino
 

Kay-Viktor Stegemann
Germany
Local time: 22:41
Member (2016)
English to German
Depends on the field and more Nov 19

The real usefulness of fuzzy matches depends on the field, the client, the language pair, even your personal style, so it's not easy to quantify. In my experience, TM matches below 85% are mostly useless, so I would try to avoid any reductions here. TM matches of 99% or higher or outright repetitions are obviously less work, but don't accept a "free" rate for anything. Even "102%" repetitions need review and checking and sometimes changing. Watch out that you don't work for free. There is some t... See more
The real usefulness of fuzzy matches depends on the field, the client, the language pair, even your personal style, so it's not easy to quantify. In my experience, TM matches below 85% are mostly useless, so I would try to avoid any reductions here. TM matches of 99% or higher or outright repetitions are obviously less work, but don't accept a "free" rate for anything. Even "102%" repetitions need review and checking and sometimes changing. Watch out that you don't work for free. There is some trial and error involved in this, so you should try to "err on the safe side" ...Collapse


Dan Lucas
Christel Zipfel
Nikki Scott-Despaigne
Oleksandr Ivanov
Jan Truper
Rachel Waddington
MollyRose
 

Christel Zipfel  Identity Verified
Member (2004)
Italian to German
+ ...
Yes, this is eminently reasonable Nov 19

sencho1999 wrote:


A fellow translator told me that they charge full standard rate for any matches that are less that 85% - I don't know if this is reasonable or not?




Everything below 85% is mostly garbage for translation purposes (except terminology) and you lose more time than you gain with your SW. Not to speak of the good money you have probably paid it and the time you spent in learning it. Why should your customer profit from all your possible advantages?


philgoddard
Nikki Scott-Despaigne
Oleksandr Ivanov
woodstar
 

Nikki Scott-Despaigne  Identity Verified
Local time: 22:41
French to English
Restrictions and careful calculations Nov 21

Each tool calculates matches with slightly varying parameters but the main idea is the same. The previous two posts (Kay-Viktor and Cristel) make important points. There has to be an interest in it for the translator too. If the interest is that we buy the expensive tool, we do more volumes for less money for the client and that we also help to make the management of language projects easier for our clients, that's fine. It should not be to the extent that the mere fact that a word is repeated i... See more
Each tool calculates matches with slightly varying parameters but the main idea is the same. The previous two posts (Kay-Viktor and Cristel) make important points. There has to be an interest in it for the translator too. If the interest is that we buy the expensive tool, we do more volumes for less money for the client and that we also help to make the management of language projects easier for our clients, that's fine. It should not be to the extent that the mere fact that a word is repeated is invoiced at zero. Translation is not just about words, characters, etc., it's about how best to express ideas, notions and turns of phrase in a given context. Volume is just one way of estimating the value of the job and volume is basically a way of considering the time needed to do the job. So be it. What bugs me with many agencies, is that because time can be reduced, the first thing that is usually suggested, on a take-it-or-leave-it basis, is a reduced rate. No thanks. At least my usual base rate then we can start talking about mutual advantages of how the use of a particular tool can be good news for both of us. A crap rate is a crap rate, however you look at it. So let's try to maintain our usual rates and then consider what factors are mutually advantageous!Collapse


Oleksandr Ivanov
Sandra& Kenneth
 

DZiW
Ukraine
English to Russian
+ ...
Technically speaking Nov 21

"Fuzzy match" is a fancy marketing name for scrambled and replaced letters or parts of words. In most cases and language pairs it's just meaningless, especially out of the same or 'techy' context. Why, even 100% matches require careful rereading and [often] restyling, taking up much paid time.

I neither work with agencies, nor give fake "discounts". However, a friend of mine says she can offer a flexible rate, providing it's not below $0.04/word flat--counting all the words. Just ea
... See more
"Fuzzy match" is a fancy marketing name for scrambled and replaced letters or parts of words. In most cases and language pairs it's just meaningless, especially out of the same or 'techy' context. Why, even 100% matches require careful rereading and [often] restyling, taking up much paid time.

I neither work with agencies, nor give fake "discounts". However, a friend of mine says she can offer a flexible rate, providing it's not below $0.04/word flat--counting all the words. Just easily find your absolute bottom)
Collapse


Bernhard Sulzer
 

Bernhard Sulzer  Identity Verified
United States
Local time: 16:41
Member (2006)
English to German
+ ...
Why you should not base your price on fuzzy matches Nov 21

As I said earlier, if a company suggests a scheme and you trust their TM, and even better, if you work with their online CAT tool that you are familiar with, whatever the fee is, it should still reflect what you would charge anyway, according to your own review of the source text. In other words, if you would have charged 15 Cents a word and let's say your total would have been $750.00 but you accept the scheme and do it for $500.00, you did it for $250.00 less (and do you really know what the ... See more
As I said earlier, if a company suggests a scheme and you trust their TM, and even better, if you work with their online CAT tool that you are familiar with, whatever the fee is, it should still reflect what you would charge anyway, according to your own review of the source text. In other words, if you would have charged 15 Cents a word and let's say your total would have been $750.00 but you accept the scheme and do it for $500.00, you did it for $250.00 less (and do you really know what the end client pays for the translation?). It doesn't matter what the machine determined to be a match or fuzzy match, without you, the translator, none of these sentences in the translation could ever been trusted. More than anything else, it's your knowledge, and the fact that only a human can really think about words, that should determine the price/fee - no matter what technical aid you apply to get to that point.


