Friday, March 6, 2009

Towards A More Equitable Pricing Structure: A Translator's Perspective

By Translator X

I have been a professional translator for over 20 years in Canada, one place in this world where translation is regarded as what it really is: transferring messages between different cultures. In Canada, we have two main official languages, English and French.

In this article, I take issue with certain claims made by translation memory (TM) and content management system (CMS) vendors. In addition to the lingo of "fuzzy matches" and "exact matches," there is the publicity claiming, "Don't pay twice to translate the same word," and "Save up to 80% on your production costs." I beg to diverge on that last one. My figures are a bit different. First, one must understand what is the target, and what can be obtained from it.

How is redundancy billed?
What does it imply if it is saved from the translation bill?

Most translation work is paid on a per word basis. This system is far from perfect, but all in all, everyone has become accustomed it. Some word combinations can easily be translated easily, while others are tough. Some parts of text are repeated, while others require endless research to solve ambiguity. In fact, it turns out that pricing based on the concept of word count is a sort of package deal that assumes that some parts will pay a lot, while others will hardly pay at all.

For translators, reusing parts of text that have been previously translated, either in the same text or in previous texts, has always been part of the price billed to customers. However, TM has radically altered the way in which translators do their work. Yet, pricing schemes and cost structures rarely take this into account.

What share of the savings should the customer expect?

The customer should expect his fair share of the savings that new technologies bring, but not those very ambitious figures provided by software vendors. The savings should be based on reality, not at the expense of individual translators.

To determine what is fair, one needs to understand the nature of texts. Some texts are what we call "narrative continuous," requiring careful attention to grammar, syntax, ambiguity, etc. "Discontinuous" texts are much easier to translate for they are usually parts lists, very standardized technical procedures and the like. They are processed on the basis of terminology rather than translation.

Outlined below are three examples that demonstrate how very often translation savings are not value-added by the new technologies.

  1. Course manuals for students and trainers. The trainer's manual contains 100% of the student's text, plus additional explanations and test answers inserted throughout the text. [3 scenarios]
  2. An updated release of a user's guide in which 80% of the words are recurring. However, there is no log of changes in the text. Since the software is a new release, the structure and formatting of the document have been changed. [2 scenarios]
  3. A list of items for a catalog of office supplies. Around 90% of the items are the same as in the previous release of the catalog, but most prices have changed.
Example 1: Course Manuals for Students and Trainers (Scenario 1)

The customer's project manager (PM) is both technically and linguistically qualified to do a good translation herself. She hopes that the student's manual will cost nothing, since the text is contained in the trainer's manual. The company has recently invested in a TM package.

  1. The PM outsources the text and requires the translator to create a TM for the user's manual. The translator uses the same software as the company, so she analyses the text and submits her bid. Since there are almost no internal matches identified by the tool, she knows that she will gain nothing in the short term while translating the trainer's manual. In addition, she will spend time to enter sentences into TM. NOTE: The trainer's manual contains 30% more text than the student's manual.
  2. During the translation of the trainer's guide, the translator retrieves around 10% of the text, i.e., sentences that are similar to others in the same text, and reuses them. Result: the translator saves time on around 10% of the sentences, but she loses 10% because of the additional time required to use TM. In addition, she must now invest a certain amount of time to modify these retrieved sentences, usually around 40% of the time required to translate the same sentences without TM. Therefore, the translator really ends up spending 10% to save 6%, giving away 4% to the customer.
  3. The translator now returns this text, along with its new TM, to the customer. The PM then sends the student's manual with the newly created TM to the same translator. The latter knows that she must carefully proofread the text to ensure that everything is correct, and that there is nothing abnormal in the memory. This type of proofreading requires approximately 20% to 30% of the time normally needed for a full translation.
  4. Time must be added to review the quality of the TM. History shows that resolving technical problems and managing TMs take around 10% of the total translation time. Obviously, when a translator has little or no experience with TM, the review time is longer.
  5. Total time spent at this point in the process by the independent professional translator: 20% (proofreading) + 10% (fixing problems and managing TMs) = 30%.
  6. In this case, the formatting of the two manuals is not identical, so the PM must spend an additional 10% of her time proofreading. This brings the total time spent to retrieve text that contains exact matches up to 40%.

