Monday, August 31, 2020

Lead and Gold: Challenging the Premium Translation Market Claims

This is a guest post that is a detailed response by Luigi Muzii and is a clear rebuttal of my previous post on the Premium Translation market. While Luigi agrees that the translation market is made up of many smaller segments, he does not see enough evidence of a clearly discernible premium market. He admits that there are premium customers who are willing to pay higher prices he questions the long-term viability of a market that ALWAYS pays higher prices for translation when reasonable lower-cost alternatives are available.

Since Luigi took the trouble to document his criticisms into a complete post, I felt it deserved to be an independent post that furthers the dialogue on the subject, by presenting another opinion and hopefully attracts further conversation from those who see the issues most clearly. 

We live today in a world where respectful dialogue is woefully inadequate, especially in national politics. While some of the comments might be seen as scathing, or overly negative, I find that his statement of his views (which I do not necessarily agree with) meets my standards for respectful professional discussion. Disagreements need to be forceful at times, and for this blog, I only expect that they do not become petty and personally disrespectful.

I also gathered that a primary motivation for his comments was his concern that premium market discussions would discourage both young translators who are just starting out and old-timers who might feel discouraged by choices they have made years ago.

There are, unfortunately, clear value associations in the very contrast created by the words "bulk" versus "premium".   The value creation aspects are more opaque in this dichotomy, as I have seen large MT projects can create more value (in monetary terms) than the most expert translators can around a single project, and that there is a place for the whole spectrum of translation production possibilities that exist in the world today.

Thus, I maintain that for those with demonstrated competence and true subject matter expertise, a premium market does exist. This means that it is not just higher priced work, but also that the client to translator engagement is much more active, collaborative, and consultative. 


Do you know the way to El Dorado?
I’ve been away so long. I may go wrong and lose my way
Do you know the way to El Dorado?
I’m going back to find some peace of mind in El Dorado.

Do you know the way to El Dorado?

The fabled El Dorado of translation, the “premium market”, is to my mind much like a losing stream. Unfortunately for the many believers who still exist, there is no El Dorado. It is a legend, nothing more than a popular topic in blogs and at conferences. There is no proof of its existence because none of its fierce advocates have ever produced any, so there is no map or instructions to get there and no one is able or willing to provide any.

A few years ago, the news circulated frantically, without any confirmation from the said client, that Le Manoir de la Régate, a gastronomic restaurant in Nantes, had paid a notorious advocate of the mythical “premium market” € 800,00 for the translation of a “postcard”.

Lately, another example has been circulating, supposedly to put an end to any controversy about the existence of the so-called “premium market”.

Now, anyone who has worked in translation for a while knows that, for the job to be sustainable, a translation should not outperform the original (especially if this is already good), and that a professional translator producing superior texts from crap is not the reincarnation of Cagliostro, but a fool.

It may, of course, happen that the translator’s writing skills are such, that they easily outperform the original from the start, but in that case, the client is a fool, willing to pay more for a derivative work than for authoring. Provided, of course, that the translator is not so foolish or bashful as to accept to work for a pittance while being aware of their skills, the poor quality of the source text, and the intended use of the translation.

No top defense counsel is going to represent a pickpocket in court, just as no pickpocket is likely to have the means to hire a top defense counsel. Maybe because no defense counsel becomes the best in court overnight.

Similarly, a wicked lawyer may counsel a pickpocket and bill a fortune, just as a translator may charge an outrageous fee for translating a postcard. But, if this is true, this is rather a matter of professional ethics. Incidentally, counseling may not save the pickpocket, and the postcard may remain an isolated marketing attempt (the restaurant’s website above, for example, is still in French only).

Anyway, a customer looking for a translation to outperform the original has much more serious problems than finding a top (premium) translator. On the other hand, it is highly unlikely that the author of a legislative text, a patent application, an economic or financial report, or legal advice, would accept any comments, remarks, or writing directions from a translator, however capable. Unless, of course, the translator is also an equally capable lawyer, engineer, scientist, or economist.

Today, all published content is indeed global, and users can easily have it machine translated if they do not master the language(s) in which it needs to be available. On the other hand, in an ideal world, content would be designed and authored with translation in mind, the software would be perfectly internationalized and multimedia ready for multilingual subtitling. Only in an ideal world.

In the real world, translation is often unappreciated, most often seen as a necessary evil, and, as such, left for the end of the content production cycle, to be done cost-efficiently. So, a client who pays more than four times the market average for the translation of a standard text is a fool, all the more so if that translation is not worth the price differential.


The commoditization that has been affecting translation for some years now has the same effects that all commodities endure.

For example, deforestation and climate change are progressively and significantly reducing coffee crops and worsening the conditions of extreme poverty in which farmers are already living, although the coffee trade generates revenues of over US$ 100 billion per year, leading many farmers to leave. The increasing sales of fine varieties like kopi luwak will be of no help.

Just as the demand for coffee, the demand for translation is widespread and increasing; the global marketplace is crowded with price-sensitive buyers and there is little point now to bring about the issue of information asymmetry and signaling, which has been regularly dismissed for years as irrelevant.

Like for coffee and kopi luwak, there is no translation “premium market”: There may be a few “premium customers” that can, at most, and with much goodwill (from my side), represent a segment. And you should be accurate with your lexicon, especially if you are a linguist and you work for the banking and financial industry.

Another example could be a bespoke, hand-sewn three-piece men's suit from a luxury tailor shop in Savile Row, which does not necessarily make any of the tailors in the shop, and maybe not even the owner, a wealthy guy. Likewise, Brioni can provide a very demanding customer with a sartorial ready-to-wear suit while Marinella can still sell its famous handmade custom ties at any of its stores around the world.

In short, the bulk-premium dichotomy is not only simplistic, Manichaean, and capricious, it is mala fide and reflects ignorance and a profound lack of respect for all those who make a more than respectful living working with “bulk-market” customers.

Finally, bringing on the example of US defense contracts to support the existence of the premium translation market is pointless. In this case, translation jobs go to some professional services contracting company, which is part of some large conglomerate, as in the case of GLS and other regular military contractors.

If “premium” simply means that the word rate is higher, possibly a few isolated, individual translators and, most probably, some sub-contracting LSPs may earn better money than average, but this definitely does not make US defense contracts any kind of  “premium market”.

In contrast, it is possible that other major institutional customers, e.g. the EU, push hard on translation prices when procuring for translations to compensate for the untenable stipends of their in-house translators, thus further contributing to commoditizing translation.

For all these reasons, should a “premium market” exist it would most probably be “fiercely guarded and (often) shrouded in secrecy to prevent additional competition”, and this would make it even harder to find and access it.

Venture Capital and Private Equity

The advocates of the so-called “premium market” have been using the interest that some private equity firms and, to a much lesser extent, a few venture-capital funds have recently shown in the translation industry to restate their arguments.

Others maintain that the Big Four accounting firms regularly approach translation boutique firms to explore potential opportunities.

Leaving aside for a moment the numerous and repeated criticisms made over the years to those Big Four firms, who figure prominently in corporate collusion allegations, and the suspicion of money laundering behind some PE transactions, recently some mid-to-high gross margin LSPs have caught the attention of PE firms because higher gross margins usually mean higher cash-flow margin to investors.

However, VC and PE firms are typically interested in short-term growth, possibly via M&A, but top growth rates are made up of, among other things, the pace of hiring, the complexity of services delivered, and the capital intensity of expansion, with respect to the market size, maturity, and competition, all things that are very hard to find in SME LSPs.

Indeed, the translation industry’s CAGR is often claimed to be steadily above the World’s GDP growth levels, following the explosion of content volume and expanded global trade. LSPs in the gaming and life sciences niches might in fact grow even faster, but PE firms usually expect to make a three-time cash-on-cash return or more on a five-years typical investment time horizon, and the expected IRR is 20-25 percent minimum. Objectively, these results are hard to achieve by investing in SMEs LSPs.


As a matter of fact, most of the people hailing the fabled “premium market” live and work outside it, and most probably do not know the path to it or the key to access it.

This, however, does not hold them from ranting against those who they believe, or maybe they just assume are the culprits, for the decadence of translation and the translation profession.

Indeed, PE firms targeting their investments in the translation industry cut freelancers out of the equation, and this gets things back to square one with the so-called “premium market”. However, the recent RWS’s takeover of SDL shows that domain expertise builds value, and this can be found not only in high-profile professionals. In this respect, back in 1993, AITI (the Italian Association of the Translation Industry) invited RWS to an international conference on translation quality assessment and report about being the first translation company certified to BS 5750, the predecessor of ISO 9002. The conference proceedings are available for download.

The fundamentalist fever against ‘corruption’ of translation also affects Academia. Actually, it started there and has always been spreading from there. Specifically, translation students are not taught to deal with the intricacies of the real market because most teachers have never translated a line in their lives. Preserving the status quo of the old curriculum, with the associated models, is a reason for the continuing survival of these attitudes.

So, a recent paper on the sustainability of the current models in the translation industry comes as no surprise, even if it comes from an otherwise seemingly innovative institution like DCU.

Sustainability is not a new topic and has often been associated with quality. Unfortunately, a major flaw in Joss Moorkens’s paper can be found right away. In actual fact, the working situation described in his paper is relatively recent and has not been “live for decades now”.

Moorkens’s paper presents the typical traits of confirmation bias, the same that can be found in most arguments from the advocates of the so-called “premium market”. Confirmation biases contribute to overconfidence in personal beliefs and can maintain or strengthen beliefs in the face of contrary evidence.

In fact, the paper does not substantiate or provide any actual evidence of the alleged widespread application to translation-work of Taylorism, which Moorkens also apparently confuses with Fordism. It may happen. However, the application of documented information as related to processes and workflows is the basis for quality management standards and even applies to the many, (somewhat poor), translation quality assurance standards. Even the existing translation quality assessment models, which come mostly from academics, are based–more or less knowingly–on “scientific management”. Finally, the standardization of production has allowed consumers to buy cars, and soldiers to be safer when using ammunition. See my further comments on the freely available A Contrarian’s View on Translation Standards.

Of course, the remuneration of translators is a critical issue, but M&A and PE firms have little to do with it. It is true, though, that remuneration has been decreasing for the last three decades due to the many technological innovations that have been introduced almost entirely from the outside into the translation industry. This is another interesting topic that academics seemingly prefer to ignore in their studies, possibly because it does not relate to their field of study and it doesn’t help to safeguard the status quo.

On the other hand, the issue of wage reduction has been at the heart of business associations’ propaganda for years, because it is easy, costs nothing, and avoids entrepreneurs from having to open their wallets to modernize production structures and processes.

Surprise, things are changing even in the temples of laissez-faire. For example, in The Economics of Belonging, the Financial Times’s European economics commentator Martin Sandbu argues that compressing labor costs reduces productivity, that higher labor costs push companies to move to more advanced production models, to make more investments, and that, as long as people are paid little, companies are settling on low value-added production. Likewise, in The Limits of the Market, Paul De Grauwe argues that if employers like to keep labor costs low, they will only succeed if they work to curb technological progress.

Wait a minute! Has this nothing to do with the Gresham’s Law? Maybe the advocates of the fabled “premium market” are short of updates in labor economics. And not just that.

However, if RWS’s takeover of SDL is good news for the translation industry as many seem to think, there will be less and less space for a “premium market”. The specialization required to become a skillful and well-paid translator with a lasting position at the high end of the translation market involves substantial commitment, time, and investment. Skills do not grow on trees or accumulate overnight; building a network of relations is toilsome; learning to exploit it may require some major changes in character, and no one can guarantee stable high prices and job satisfaction. Time is crucial unless your parents or your spouse can indefinitely pay for your continuing education while you are stuck in the “bulk market.” And possibly find you a permanent position in some financial or military institution.

Do advocates of the “premium market” ever tell this to their pupils willing to hear their fairy tales and feed their wishful thinking?

To be honest, in his paper, at least Joss Moorkens admits that “the hollowing out of the middle section of the market may make it more difficult to climb to the high end”.


Luigi Muzii's profile photo

Luigi Muzii has been in the "translation business" since 1982 and has been a business consultant since 2002, in the translation and localization industry through his firm. He focuses on helping customers choose and implement best-suited technologies and redesign their business processes for the greatest effectiveness of translation and localization-related work.

This link provides access to his other blog posts.


Post Script Addendum

I dug into the €800,00 postcard translation example given above and found out some facts on this that I think are worth sharing to give it an accurate context.

The example comes from Chris Durban who uses it as a teaching aid (shown above) in a classroom setting to explore and show different aspects of value-added translation work. It was done for an FDI client of hers whose primary focus was on texts that were intended to attract foreign investment into a region of France to increase employment. This client was sending a team to Davos, and the postcard was part of a press kit pointing to the quality of life characteristics of the region and was intended to attract investment, and expatriates to consider the region for new business initiatives. This was to be used at the WEF conference, where it would be compared to other premium marketing and communication messaging.

"My point with this exercise is generally to introduce (and raise awareness of) value pricing, expertise, context (purpose of the text, client's communication goal), and time factors."  

It required the contribution of a specialist cookbook translator working together with Chris working within tight deadlines to make it work with the quality-of-life theme they were trying to promote in the press kit. The point of the example is to show how value is added when one looks beyond the words that need to be translated and focus on the broader intent of the communication, which typically requires more elaborate and knowledgeable integration procedures.   

Thursday, August 20, 2020

The Premium Translation Market: Hiding In Plain Sight

 It is challenging to get a credible and widely accepted estimate of the size of the "business translation" market. The researchers who generate the market forecasts focus heavily on publicly available, or voluntary survey data provided by Language Service Providers (LSPs) in the most visible part of the translation market, yet they seem to come up with somewhat different market sizing estimates. One is often left wondering what they are counting, and where and how they are getting the supporting data. The variations in the market estimates are quite different, as we see below. 

  • The Slator 2020 Language Industry Market Report provides a view of the global language services and technology industry, which, according to Slator, grew to a USD 24.2 billion market in 2019.
  • According to The Nimdzi 100 annual report, the language services industry will reach USD 53.6 billion in 2019 and is projected to reach USD 70 billion by 2023.
  • The market for outsourced language services and supporting technology grew 6.62% to USD 49.60 billion from 2018 to 2019, according to market research firm CSA Research.
Differences of this magnitude suggest that not only is the research methodology varied and quite possibly flawed, but also that reliable data is hard to find. The problem begins with differing market definitions, which, of course, leads to consistency problems across the research estimates. The extreme fragmentation in translation-focused businesses, and "hidden or less visible" market segments make the sizing task especially challenging. While the market estimates shown above are all valid to some extent, one wonders if they capture the reality of the broader translation market? This Nimdzi webinar commentary describes the most widely accepted views on the translation (localization) market, which we can see is very LSP and localization-centric.

The industry is highly fragmented, and there are very low barriers to entry for "basic translation service" work. Most will agree that translation work does not have the same cachet as jobs done by accountants, nurses, lawyers, teachers, and many other professions in the public eye. Because the need for translation is so widespread, barely bilingual amateurs who work part-time can pretend to have professional status and target price-sensitive buyers. Information asymmetry around buyer perceptions of quality, a solitary freelance work-from-home culture, and a lack of clear professional accreditation signals add to the problem. This phenomenon undermines truly professional workers, but it is an unfortunate fact of the translation business today. The presence of these factors does not mean that a professional and respected cadre does not exist but perhaps makes it less visible even though it is very much present.

Are we missing something in the market estimation exercise? One of the most popular articles on this blog over the last ten years is: The Translation Market – Is it Really Understood? a post that was written by Kevin Hendzel, a long-term participant and service provider in the "premium translation market." He unequivocally challenges the conventional market estimates, which barely acknowledge the existence of a premium sector.

And there are government contract announcements that suggest that he is probably right. For example, in a statement from March 2017, the US DoD announced the winners of a super-sized, 10-year language services contract. The USD 9.86 billion Army contract will be divided among nine companies (unknown names in the localization market) and will run through 2027. The need for foreign language support is greater than the Army's ability to resource with military-trained linguists. As a result, contract linguists are needed to work to support a wide variety of military activities ranging from international conferences and criminal legal proceedings to real-world military operations, social data analysis, and intelligence functions. The objective of these services is to save soldiers' lives, and these translation services are vital to successful combat and military intelligence operations. These companies do not show up in traditional localization market research. A description of one of the awardees' language services capabilities (copied from their website) is shown below and describes the premium (Level 2 and 3) translation activities.

And they are not alone: Here is another government market player showing a more complete client list that indicates the breadth of this specialized market need. If we consider that DLITE is a single large contract that is ~30% of the Slator market estimate and that there are many agencies with similar (though smaller) mission requirements spread across the government, we have a sense of what the market estimates may be missing. Add the invisible, classified, clandestine community translation work to the DoD work, and then consider that these IC needs are common to all G7 governments around the world. We can then see that we have a sizable invisible market just with national security and the intelligence community sectors. Many in the localization industry forget that DARPA (Defense Advanced Research Projects Agency) was a primary research funding force in the development of data-driven SMT and STS translation. National security, defense, and global law enforcement customers also interestingly require capabilities at both ends of the translation spectrum, i.e., raw MT for information triage, and specialized SME translators with operational expertise in military and intelligence analysis processes.

There is evidence to suggest that there may actually be several translation markets, and the underlying assumption by the market researchers about a single market may be the reason they are all perhaps only presenting a partial and incomplete picture. Chris Durban, who was instrumental and helpful in my research for this post, characterizes the overall translation market as consisting of three different markets. The translation market can be presented as a continuum, where some overlap may be possible: Bulk > Added Value > Premium.

Kevin describes, "[….] a very long continuum [in the overall market] that encompasses all market segments, with raw, free MT at one end, and $25,000 tagline translations of 3 words [or low word volume high-impact communications] at the other." 

Many (especially those firmly entrenched in the bulk localization translation market) immediately say that premium simply means that the word rate is higher. However, we all understand that a higher price without discernible higher value delivered is generally an unsustainable business proposition. This post attempts to delineate the differences more clearly, as I have observed that the premium market is often dismissed as a myth. I am not saying that premium work is always more valuable than bulk translation, and the "bulk versus premium" phrasing can often be problematic. Both kinds of translation work are necessary and essential for any enterprise with critical global initiatives. Yet while both are necessary, we should also acknowledge that expertise, knowledge, compensation, and other attributes are different.

As Kevin states: "It's far more accurate to characterize the [translation work] continuum in terms of gradual and consistent gradations of shade rather than in terms of clear differentiating boundary lines. The "premium versus bulk" dichotomy is a form of shorthand only. That also applies to price and quality, since the correlation between the two is not always linear. The premium sector includes commercial segments that are fiercely guarded and (often) shrouded in secrecy to prevent additional competition." 

While the localization community continues to overlook the premium segment, potential investors seem much more interested. I gathered from my research that Private Equity and Big Four accounting firms regularly approach boutique premium firms to explore potential opportunities.

What is the premium translation market?

A premium translation market segment is where the impact/cost of lackluster, literal, or inadequate translation is dramatically higher than the initial higher price paid to produce the translation product.

"Markets where the cost of failure is dramatically higher than the cost of performance."   Kevin Hendzel

An admittedly dramatic example is useful to help illustrate this. The NSA has provided documentation on how translators unfamiliar with Japanese diplomat-speak may have inadvertently caused the US to proceed with the bombing of Hiroshima. 

Mokusatsu (黙殺) is a Japanese word composed of two kanji characters: 黙 (moku "silence") and 殺 (satsu "killing"). It is one of the terms frequently cited to argue that problems encountered by Japanese in the sphere of international politics arise from misunderstandings or mistranslations of their language. The Japanese verb for withholding comment is mokusatsu(suru), which could be better understood culturally as, "We'll wait in silence until we can speak with wisdom."

The meaning of mokusatsu can either be "ignore/not pay attention to," or "refrain from any comment," depending on the specific circumstance and context of the usage.

The word Mokusatsu was used in a Japanese government response to the demand that Japan surrender unconditionally to Allied forces. Unfortunately, international news agencies saw fit to tell the world, that in the eyes of the Japanese government, the ultimatum was "not worthy of comment." The response was thus misunderstood to mean that Japan had rejected the terms. This misunderstanding contributed to Truman's decision to carry out the atomic bombings of Hiroshima and Nagasaki. We know today that around 150,000 people died as a result of this unfortunate action.

In the global business context, similar misunderstandings around critical terms can also happen around the reporting of financial results in international markets. The operation and use of complex products like naval ships, jet bombers, nuclear reactors, large-scale construction, and medical equipment also require expertise and experience with the subject matter at hand as well as linguistic issues.

Characteristics of Translation Market Segments

The following characterizations provide some useful clarification on some fundamental differences in these markets.

Bulk Localization: A key objective of most bulk localization projects is cost and production efficiency. Fast-flowing corporate web and product description-related content needs to be rapidly available in a large number of languages. Thus, high volumes of "good enough" quality are often a characteristic of translation here.

Value-Added: A key objective here is the efficient handling of more complex source data that requires a more sophisticated technological, domain issue-focused, or culturally adjusted production process, e.g., subtitling, multimedia data, and culturally appropriate advertising.

Premium: A key objective here is engaged, proven content expertise (SME) in addition to linguistic competence to ensure effective communication focused on high-impact content. Subject matter experts do the translation in deep engagement with the content producers. It has a much stronger focus on the communication impact of the target translations. Translators here tend to be deeply specialized in a specific domain rather than generalists who jump from subject to subject. I would characterize most of what is called "transcreation" as a value-added translation, NOT a premium translation. While financial or SME-driven transcreation is possible in the premium market, generic transcreation is probably closer to the bulk market or value-added sector at best.

I have found the writings and presentations of David Jemielity, Head of Translations at Banque Cantonale Vaudoise (BCV), particularly useful in understanding the work of translators in the premium sector. He has characterized the business translation market as a large and growing economic sector that "punches below its weight" in terms of visibility.

 Here is a recent presentation where David provides much clarification on the characteristics of the premium market and his observations on value-added translation in general. This clip provides some specific samples of the finesse and expertise that is expected in the premium sector.

Most of the dismissive comments on the premium market focus on just two aspects of the market reality, price, and specialization. Many bulk market critics claim that they, too, have higher price segments, and have domain specialization when needed for regulated markets like Pharma and Finance. However, as one looks more closely at the bulk vs. premium approaches, one sees that the differences are indeed more substantial and nuanced. A fundamental difference that I noticed in my interactions with several premium market service providers is the increased focus on communication rather than translation, and the degree and level of engagement they have with their customers.

Relationship to Content Creation

Most bulk localization work tends to happen downstream, with translators having little or no say, or interaction with the content creation process. Source text is "given from somewhere" and is expected to be converted to many target languages quickly and cost-effectively. And thus, translations tend to be word-for-word and often not fit for purpose due to expediency and lack of any client contact.

Premium translation is a more collaborative, consultative, and engaged process with the content creators to ensure that content is optimally translated, which often means "no excuses effective communication" in the target text. This increased engagement is necessary to rise from "Is this a good translation?" to "Is this effective target-language communication?" 

Effective communication outcomes have a higher value for all stakeholders involved than "good translation." This means that ensuring the communication intent is preserved has a higher value than ensuring all the source words are translated. In many cases, as David's experience at BCV shows, this can even mean regular interaction with the CEO and CFO around critical financial market communications. Facetime with key executive stakeholders is a characteristic of the premium segment, which also means that translators can influence the way new and future content is created once they have established themselves as trusted service providers.

Lisa Rüth, describes the buyer and seller relationship as a partnership in the premium market. "The translator is not a mere supplier, and both parties see translation as a service, not a commodity. Premium-market translators usually provide excellent service. They will often go the extra mile – and be paid for it as a partner can be expected to be."

Translator Specialization

In the bulk localization market, large LSPs often source, barely vetted translators out from "vendor" databases that may contain 150,000+ translator resumes. There is much room for problems in such an approach. Long-term focus on a limited subject domain is a necessary condition for building credible expertise, and most of these databases are focused on cost and turnaround efficiency attributes. Proven expertise capabilities are not easily embedded into these databases.

Translators in the premium market are expected to have deep subject matter expertise. SME means verifiable field-specific knowledge. Specialization means that work is limited to a specific and narrow subject domain continuously, i.e., five days a week for years, not three days a week for a month. Subject matter expertise can only be built by a sustained and long-term focus in a specialized field.

As David explains: " A [financial] translator who is truly specialized in the field will give the CEO tons of different ways to talk about this uncomfortable [financial] thing that the CEO doesn't really want to talk about (because it might send the share-price down). In brief, a truly specialized translator knows his field so well he's practically a walking thesaurus in it. The translator has harnessed high-end specialized skills, "blended" skills that combine overall verbal agility with deep knowledge of a specific field —and particularly, of its discourse. How deep? Deep enough to be impressively resourceful in a content-brainstorming session with other native speakers who work in that field. If you can do that, you've passed the specialization test." Real specialization requires that many different subject areas are permanently excluded from ongoing consideration. Financial premium market translators do not volunteer to translate virus detection diagnostic equipment related material because the money is good. Ever. Just as a top dermatologist would not agree to perform cardiac bypass surgery. Ever.

Lisa Rüth states the key characteristics of premium market translators as: "Firstly possessing excellent writing skills and knowing how to apply them to the client and the situation (think "purpose" and "target groups"). Secondly, it means possessing expertise in what you are writing about – fluent subject-matter knowledge." This expertise also means that a premium translator is more likely to use more authentic, field-specific terminology that has a desirable and favorable communication impact.

True specialization and real expertise will matter even more in this age of increasingly good automated translation. The MT will continue to improve, but we are still a considerable distance from computers that understand and can infer the communication intent and the impact of a translation. The premium market and translators who have proven subject matter expertise are likely to see little effect from better MT. However, for translators who work with bulk, low-value, short shelf-life, commodity content NMT is already producing compelling "good enough" translation in many language combinations. It is reasonable to expect that many bulk market generic (unspecialized) translators will feel the downward price pressure and displacement impact of improving MT in the future.