Fuzzy matches are not really matches. An orange often turns out to be a peach or even a piece of asparagus, and you need to throw out the orange. That's mostly the case with sentences that are a little more complicated than "Bill bought a cheap translation today" which in your own new text should be "Bill bought a bad translation today." Although in this particular case, not too much harm is done if you use the suggestion.

When an analysis of source and TM determines a fuzzy match, only one thing is for sure: it is NOT an exact match. It's not even almost a match. At least it is NOT GUARANTEED. What does an 85% (or higher or lower) match really mean? A certain order of words/certain words have been found before in a different (mostly a very different) context (unless you are doing updates of the same text with minimal changes - and still in my areas of expertise that means I need to restructure sentence parts) - and now the new source text has been recognized as somewhat similar to an old source-target match/TM and you get a phrase or sentence suggestion. Then you look at that and find that you have to rearrange, delete and insert other - the correct - words.

Or, you say, wow, that's pretty close to what I need and you might only change a little before using the suggestion. In any case, only you can determine what fits and what doesn't and whether or not the phrase or sentences your CAT tool suggests are correct or can really be used in some way. That goes for the whole text, including the so-called 100% matches. Again, meanings of words depend on context and the same word in different contexts/subject areas can be a completely different word or phrase (throw in idiomatic expressions) in the new target text.

There's nothing wrong with using TMs, but I can't recommend accepting a crude analysis result such as fuzzy matches to determine the price and discount for any translation. Should you determine, after you review a source text that is very similar to what you did before with a TM you trust, that it is an easy task, it's up to you to give a discount, a discount you determine based on your own considerations, not that of a machine. But just because something is easier for you to do doesn't automatically mean it warrants a discount. It's the value of your assessment during your work that the client should pay for -- so your work might be worth much more despite the fact that you used an already existing TM or not.

But don't just accept a scheme because you think you have to do it. It's giving in to the agency; you become a willing cog in their machine. And be very careful about TMs you get anyway. And how do you know that the fuzzy analysis is even correct? And don't trust "fuzzy" schemes that are not even based on TMs, meaning matches purely based on the source text.


[Edited at 2019-11-21 15:36 GMT]
Collapse


 

Samuel Murray  Identity Verified
Netherlands
Local time: 22:41
Member (2006)
English to Afrikaans
+ ...
@Bernhard Nov 21

Bernhard Sulzer wrote:
And don't trust "fuzzy" schemes that are not even based on TMs, meaning matches purely based on the source text.


I don't think I've said this in a while, so allow me to disagree again: if you do give discounts for fuzzy matches, then logically speaking fuzzy matches that are based on TMs should be charged higher than fuzzy matches that are based on the source text itself, because fuzzy matches from TMs are the work of other translators (and you can't know beforehand if you trust them), whereas fuzzy matches from the source text are the work of yourself, whom you can always trust.

Previous generations of CAT tools were just incapable of calculating fuzzy matches based on the source text, which is why translators and clients are "used to" calculating discounts for fuzzy matches based on TMs. Grids that do give discounts for fuzzy matches based on the source text tend to give them equal weight as fuzzy matches based on TMs, which does not make logical sense, but at least the error is in the translator's favour.


 

Bernhard Sulzer  Identity Verified
United States
Local time: 16:41
Member (2006)
English to German
+ ...
Discount for source word fuzzies? Nov 21

Samuel Murray wrote:

Bernhard Sulzer wrote:
And don't trust "fuzzy" schemes that are not even based on TMs, meaning matches purely based on the source text.


I don't think I've said this in a while, so allow me to disagree again: if you do give discounts for fuzzy matches, then logically speaking fuzzy matches that are based on TMs should be charged higher than fuzzy matches that are based on the source text itself, because fuzzy matches from TMs are the work of other translators (and you can't know beforehand if you trust them), whereas fuzzy matches from the source text are the work of yourself, whom you can always trust.

Previous generations of CAT tools were just incapable of calculating fuzzy matches based on the source text, which is why translators and clients are "used to" calculating discounts for fuzzy matches based on TMs. Grids that do give discounts for fuzzy matches based on the source text tend to give them equal weight as fuzzy matches based on TMs, which does not make logical sense, but at least the error is in the translator's favour.