FINAL RESULT: In this case, the customer will almost certainly be frustrated, since her expectation was a 100% savings on the user's manual. However, her savings has shrunk to 60%, even though the translator has given up 4%, too, rather than paying for her software. There is a "negative added value" for the translator. The overall cost of the process is 44%.

An archaic invoice for this project would look as follows:

Archaic invoice

A more realistic invoice for this project would look as follows:

Realistic invoice example

There is only a difference of $785.00 between the two types of invoices: a pittance for most companies, small or large, yet very important to an independent professional translator.

Example 1: Course Manuals for Students and Trainers (Scenario 2)

The customer expects a 100% saving since he has just bought a new software program that stores phrases in terminological databases and allows him to modify text alignment at will.

  1. The PM sends the text to be translated at the normal rate.
  2. When the translation returns, he aligns it very quickly.
  3. He then validates the alignment, which should take around 3% of the normal translation time. Since he needs to align 130% of the student's manual, 3% becomes 4%. He must also create entries in the database, which takes around 10 % of the translation time.
  4. The PM must invest approximately 10% of his time to maintain the memories.
  5. He applies the translation memory to the text of the student's guide.
  6. He must proofread the text to make sure that it does not contain any false translations. This requires at least 20% of translation time.
  7. Partial totals for time spent: 4% for alignment, 10% for entries, 20% for proofreading, 10% for TM management, which add up to 44%.

The customer expected a 100% saving, which has now be reduced to only 56%. What does it take to save the 70% and more promised by TM vendors?

Example 1: Course Manuals for Students and Trainers (Scenario 3)

The customer expects a saving of 100% on the student's guide, even though his company has not invested in TM software. Instead, he hires a translator, who is the proud owner of this state-of-the-art new software, and agrees to bill only 10% on the 100% matches, since it is the machine that will do all of the work. The translator assumes that 10% for a few minutes of work is good enough.

  1. The translator translates the manual and enters each and every sentence in the memory. Doing so lengthens his working time by 10%. Since his TM software does not recognize fuzzy matches, he receives no compensation for those. Of course, managing his translation memories also takes a toll of 10% on his time. However, he does this willingly, assuming that his investment will pay off someday.
  2. When the translator receives the second document, he is able to retrieve most of the text. However, he must now spend approximately 10% of his translation time in formatting.
  3. The translator must then verify that every translation is in the correct place, which takes around 10% of the translation time, and then translate the few phrases and sentences that were not retrieved by the software (e.g., the insertion of a paragraph in the middle of a sentence, etc.).
  4. TOTAL: the translator has now spent an additional 40% of the translation time, but has only billed 10%. This represents a mortgage on holidays for a few years if he continues to work this way!

FINAL RESULT: The customer is happy, having almost realized his expected 100% savings. He also believes that he has been generous with the translator. He will tell everybody that implementation of such software can generate savings of up to 90%.

Of course, the reality is that a large part of the savings has come directly out of the translator's pocket. The added value has been largely overstated and constituted a transfer of expenses from the customer to the supplier.

Example 2: Updated Release of a User's Guide (Scenario 1)

The customer expects savings of 80%, since the TM software recognizes 80% of the words.