Chris Durban regularly illustrates this point through side-by-side comparisons that underscore the widely varying skills of translation suppliers. Her aim is twofold: demonstrate that demand exists for genuinely expert translators who have developed the high-level skills described by Rüth. And also point to trouble ahead for suppliers delivering work that, while usually devoid of grammatical and spelling errors, mimics source-text syntax and offers little added value on MT as shown below.

Source: Chris Durban contrasting translation output from MT, Bulk and Premium translation

Source: Chris Durban, Table showing MT, Bulk, and Premium translation comparison.

Translation Process Modifications

Premium translation is focused on producing useful and natural communication in the target text, which means that translators are allowed and expected to ask questions about the conversation and brand intent in all high-impact source-text messaging. Translators are advising stakeholders on the best and most natural communication practices in the target language to ensure that communication is effective. This advice can involve source reconfiguration and the development of non-literal translations that are simply more effective in the target text. Of course, this cannot be done for all the documents that are translated in the modern enterprise. Thus, information triage is part of the consultative and collaboration process, and decisions are made to reserve this more engaged process for "content that really matters." This type of content will include high-impact financial market communications, operation of nuclear reactor controls, military and weapons systems, science, and technology research, critical medical care related material, cybersecurity processes, and other enterprise survival and life and death outcome related content.

Banks and other firms looking to optimize multilingual communications processes need to start by identifying [high-impact, e.g., earnings reports] image-critical content and other key texts. Then, they need to develop differentiated business processes [to handle these] as part of an overall multilingual content strategy.

David Jemielty, BCV

More Frequent and Engaged Customer/Translator Meetings

Keeping translations on message and keeping the communication intent clear in the target text is something that can only be done through in-depth engagement with content producers and stakeholders. However, as these meetings can sometimes be with C-suite executives, the translators involved need to be efficient, precise, and prepared for these interactions. Sessions are likely to be short, efficient, and focused on the perceived impact of the translation text, rather than on linguistic accuracy of the translation per se. As David summarizes, "You should fuss lovingly over the meetings in the same way that you fuss lovingly over your translations." These meetings are sometimes needed to address low-quality source text issues and ensure that the impact of the target text communication is not compromised because of this.

Again, David says it best: "The most successful firms and organizations foster a dialog-based culture of translation for their high-end, image-critical content. At the most basic level, this means that translators should ask questions because that is the only way to successfully negotiate the "incongruencies" that exist between languages (i.e., the differences in the way languages are structured and the differences in their stylistic preferences). However, a truly dialog-based culture of translation goes beyond Q&A. When it comes to translating high-end, image-critical content, "dialog" means just that: there should be an ongoing conversation between the translators and the text owners on how to best render the key phrases, terms, and ideas in each text. When the process is working correctly, each side makes concessions.

What is notable here is a willingness to sacrifice consistency among the various linguistic versions when there is a good reason for doing so, such as a marked difference in stylistic norms between the languages concerned. As the American writer, Ralph Waldo Emerson famously wrote, "A foolish consistency is the hobgoblin of small minds. "

Such processes take time, however, and not just translator time, but also text-owner time (which can be rather more expensive). Therefore, they must be employed sparingly, on the texts that count. In addition, translators must be hired and trained to perform well in such "give and take" situations."

Management of Perceived Quality – Buy-In

Another purpose of the increased buyer/translator engagement is to build and establish the reputation of the translators as experts who are aligned with the higher organizational purpose of producing effective communications across languages to further the international mission. Executive buy-in is obtained by clearly showing the benefit of a recommended communication (translation) approach and by providing ongoing examples (benchmarks) of value-add that go beyond the literal translation of the source text. 

It seems that in many organizations, the premium market needs and bulk market expenditures co-exist. All content is not equivalent, and there is a role for both kinds of translation services.

Comparative Translator Culture

The culture among premium translators is much less concerned about competition, and peers are much more likely to be friendly and collegial even when they work for the same client.  There is a striking contrast to the zero-sum game, high price pressure, and rapid turnaround expectation that often characterizes bulk market relationships, which are also less sticky. Clients who build trustful relationships with premium SME translators seem to want to keep these in place. Trust is not easily replaced and is earned over time, usually with long-term competence and demonstrated expertise. The presence of trust also engenders mutual respect.

Premium translators are also much less likely to be present in social media and more likely to commit to ironclad NDAs and data security requirements, which also limits the number of customer case studies that document their work.

I also noticed that premium market translators are much more willing to share examples of their work products and demonstrated competence and often have a work portfolio ready to document expertise properly.

"Subject-matter knowledge, writing skills, willingness (and ability) to ask questions—those are three skills that my segment of the market demands" says Chris Durban. She characterizes the premium market culture; thus: " Full-time translators with genuine specializations regularly point out, the premium end of the market is crying out for skilled practitioners. And based on their accounts, what's not to like? Intellectually challenging texts plus the satisfaction of giving these a voice; engaged clients eager to discuss nuance and impact; recognition—even gratitude—from demanding buyers; attractive remuneration." The premium sector is an attractive and growing market, but one that requires practitioners to master an exceptional skill set described here.

It is also unlikely that we would see many premium market participants at localization industry events. They are much more likely to focus on industry conferences that their clients focus on, or on translator events where accreditation and professional qualifications are more visible. Some examples include ,, BDU events, and specialized events like this.

Concluding Remarks

There is indeed a premium market, probably quite substantial in size, even though it is hard to discern for those whose primary focus is the bulk localization market. This market has different professionalism and expertise requirements and has a mostly different business model. It has always been visible to us in the MT community as one end of the information triage chain. Fifteen years ago, I was surprised to discover how large it was, even for the US intelligence community, one of the most significant users of MT worldwide. Many are not aware of the fact that their spending on highly specialized human translators was fifty times higher than their spending on MT. That's a remarkable fact when we consider that the US intelligence community has always been a pioneer of AI and MT, a trend that continues today. However, they understand that MT (technology) and expert humans occupy different parts of the value chain.

The premium market is an opportunity lost and missed by the bulk localization market players whose attention is consumed by the reality of survival and competition in the bulk translation world. Most of the private equity bets in the translation market are focused on trying to make bulk more efficient and streamlined, but IMO we have yet to see any of these investments bear fruit. The fact that expert-driven professional service giants are looking around the premium market suggests that this could change.

The translation service business is built with interacting technology, business processes, expert human input, and clients with a global mission. To me, the most valuable element here has ALWAYS been expert human input. However, I have yet to see this properly reflected in the localization market, even though it seems glaringly apparent in the premium translation market. It will not surprise me if the disintermediation of the localization industry comes quietly in the night through an opening in the premium market.

An opposing opinion by Luigi Muzii on the Premium Market can be viewed here: Lead and Gold: Challenging the Premium Translation Market Claims

This is a link to a response by Kevin Hendzel to the opposing view stated in the link above: The Premium Translation Market: Come On In. The Water’s Perfect.

Tuesday, August 11, 2020

The Continuing Evolution of Enterprise MT

I recently hosted a round table panel discussion organized by Multilingual with three leading enterprise MT players, to discuss and share thoughts on the state of Enterprise MT today. The panelists are all firmly established in the enterprise MT arena, and a key objective of the session was to contrast the focus and differing requirements of enterprise MT from the generic consumer portal MT that is so easily available today.  

MT technology has become pervasive, and today more than a trillion words are translated every single day by the many free MT portals on the web. However, the MT needs of the enterprise are different and much more specialized. Generic MT technology is not always adequate for the many new uses that a global enterprise might have for high volume translation. In this panel discussion, we talk to three leaders working in the Enterprise MT space about:

  • Customization and adaptation requirements for the enterprise
  • Issues in measuring and understanding relative MT output quality
  • The outlook for continuing improvements in core MT technology
  • Use cases for MT beyond localization

The MultiLingual Summer Series: Meaningful Conversations with Thought Leaders have other interesting sessions coming up in August.

The panelists present in the session included:

Chris Wendt graduated as Diplom-Informatiker from the University of Hamburg, Germany, and subsequently spent a decade on software internationalization for a multitude of Microsoft products including Windows, Internet Explorer, MSN, and Bing – bringing these products to market with equal functionality worldwide. Since 2020 he is leading the group of language services in the Azure Cognitive Services family, bringing natural language processing capabilities to its customers.
Alon Lavie leads and manages Unbabel’s US AI lab based in Pittsburgh, and provides strategic leadership for the AI R&D teams company-wide. For almost 20 years (1996-2015) Alon was a Research Professor at the Language Technologies Institute at Carnegie Mellon University, where he continues to serve as an adjunct professor.
 Joern Wuebker, Director of Research at Lilt. the AI-powered enterprise translation software and services company. Prior to joining LiltJoern earned a Ph.D. in computer science from RWTH Aachen, where he focused on machine translation research.

The complete session is available as a video on the Multilingual site at the links shown above, however, it requires a viewer to sit through the full session. While there are some solutions that allow video content to be indexed and annotated, and thus make the content discovery easier, these are not in wide use yet. Thus, I thought it would be useful to provide subject-related links to the different sections of the full session in this post to allow a viewer to selectively focus on the subject matter of greatest interest. I have also summarized some of the important points in my own words and added new comments where I thought it might add greater clarity.

What are the characteristics and requirements of enterprise MT? How does it differ from generic MT?

  • Strong focus on data security and privacy of customer data
  • The adaptability and customization possibilities available to tailor the MT system to unique customer requirements
  • Service level requirements to meet corporate IT requirements: Scalability, Batch vs. Interactive
  • Matching and adjusting MT quality to different enterprise use case requirements: correct translation in context and proper handling of relevant content types especially when MT is used without downstream human adjustment
  • Handling of varied content types (structured documents) and formats in different business workflows
  • Maintaining data privacy in dynamic workflows and data interactions
  • The growing need for interactivity and dynamic adaptation to meet organizational requirements as rapidly and efficiently as possible

What do we mean by adaptation and customization to meet unique organizational needs? 

  • Offline adaptation can often be ineffective referring to TAUS study according to Lilt who favors dynamic and interactive human-in-the-loop adaptation which seems especially well suited for localization use cases but could be less useful for very high volume use cases like eDiscovery and eCommerce
  • The TAUS COVID MT study should not be used to draw larger conclusions
  • Enterprise content is constantly changing and so MT engines need to evolve and cannot be static for best outcomes when using the technology
  •  Adaptation should be driven by content variations and evolution
  • Microsoft has seen that on average their customers get 10+ BLEU improvements from their customization efforts
  • Handling of unique terminology, style, and branding correctly is increasingly seen as a critical need for the enterprise

How do we handle the problem that there never seems to be enough data to properly customize or adapt an MT engine to unique customer requirements?

  • Lilt claims they don't need large data volumes to adapt an engine, but rather they need very specific and relevant in-context data that is current and reflects the immediate focus to enable effective adaptation
  • Adaptation can start with just a few key terms and evolve with TM and/or human feedback that can be gathered over time
  • Context and high relevance is now increasingly being seen as more important than data volume
  • Make the customization process demand-driven. Focus your correction and improvement processes on the content that matters the most
  • Data manufacturing is of dubious value but there are some synthetic data strategies that make some sense e.g. back translations which can be "surprisingly effective".

How do we address the challenges around evaluating MT quality given the increasing awareness and recognition that BLEU scores are not as useful with NMT?

  • Understand the source data and choose the right metrics to determine accuracy. At Lilt, they use "next word accuracy" as a key measure to determine quality evolution progress
  • The purpose of  MT deployment needs to be properly understood first? Does the MT solve the problem? Does the machine translation improve user experience even though it is not linguistically perfect?
  • BLEU is a better laboratory evaluation rather than a real-world evaluation and still has value to developers
  • A human evaluation has to be involved to get a better picture of the MT quality reality but these are expensive and slow 
  • Unbabel uses MQM based error evaluation techniques to develop improved quality predictors 
  • Unbabel has developed a neural-based evaluation approach that is showing great promise with an initiative called COMET. Details will be coming forth soon.

How does Lilt measure MT output quality improvements given that their engine is changing dynamically?

  • Lilt measures every sentence and learns from every single sentence
  • They can tell from the improvements in "next word accuracy" and see how it evolved from an initial engine
  • GPT-3 next word prediction is not a threat to our approach as it is monolingual and requires a huge model that would make it impractical

How do we handle MT use cases beyond localization where potentially billions of words are being translated and a human-in-the-loop deployment is not practical?

  • Raw MT can be used to make much more content available to global customers even if the value of this is unknown. The enterprise can make more content available to identify what content has the greatest value in different markets
  • User-generated content (UGC) is hard to customize around and generic engines may work just as well
  • Communication scenarios are much more important to the business mission usually than UGC, even when an enterprise has very short shelf-life content and thus quality measurement happens in realtime e.g. chatbots, live support
  • Communication content strategies vary for inbound (UGC) and outbound (customer support). Outbound communications can involve human-in-the-loop (HITL) if the content turnaround times allow this or if the content needs it.
  • Realtime quality estimation capabilities become a much more critical enabling element for other use cases especially with social media and user forum communication where it is desirable to use raw MT.

What is the difference between linguistic steering and post-editing?

  • Focusing on high-density language patterns rather than the full corpus to ensure that critical word patterns are properly learned and handled. PEMT is a full corpus evaluation strategy
  • Unbabel develops dedicated test suites to ensure that the quality is acceptable for the linguistic material that really matters
  • Raw MT can create the danger of making a catastrophic mistake and experts need to find ways to identify and handle these kinds of errors
  • Lilt feels that they could handle the translation of a 25M word corpus through their interactive and dynamic evolution and predictions will get better and better. Lilt can push the boundaries of HITL to handle a very large corpus.
  • Text-generation technologies are very cool but they actually do NOT really understand the text
  • Catastrophic failure will be common and easy to achieve :-)
  • Large LM approaches can help with document-level context but they are mostly monolingual 
  • Also, these approaches are not as effective as dedicated NMT engines on handling the translation task
  • These new breakthroughs can be used to leverage other downstream NLP tasks
  • The primary issue is that they are mostly monolingual at the moment, but they could be valuable as they become more multilingual
  • COMET is built on top of some of these initiatives
  • Multilingual training experiments have shown that there can be improvements for low resource languages even if they are unrelated

Audience question: Given Cybernetics 2.0, how crucial is human feedback for NMT?

  • Human feedback will always be required 
  • Language is a living thing and will continue to evolve
  • Consider COVID 19, how many MT engines knew about this even 6 months ago?

Do you see continuously improving NMT displacing ever greater numbers of human translators?

  • Machines are already translating 1000X what humans are doing  so this is already true in terms of volume but humans are the primary communicators 
  • The roles of translators are evolving as technology improves but there are a finite number of human translators and HT will always be needed
  • "Human Translation" is not monolithic, and there are many kinds of translators, thus, we should be careful to not lump all translators as equivalent, as there is a wide range in terms of competence and expertise
  • MT cannot replace competent, subject matter expert translators who understand the communication intent, the semantic core and comprehend the communication impact of a translation
  • Competent HT will always be the final measure for quality 
  • As long as there are humans there will be a need for human translation.
This last question evoked some spirited social media response and I saw this comment being retweeted many times on Twitter.

The full session can be viewed below and I will add a link of the KUDO version which will also provide interpretation in Chinese, Italian, Russian, and Spanish when it becomes available.