As you say that's your own work. You have to still translate the source text. You're not getting any suggestions.
I wouldn't see a reason for a discount. Are you saying because you/the translation agency detected fuzzy matches in the source text, you'd be willing to give discounts, theoretically even charge less than for TM fuzzies?

[Edited at 2019-11-22 14:08 GMT]


 

Bernhard Sulzer  Identity Verified
United States
Local time: 16:41
Member (2006)
English to German
+ ...
Don't trust machine analyses Nov 25

As becomes clear by reading the thread below, analyses of new texts and old TMs can offer various different results as far as counts for matching words are concerned. Which analysis can you trust? In any case, I don't base my fee on analyses. As I said before, if an outsourcer insists on using the analysis to arrive at a price (after you agreed to the scheme), the total fee should still reflect what you would charge based on your own "human" analysis and the expected quality of the translation.... See more
As becomes clear by reading the thread below, analyses of new texts and old TMs can offer various different results as far as counts for matching words are concerned. Which analysis can you trust? In any case, I don't base my fee on analyses. As I said before, if an outsourcer insists on using the analysis to arrive at a price (after you agreed to the scheme), the total fee should still reflect what you would charge based on your own "human" analysis and the expected quality of the translation.

If the total price is below your threshold for doing the translation, then don't do it. I wouldn't. You're just selling out. Whatever tools you use to help you carry out the translation is just a technical aspect. You are expected to provide an excellent text in the target language. That ultimately depends on your "human" expertise and thinking. It's the expected quality of the translation that should be considered when quoting a price, not a machine counting matches of words, phrases and sentences. My opinion.

https://www.proz.com/forum/cat_tools_technical_help/283468-baffling_difference_between_memoq_and_sdl_trados_studio_word_counts.html

[Edited at 2019-11-25 18:32 GMT]
Collapse


 

Nikolay Novitskiy  Identity Verified
Russian Federation
Local time: 02:41
Member (2018)
English to Russian
+ ...
Why should you? Nov 28

You should not, and there are 3 reasons for this.

1) Agencies don't give fuzzy matches discounts to their direct clients. They charge them full price - for repetitions, 99% matches, whatever! It means that accepting this policy you just give your earned money to your agency.

2) Buying and mastering a CAT-tool is an investment which boosts your productivity. Why should you reject a cash bonus this boost gives you? Investments should bring you profits, not losses.
<
... See more
You should not, and there are 3 reasons for this.

1) Agencies don't give fuzzy matches discounts to their direct clients. They charge them full price - for repetitions, 99% matches, whatever! It means that accepting this policy you just give your earned money to your agency.

2) Buying and mastering a CAT-tool is an investment which boosts your productivity. Why should you reject a cash bonus this boost gives you? Investments should bring you profits, not losses.

3) Not giving fuzzy matches discounts gives you more money. More money is better than less money, isnt't it?



[Edited at 2019-11-28 07:52 GMT]

[Edited at 2019-11-28 07:53 GMT]
Collapse


 

Samuel Murray  Identity Verified
Netherlands
Local time: 22:41
Member (2006)
English to Afrikaans
+ ...
@Bernhard Nov 28

Bernhard Sulzer wrote:
As you say, that's your own work. You have to still translate the source text. You're not getting any suggestions.


You *are* getting suggestions after you translate the first one.

The internal fuzzy discount does not apply to the first segment that you translate. It applies only to segments that get suggestions.

So, imagine you have these four segments, only segment 2 and segment 3 are discounted. Segment 1 is paid at the full rate, because you're not getting any suggestions, and discounts only apply to segments that have suggestions.

#1 The quick brown fox jumps over the lazy dog = entire segment is new, therefore NOT discounted
#2 The quick black fox jumps over the thrifty dog = high fuzzy match against segment #1, discounted
#3 The slow brown fox jumps over the lazy cat = high fuzzy match against segment #1, discounted
#4 The fox and the hound jump over and over = low fuzzy match against segment #1, possibly discounted


Angie Garbarino
 
Pages in topic:   [1 2] >


To report site rules violations or get help, contact a site moderator:


You can also contact site staff by submitting a support request »

Do you give reduced rates for CAT tool generated fuzzy matches?

Advanced search







Déjà Vu X3
Try it, Love it

Find out why Déjà Vu is today the most flexible, customizable and user-friendly tool on the market. See the brand new features in action: *Completely redesigned user interface *Live Preview *Inline spell checking *Inline

More info »
Anycount & Translation Office 3000
Translation Office 3000

Translation Office 3000 is an advanced accounting tool for freelance translators and small agencies. TO3000 easily and seamlessly integrates with the business life of professional freelance translators.

More info »



Forums
  • All of ProZ.com
  • Term search
  • Jobs
  • Forums
  • Multiple search