  1. The PM sends the project out to bid and finds a translator willing to accept an 80% reduction in his normal rate. However, he requires the customer's TM.
  2. The PM provides the TM, but this product stores the sentences and the terminology in the same repository, handling complete sentences and phrases in the same manner. The translator, on the other hand, has invested in an industry-standard product that distinguishes between terminology, phrases and sentences. Therefore, a large volume of the words recognized by the software of the customer is of no use at all to the translator.
  3. The translator imports the memory, only to discover that a large percentage of words recognized by the customer's software are not recognized by his package. At this point, only 36%, instead of 80%, of the text is recognized.
  4. The translator has just lost a full hour playing with the content and the conversion. He calls the PM and explains the situation. The PM does not need his software right now, so he sends the dongle and the software to the translator, who then invests a few more hours to familiarize himself with the product.
  5. This time, the results are 80%. However, since the software does not always insert the correct gender, i.e. feminine or masculine, in phrases that make up translation units, the translator realizes that he must spend as much time using TM as it would have taken to do the full translation. Therefore, his real savings are approximately 25%, and when he finishes the job, he is totally exhausted.
  6. TOTAL: the translator bills 20% for 75% of his effort, plus he does not charge for the hours required to convert the TM, learn the new product, etc. On top of this, if he doesn't win more contracts in which he can use the same software, he will lose the knowledge that he has gained.

RESULT: The customer is satisfied, but the translator is short-changed, similar to the crow in a Lafontaine tale.

Example 2: Updated Release of a User's Guide (Scenario 2)

The customer has been told that 80% of the words are the same as in the last version. He believes the text comparison feature in his software and therefore expects an 80% savings.

  1. His PM has processed the previous text with an industry-standard TM product that handles sentences and phrases separately. Around 30% of the sentences are exact matches, while 50% are fuzzy matches.
  2. When sent out for bid, translators respond with a cost of 20% for exact matches and 75% for fuzzy matches. After some negotiation, it is agreed that it will be 10% for exact matches and 50% for fuzzy matches. Translators often consider that reviewing context takes roughly 10% of the translation time, while fuzzy matches require between 30% and 75% (due to gender, verb tenses and the like).
  3. The PM sends the text to the freelance translator who processes the text. Of course, he has to work a lot on the formatting, which varies from the last version, because customers always want their texts to look "better and improved." However, the translator is satisfied overall.
  4. The translator then bills 50% of words at 50% for fuzzy matches, that is 25% + 10% of the 30% (exact matches), that is 3%, plus the remaining full 20% which is brand new text.

RESULT: Rather than paying 20% of the total bill, the customer will pay a more equitable 58%. The customer feels a bit of frustration, but in fact, he just paid the right price.

Example 3: Updated Office Supply Catalog

Here is a glorious exception where the customer can save 70%, 80% or more, especially if there is an item ID that clearly identifies the item. In this case, the translation probably doesn't even need to be proofread since only the prices have changed. But...do we really need TM for such projects? Wouldn't a database containing the description in the appropriate languages, a place for a conversion factor for the various currencies and a price field be a better solution?

quote

A translator is a language specialist who understands creative uses of language and that sentences can be ambiguous due to neologisms, false friends, etc. He is familiar with and knows how to handle the many other challenges in correctly translating the reality of one culture into another. A translator's compensation should reflect this. A more equitable bill for some of the work described in the examples above would look something like this:

Equitable invoice example

Again, however, the normal bill usually looks as follows:

Normal invoice example

This is a difference of $1,195… equal to 87% of what the translator was paid! In other words, the translator has been paid a little bit more than half of what he should be! However, everyone seems to find it normal that the parts that are particularly lucrative should be deducted from the translator's wages.

Let me be clear, I don't want to say that we should reject TM software. It is clear that these software are here to stay, and that they are very often productivity enhancers. However, the added value must be shared with translators, instead of becoming a pretext to minimize the importance of good work.

As a matter of fact, many of the words counted as "saved" by TM tools are, in fact, as easy and fast (sometimes faster) to handle through a good global replacement feature. E.g., when there is a sentence in English instructing a user to do XXX and then to "press Enter," it can easily and quickly be translated into French with the same text every time ["et appuyer sur la touche Entre (Enter)"].

Preprocessing text with Content Management Systems

This is worst than bad! CMS vendors claim, "You only have to pay for the new content." In real life, that means that the translator receives a few sentences, or a paragraph, and must read many paragraphs to understand the context each time.

