Discovering Hot Water

Reinventing the wheel

According to Common Sense Advisory’s (CSA) analyst Rebecca Ray, “the majority of the world’s translators (who deal with written words) do not translate books. The largest amount of work in this field comes from the manufacturing sector.”

In an interview with Karen Tkaczyk, Mrs. Ray also said that “the majority of language suppliers still aren’t differentiating or marketing themselves very well to this vertical” and that “buyers in this vertical told us that they rely principally on three methods to identify potential suppliers: their own experience, recommendations from peers, and translator portals.”

CSA’s research report on the manufacturing sector concludes that there is opportunity “for suppliers to differentiate themselves through targeted web content, smart use of SEO, and the purchase of URLs related to manufacturing and translation.”

Also, in an article for the online marketing magazine Business 2 Community, Mrs. Ray argues that 72.1% of international consumers spend most or all of their time on sites in their own language, and that 13 languages (English, Japanese, German, Spanish, French, simplified Chinese, Italian, Portuguese, Dutch, Korean, Arabic, Russian, and Swedish) allow for access to 90% of the world’s online wallet.

With these premises, translation could even cost as little as 10 cents (supposedly of US dollars) per word, especially when considering that the average cost to create original content is 65 cents (again supposedly of US dollars) per word, even though interviewees told CSA for its first research into content management that they really don’t know and estimates vary from US$5 to $100 per word.

Anyway, 10 out of 65 is some 15% of content production cost; translation in 13 languages would be twice as much.

Also, translation doesn’t necessarily have to be difficult, even though people looking at the world as a giant market to exploit only for domestic purposes should gracefully show a better attitude towards languages and culture. Not surprisingly, still, courses in translation and interpreting in the U.S. are admittedly scarce — at least in comparison with Europe — despite the need for language specialists in government and commerce.

The main translation market is overseas, and yet so many experts feel the urgency of reminding their countrymen that many companies have already learned the most painful lessons in global markets.

In reality, they all know that translation is not easy, although others contend this, pretending that adding translation memory to the mix and eliminating repeated translation costs entirely would send ROI to infinity and beyond, as it were just like adding water to an instant coffee powder in a mug.

The first question to ask is: do the many Seth Godin’s fans really learn his lessons? In a blog post dating back to January 2010 (eons ago on a Web 2.0 timeline), Seth Godin wrote that all customers simply want the cheapest product or, more precisely, the product with “the right specs at the right price,” since “they have been seduced into believing that all options are the same, and they are only seeing price.”

Therefore, aiming at educating your customers to differentiate yourself cannot but prove ineffective, and rather than trying to fix the market, one should better look for a new market and a new model.

According to another CSA report, freelancers are at the very end of the translation supply chain, but their views play a significant role in the market. Is this any news for anyone?

The report was based on a survey of 3,165 freelance translators throughout the world, when major experts from CSA itself repeatedly maintained that translators are more than 300,000 in the world while LSPs are more than 22,000. According to such estimates, each LSP should be able to count on a stable workflow of less than 15 linguists to cover all language combinations for any kind of job.

The report also names translation agencies that were listed by freelancers as having reputations as poor payers, as well as those that had reputations either as high-quality or low-quality providers. Companies mentioned in the report include ASET International Services LLC, CETRA Language Solutions, CLS Communications, Corporate Translations, Crimson Language Services (a division of TransPerfect), euroscript International S.A., Geotext Translations, Lingo24, Lionbridge Technologies, Moravia Worldwide, thebigword Group, RR Donnelley, SDL, and Translated.

Not surprisingly, CETRA Language Solutions’s founder and president is former ATA President Jiri Stejskal, whose office will be remembered for his letter on machine translation to President Barack Obama, TransPerfect boasts its quality management system as certified to ISO 9001:2008 and EN 15038:2006, and Lionbridge, RR Donnelley and SDL are public companies listed in stock exchanges.

According to the CSA report, on average, freelancers receive approximately two-thirds of their income from translation agencies. More than one third have been victims of a translation agency failing to pay them for work completed. Even more translators have turned down jobs from a translation agency because colleagues had warned them about the agency’s reputation, and the vast majority has turned down work because the agency’s rates are too low.

Not surprisingly, among the many Seth Godin’s fans there are also a few other self-avowed hypocrite entrepreneurs with an urge to educate rookies on the rules of marketing, from the shallows of their ineffective experience, unable as they were to apply the same rules to their own businesses.

Could this not be a good starting point for business model innovation? Name it sincerity, fairness, righteousness, morality, reliability, professionalism or, simply, decency.

Technology-driven Innovation

In a recent TechCrunch article, the 147-year-old Finnish company Nokia is cited as an example of decline for lack of innovation, after a long period when it was a global technology star. The cause of the company’s decline looks very simple with hindsight: Nokia should have moved off its strategy years sooner than it did.

The core problem that brought Nokia down is not unusual for successful public companies that worked their way into a prominent marketplace position over a period of years. This success evidently made it harder for the management to change the company’s business to react to the looming threats from Internet-focused companies, centering on incremental innovation of existing products rather than aggressively pushing a disruptive innovation.

There is not one clear event to which one can point and say “this is when it happened.” Instead, technology slowly insinuated itself until it reaches a critical mass, to be suddenly noticed.

Economist Joseph Schumpeter argued that industries must incessantly revolutionize the economic structure from within, i.e. innovate with better or more effective processes and products.

The attitude of the language industry towards innovation is still technology push, rather than market pull or demand pull. This explains why innovations in the language industry have always been introduced by outsiders.

On the other hand, many efforts are still made on minor aspects of the business. Too much attention is still given to quality following the usually vague and outdated approach. According to insiders, for example, current efforts in the ASTM Committee F43 to create standards for language service companies are still addressing process and quality standards. This multi-year effort is still afar from a draft standard, although it will hopefully go to ballot sometime in 2013. Parkinson’s Law states that work expands so as to fill the time available for its completion…

Unfortunately, reacting when a technology advance is pervasively deployed is too late to fully profit from it. The Internet is still at an exploitable stage, even though it is based on a mature technology and is widely deployed.

On the other hand, the organizational structure of all major LSPs is still based on achievement, legacy, and loyalty. At least ostensibly. In this respect, the Peter principle and the Dilbert principle are paradoxically actual: companies tend to systematically promote their least-competent staff beyond their level of ability to limit the amount of damage they are capable of doing. This explains why people have the impression of being surrounded by incompetents when dealing with translation businesses. And why ‘educating the customer’ is nonsense.

More Technology or New Business Models?

In his presentation on the top 10 trends and predictions for the professional translation industry for 2013 and beyond, Dion Wiggins, Asia Online’s co-founder and CEO, reaffirms that the world is increasingly instrumented and interconnected, with the volume of data produced doubling every 18 months.

Everyday 15 million of gigabytes of new data are being generated. The data generated in 2012 was 5 times as much that generated in 2008. The data generated by 2020 is expected to be 20 times as much that existing today.

Does more data mean more translation?

Square peg, round holeCSA estimates that the demand for translation is 5 times that of the current capacity of all LSPs that translate a mere 0.00000067% of the text information created every day. Google Translate translates more in 1 day than all human translations in 1 year.

This huge amount of data opens new markets for translation: for enterprise information (human resources, training, reports), e-mails and instant messaging, support information and knowledge bases (call center, help desks), and the so-called user-generated content (social media, forums, reviews). This huge amount of data needs to be analyzed and possibly translated quickly and cheaply.

Translation automation will be the answer, at least for discovering information in cross-border litigations, patents, research, market analysis, sentiment analysis. It will also be an essential industry tool, not only for the language industry, and innovation will again be coming from outside.

Even though translation productivity rocketed in the last two decades thanks to translation technology and translation memories, the current level is dramatically inadequate. Moreover, LSPs are having trouble finding qualified and skilled translators, mostly due to lower rates in the market and more competition for the same resources.

Many have already been trying to capitalize on the market opportunity created by translator shortage, but delivering sub-standard services. New LSPs and translators appear on the scene every day, with obvious lack of experience and lower quality offer.

In posing the Edge 2006 annual question, Steven Pinker wrote that “the history of science is replete with discoveries that were considered socially, morally, or emotionally dangerous in their time.”

Business model innovation requires “dangerous ideas.” In fact, as CSA reports and Mrs. Ray’s comments confirm, translation buyers are not improving their understanding of the importance of translation, even though enterprise globalization awareness is fairly widespread and mature.

In the last three decades, however, translation buyers have been learning how to improve their processes, especially as to requisition, purchasing, management, and metrics, and this is affecting every procurement element, including translation, which is increasingly treated the same way as any other supply.

Industry fragmentation is resulting in a few incumbent MLVs and some dominant SLVs in a number of markets, with MLVs using more SLVs and less freelancers. This is partly due to most MLVs’ inefficient business models. MLVs will keep exerting their pressure on price to their vendors, and this compression will drive smaller/slower SLVs and freelancers out of business.

Using Technology to Solve Problems

Albert Einstein is often reported to have said that a problem cannot be solved at the same level it was created. For this reason, before resorting to technology to find a solution to any problems, a thorough analysis of the current scenario is necessary to detect any organizational or process flaws. These flaws can hardly be fixed by using technology alone.

Nevertheless, in the automation age, machine translation seems a panacea: it is fast and cheap. In the light of this reflection and Seth Godin’s considerations above, the natural inclination of any customer is to wonder why going to the expense of human translators at all.

Many LSPs are then recklessly trying to ride the tiger, adding machine translation and post-editing of machine translation to their offerings, with no adjustments to their business models. However, even in the second decade of the XXI century, most LSPs are still tied to century-old — and long outdated — production and pricing models like TEP and word count. The same applies even for new collaboration platforms, clearly designed to meet the needs of outmoded users.

And yet, every day countless tales of woes are circulated every day due to reliance on cheap machine translation. Many others tell of unprofessional translators, possibly posted by disappointed customers and users as well as overzealous professionals who look at the speck of sawdust in their colleagues’ eye and pay no attention to the plank in their own eyes.

Nevertheless, no disappointed customer, although faced with a long chain of evidence of just how wrong cheap (machine) translation can go, will jump to the conclusion that going solely with high-end human translators is the only safe course. Mostly because — contrary to what some experts may tell — translation could be a vehicle to increase sales in markets abroad, but its ROI is never ascertained. And even the user manual of a very simple appliance, or a small typical website, when translated manually into just one foreign language, would cost a lot and take weeks.

The customer willing to sustain the reputational damage caused by a bad translation probably has no reputation to keep safe, and is eager to reach any prospect customers or increase sales. To the customer with a reputation, no loss of prospect sales or customers can equal the brunt of law suits for damaging, badly translated content.


No celebrated TMS can improve poor human translation. Instead, good specification and good sourcing possibly will.

Good sourcing is essential to have clean data to use with statistical machine translation (SMT). In fact, even SMT engines do not learn about terminology, grammar, and style characteristics of a particular company’s content, at least alone or simply by using bi-corpora.

For effective SMT, language corpora should be inherently consistent. Term bases are important even for SMT, and plain texts can help keep data clean. To this end, they could possibly be authored using controlled languages.

To sum up, it is true that the cost of human translation would probably continue to decrease, but this would definitely not be due to improved tools and more efficient production models.

The quest of customers for one-stop shops is partly at the origin of LSPs’ absolute faith in translation technology. This faith comes from a lack of entrepreneurship in most LSPs, that try to exploit automation to make their business most profitable, and resort to low-cost high-profile (unfortunately not an oxymoron) resources, investing as less as possible.

On the other hand, technology today is an integral part of our daily lives, and this makes it increasingly unlikely to do without.

The quest for one-stop shops is due to the perception of translation as non-strategic. In Too Big to Know, David Weinberger offers interesting food for thought. Today, employers may have labor above average, at a cost lower than the average, from the same countries they are targeting, while automation software is getting cheaper and cheaper.

One-stop shops allow customers to drain value through the unique value contribution that makes the difference between them.

Translation proxyOutsourcing also allows for using labor force as a financial leverage: reducing flows when any problems are looming and then exaggerate in the opposite direction at the first sign of recovery.

Fragmentation is the industry’s real essence, not just a deprecated effect. Fragmentation makes disintermediation extremely hard and facilitates key players only — the largest ones, those with the highest economic capacity.

Disintermediation in the translation industry could only be technology-driven and produce and/or be produced by a radical change in pricing models. Machine translation could be the main engine of this change.

Times and technology are mature for disintermediation. In the near future, MLVs will be replaced by those SLVs capable of competing on price by squeezing their overhead and allocating savings to compensation to retain the best resources.

The Internet is still a tool, largely unexplored, for disintermediation and OpEx compression. The first stage in disintermediation consists of translation proxies.

The second stage is integrating translation proxies with (collaborative) translation platforms, possibly at an LSP’s, where customers could pick resources from a freelancer/vendor database.

Both platforms are not ‘invest and rest’ or ‘fire and forget’ service. They require a level of attention and planning from both parties commensurate with the level of work done at a content production level.

The success of disintermediation will depend on the ability of LSPs to keep offering the typical value added services of intermediaries, such as counseling.

Translation as Facilitator

In Understanding the Translation Buyer, Kirti Vashee recalls that a large part of translation is related to content facilitating and enabling international commerce. Much of this content is mandatory and/or necessary to sustain the presence in target international markets.

This the reason for outsourcing translation to LSPs, and for the quest of one-stop shops. While content production still make economic sense for companies to be internal, translation does not.

According to the Economist Intelligence Unit, the independent business of The Economist Group providing forecasting and advisory services, none of the Western economies will grow as fast as the Eastern ones. Nevertheless, this does not mean that they will prevail, even on some of the fastest shrinking economies. Therefore, the FIGS+BRICS combination will still be allowing to access more than 90% of the world’s online wallet.

However, 2012 definitely showed that, to be actually global, companies need to do more to build a satisfied and loyal customer base. Marketing messages are now indistinguishable from propaganda and more and more customers look elsewhere to determine the real truth about anything they might consider buying.

This happens in all industries, including the translation industry, and yet few LSP are showing to have learnt the lesson. GALA recently alerted their members of the need to cooperate, collaborate and develop meaningful standards and be smarter than in the past, as those who don’t will soon find themselves losing out on opportunities and falling behind.

Lagging Behind

Despite the alert for collaboration, in the same December 2012 issue of the GALAxy Newsletter, Sue Ellen Wright declared her support to ISO/CD 17100 (which is still at the initial ballot stage after more than two years) as “essentially a process standard […] with the potential to serve as a replacement for CEN’s EN 15038 as basis for the certification of translation service providers (TSPs).

At the same time, ASTM’s International Technical Committee F43 is struggling with the revision of the practically unnoticed F2575 – 06 standardto meet current industry practices.”

As Mrs. Wright rightly put it, a great effort is needed “to reconcile business interests and cultural differences.”

Not surprisingly, customers with backgrounds other than translation frequently do not understand the industry, feel strange about the translation process, and always ask about metrics and measures, reflecting a desire for quantifiable data to assess performances.

Mrs. Wright also pointed out that ISO/WD (Working Draft) 14080, Assessment of translations, was hastily terminated in May 2012 due to unresolved issues and administrative hurdles, while “many in the industry are clamoring for simple, transparent solutions, but unfortunately, it is not a simple problem.

A translation for humans to read of this last statement could be: when not enough real industry experts are gathered to produce a useful standard a failure is unavoidable. Translation is money, not philosophy; when buyers realize translation academics and pundits are incapable of talking business they leave them to their destiny: fiasco.

The translation industry cannot stand any academic exercises any longer, especially if these are aimed at freezing the existing obsolete framework. The elite of the industry, if any exists, should make a real effort to innovate.

Rocking and Innovating

Recently, Marta Stelmaszak and Anne Diamantidis of GxP Language Services started People Who Rock the Translation Industry!, a series of interviews to “people who have made a positive contribution, no matter how small or large, to the translation industry – at the international, national or local level.”

The series started with a mutual interview and continues now with weekly lessons on doing business, marketing, and development in translation for freelancers, project managers, newcomers, and anyone who’s interested in growing their careers with languages. Ostensibly Mrs. Stelmaszak does not lack confidence in asking for absolute trust, although she has been in business for 6 years only.

The fourth episode of the series dedicated to the industry’s rockers was Aurora Humarán’s turn to attack crowdsourcing, machine translation and post-editing.

Much too easy and predictable.

No machine translation vendor is so stupid to offer cost reductions, or time reductions, or both to prospect customers without warning them about the risk of unknowingly using translation automation. It could backfire.

No customers is so stupid either to invest some hundred thousand dollars in a machine translation engine to achieve immediate savings. Those who are so stupid to do so are probably uninterested in buying translation at all.

Take Google Translate. It is not just a tool enabling people to translate information on the web. It is a strategic tool for Google itself, whose self-declared mission is “to organize the world’s information and make it universally accessible and useful.” It is impossible for Google to accomplish this without translation.

As Jost Zetzsche reminds us in the latest issue of his newsletter, “Machine translation has found its way into translation environment tools and the production process of many professional translators.

Someday soon, we won’t talk about machine translation any more, and yet we could be still talking about translation, as much as content development.

So, is blaming translation agencies and machine translation vendors rocking the industry? Or is it just craving attention? After all, people in the translation industry have never been prone to dangerous ideas.

The sheer rejection of machine translation is based on the same old pattern as ever, as with typewriters, PCs, translation memories, etc.

Is conformism rocking the industry?

The translation industry follows a typical conservative approach: any change is seen as a major change.

Our world is getting faster and faster, companies produce more and more content, and run business in more and more countries, but the translation industry is still at a vendor management stage.

Something is wrong in the translation industry if ‘experts’ intend as large a company with more than 5 employees. Things are getting worse when considering that these companies can hardly even know the name of the people they owe their success, translators, even when they have been relying on them for years.

Not surprisingly these very companies crowd vendor management workshops. And yet, the first and foremost advice given in these workshops is to keep the vendor database up to date.

Vendor management is the largest cost budget item, and could be huge for companies with hundreds or thousands of vendors. It could prove uneconomical when sharing the same pool of resources with other companies, and still going through testing vendors, which is expensive and not reliable.

Few concepts in business have been as popular and appealing in recent years as the emerging discipline of open innovation, based on sharing risks and rewards with partners.

It requires ideas, external and internal. So the final questions are: who needs such ‘experts’ as those named above, especially to drive innovation? And, most importantly, is innovation possible with this crowd and the approach described above? Where is the CSA of The End of Localization Taylorism and Localization Maturity Model?