In addition, language aspects such as gender, formality, etc. are not really handled by these systems that pretend to manage content. In fact, these programs only manage changes, and usually do it right only in English. In some instances, a gender change may imply in a language such as French or Spanish that all other sentences with gender-specific content must also be reviewed and changed. Also, depending on who is speaking, the social conventions may change completely the way things are said in another language. For example, in Canada, if I don't know the person to whom I am speaking, I use the pronoun "vous", while I use "tu" if I am speaking to a friend. To be polite in some countries may mean doing business or not (as in Japan).

Elements that should be considered for an equitable evaluation Text Type: Continuous or discontinuous

Usually, a translator earns more for continuous text than for discontinuous text.

Sentence Length

As mentioned earlier, headers, footers and titles in general are easy to translate. The fact that a program can match them to their respective translations automatically does not necessarily speed up translation. Even worse is the fact that some tools force translators to confirm each occurrence of a phrase, actually making the job much, much longer. Yet, customers demand rebates for the imagined savings, at the translator's expense. Shame!

Customer Loyalty

If customers always sent their texts to the same translator, at least the investment in TM would make financial sense for the translator. However, reality is a bit different. Large corporations usually have more than one translation provider, while some prefer to handle their own TMs.

Far too often, the following happens: a translator translates a text for the first time, inserting all of the translation into a TM. However, when the corporation updates the product, it sends the first TM to a different translator who bid half a cent less than the first translator. Translators should charge to use an empty translation memory, since next time, the company will demand that the price of the translation be lower.

Why not be equitable?

Some people say that translation is the main cost of some products or services. Then, if it is the main cost, it is probably also the main revenue multiplier.

Suppose that a company produces something in English, and then translates into languages that have ten times less people in terms of customers. With each new translation, there is a new market, new opportunities. Butif you consider it only as a cost, then all you want is to lower that cost.

Let's look at two examples. A home appliance manufacturer in Canada is able to produce a microwave at an excellent price/quality ratio. The company is proud of its product and decides to include a book of recipes with lots of illustrations on high gloss paper. Making such a book really costs a lot, so they decide to "save" on translation.

The book has approximately 150 pages with around 10,000 words. The company provides each customer with two books, one in French and one in English. A very well-paid translator charges 25 cents a word, so the total translation bill is $2,500.

At the risk of damaging its image with a poor translation, the company might be able to save at most $2,000 by sending the text to a less-qualified translator. People who view translation only as a cost factor probably think this is great. However, customers reading a poorly translated book will probably think that anyone making such a decision should be fired.

Hardly anyone will contest the claim that translation is one of the main costs in the software industry. I say that this is perfectly false! Let us suppose that a good programmer writes an award-winning computer-based training program to teach Java. He sells 1,000,000 copies in English. When he and his editor consider the second release, they want to increase sales by improving the software. However, the real revenue-generating opportunity lies in allowing people who do not speak English access to the software by translating it.

Each new language implies a cost, yet generates a certain amount of income… not bad! Eventually, if a good translation always generates 100 or 1000 times its cost, all one can hope is that this type of cost will reach millions.

quote

Translation should not be considered only as an expense, but rather as an essential part of development. It should receive as much funding as any other essential activity. A poor translation causes as much trouble as poor writing. And poor writing is like delivering a program made at lower costs, but full of bugs. It is a guaranteed path to bankruptcy.

Customers do not buy services that are under a certain level of quality. The translation delivered in high tech industries very often is just on that tiny line.

However, trying to reduce translation costs is normal when it is done in the same manner as trying to reduce other costs. Would you change your programmers for a matter of 2% in their salaries? Well, lots of companies do just that with their translation providers, without having the faintest idea as to whether the lowest bidder can actually deliver reasonable-quality translation at a fair price.


- Translator X from Canada


Reprinted by permission from the Globalization Insider,
25 February 2003, Volume XII, Issue 1.4.

Copyright the Localization Industry Standards Association
(Globalization Insider: www.localization.org, LISA:
www.lisa.org)
and S.M.P. Marketing Sarl (SMP) 2004


Corporate Blog of Elite - Professional Translation Services serving ASEAN & East Asia

No comments: