Cost-cutting, in the translation industry, is often seen like dropping money into a piggy bank. However, it leads to a paradoxical behavior: translation company managers wind up ordering “human” translations they know are nothing more than slightly polished machine translations. I think companies miss this fact because their financial models do not take supply chain management costs charged by translation companies into account, despite these same translation companies listing such expenses in their financial statements and industry magazines openly and bluntly discussing “supply chains” for translation. After the middleman’s fees are taken out, corporations adhering to best practices wind up paying Chinese translators less than teenagers flipping burgers at McDonald’s. This grim reality has birthed a whole culture between translation company project managers and individual translators; one that normalizes the pattern of fraud that results when translators try to earn a living wage by resorting to artificial translation tools to instantly generate completed translations. These translations have no value over what an otherwise free machine translation could provide. As a solution to this paradox, I propose that companies calculate reasonable translation compensation by including these supply chain costs and using an effective quality management system to ensure that value is delivered.
Irrational Choice Paradox
A fascinating phenomenon in China is that most workers are reluctant to use machine translation for communication, whereas most translations to English are minimally revised machine translations. Chinese-speaking WeChat users are often surprised when English-speakers use the program’s built-in translation function and understand the messages being sent. So, why the dichotomy? Despite the marketing and hype, Chinese workers show extremely low trust in the use of machine translation for any purpose. However, the translation company as middleman model, which hides identities of translators, makes it extremely easy for those translators to cheat clients. Given that machine translation has caught up with and surpassed the incompetent casual translator level, it’s difficult for companies to use internal resources to spot fraud. Moreover, the cost-cutting rates used by these translation companies makes it impossible for them to find an honest translator. If honest translators were so cheap, they’d starve to death!
This fact is evident when going over typical financial reports issued by large translation companies and taking a look at what percent of their margins go to subcontractors, as well as those same subcontractors’ own reports on what percentage of their margins go to translators. In my own checks, I found that 60% is a typical figure paid to translators in big UK or Australian translation companies. Subcontractors doing the actual staff management, also registered in the UK, then paid about 50% to their translators. So, half of 60% is 30%. The translation company working for a rate of 14 cents per word would pay its subcontractor 8.5 cents per word, and the subcontractor would pay its UK staff 4 cents per word. At a contracted human translation rate of $10/hour, this amounts to $12,000 for a typical work-year of 1,200 hours. Therefore, these translators would be better off working at McDonald’s. These human resources subcontractors often outsource again to China, where, with another supply chain cost, the average rate paid to translators is 1.6 cents per word, according to surveys.
Now the nominal pay rate is about $4/hour. With 1,200 hours of work available, this is $4,800 per year. This being for college-educated people, in a country where computer language graduates are making $26,000 per year. The corporate financial analysis is only considering cost cutting, and not whether it’s feasible to deliver the service ordered at these prices. The reality is that, to cover the distance to say, $15,000 a year, these translators need to do only one-third of the contracted work and create a fiction about the other two-thirds. At CBL, we provide an audit service as to what, exactly, value has been invested in work contracted and paid for as human translation. Usually, the audits reveal that the translation has actually been degraded from Google Translate quality.
Normalization of Fraud
A great case study on how translation fraud is normalized, and an essential feature of many companies’ and governments’ work process, is an audit of a translation provided for the San Francisco government. It was revealed to media that Google Translate could, by luck, produce an accurate translation for notices about a voter matter. However, it also revealed that the translator produced an incorrect translation that would mislead voters–in fact, would discourage people from voting in my opinion—and the degradation was perplexing to observers. I am not perplexed at all. This is a very common method: a translator for a low-cost intermediary can typically take Google Translate and put it into a scrambling tool like QuillBot, which is used in the content industry to achieve plagiarism on content articles without purchasers knowing. QuillBot will then generate a unique Google Translated result each time, making back-tracking impossible. However, QuillBot will materially change the meaning of a translation, thus is much less accurate than Google Translate.
Why is this issue only caught when it causes a big explosion, and not death by a thousand cuts? Translator compensation predicts maximum translator performance. A business could go directly to translators to place orders for 14 cents per word, which would make for about $42,000 per year of income for a translator. Indeed, this is what nominal translation intermediary rates are, in theory, using. They could go to an ATA Certified translator off the directory for 30 cents per word, which equals about $80,000 a year—what Washington, DC, area linguists (national security people) report to the IRS. This is what a “top” legal translation agency might charge. The ATA Certified translator would absolutely not cheat a business in this circumstance, because the other certified translators could easily find out. Working with a client at those rates, either the journeyman or master translator would get adequate results.
However, businesses are generally not working directly with translators like they do with accountants or lawyers. Rather, they are working with project managers as intermediaries to find translators. And most project managers say that the pressure to cut costs and lower turnarounds is putting them at a breaking point—much less the people doing the actual work. Project managers’ jobs have largely become about finding ways to negotiate with translators to get them to accept rates that nominally equal a 70-80% salary cut over what they could fairly earn if they simply retired from translation and worked elsewhere. Most project managers are actually right out of school and very young, but the ones with experience know that they are engaging in a wink-and-nod type transaction. They will typically tell the translator that they need to meet a number of warranties, but that they would try to sell the client on accepting the translator’s sub-standard work if the latter complains about the warranties not being satisfied. For a project manager, doing otherwise would risk bankruptcy. These facts are based on interviews with people I find at conferences while doing research, but you can personally verify these views by looking at employee reviews on Glassdoor.com or translator blacklists of certain translation companies.
The normalization of fraud is also why students at China’s Masters of Translation and Interpreting programs generally exit the translation field on, or even before, graduation. They could nominally earn triple the salary just by being a Human Resources Assistant at Disney (true story, based on LinkedIn profiles, no insider sources). At translation programs, we don’t train students in the art of fraud. We train them to be honest professionals. This is why they exit the industry: when they go looking for translation company jobs in China, the only opportunity for them is to work in an environment where everyone is treating each other unethically. The focus among translators is largely shifting to passing off the same machine translations workers are reluctant to use, as the “real deal,” usually by checking the document, identifying places where the software broke down, and using rewriting tools to cover their tracks. Overall, about 5% of their time is spent on actual translation, and 25% on covering their tracks. After the 70% cost-cutting, the company has paid 30% of a proper translation’s price for 1.5% of the value. That is to say, if the company were not stuck in this cost-cutting paradox, it would be receiving back 20 times more value per dollar invested. The company is primarily investing in false assurances and self-deception: a kind of “linguistics theatre” where putting on a show is more important that material economic value.
Solutions to the Paradox
Solving the cost-cutting paradox is deceptively simple, but achieving traction and achieving results will require many years of effort and work simply because deception as a translation service has become so normalized. The seed to getting solutions underway starts in the CEO’s office: currently, industry journalists talking to CEOs of companies say that translation and even localization is viewed as a cost center. This is a misleading analysis, because translation directly adds to a company’s profit (or government’s mission)—and therefore is an investment and not a cost center. The problem here is that companies are not doing valuation on the translation work. I previously covered and described how to assess profit-generating valuations for even non-product related translations, such as legal translation. Translation company CEOs actually do know how to value translation using this method, and do it much differently from non-translation industry companies.
For the C-suite, a good comparison case study is to look at websites of big translation companies in China versus the websites of their Chinese clients. Typically, what you will see is that the translation companies are using high-quality copy with very accurate translations; you might be forgiven for thinking they are not from China. On the other hand, their clients, indeed even many of China’s biggest and most prestigious companies, are putting out websites that use broken English. Tencent’s WeChat North America launch failed famously over bad translation; HSBC famously wasted $10 million on an Asia marketing campaign over incorrect, even offensive translations. Even writing-oriented professions, like law firms, put out websites in broken English. Translation company CEOs have something very interesting to say about their CEO industry, and it’s the powerlessness to help CEOs. That’s right: people in the industry want to help CEOs, and CEOs self-sabotage.
Specifically, the translation company CEOs who provide middleman services I asked, particularly in China, and those who gave interviews elsewhere throughout the media basically consistently complained that they lack the negotiating position to provide companies with good work. An economic model produced by MIT economics professor Dan Ariely basically explains that industry CEOs and managers’ economic behavior is irrational. Specifically, they do not value translation services in terms of the absolute value they can produce for the company, but rather determine the value of a translation service based on its price relative to other services. Nimdzi reported that translation service pricing ,as a result, is already deflated by about 50% under what it logically should be in some markets.
That doesn’t mean that translators are living in poverty, however. Rather, it means that the value provided by those translators is being provided relative to other providers, and if those other providers are only offering linguistic theatre, then these translators also will provide only linguistic theatre. This leads to the second solution: financially analyze the value a translator is receiving and therefore providing. In the above example, I pointed out that a translator doing 100% human translation, and doing it properly, is devoting 100% of effort into the translation itself. If asked to give a 70% discount over the logical minimum rate, this means they need to use artificial intelligence to create a superficially passable translation that’s nonetheless a fake. In this case, 95% of their time is invested in making a fake look real, and only 5% of the time is in the actual translation. In this valuation, you could receive 100 units of translator effort with 100 cost units, or 1.5 units of effort for 30 cost units.
This of course recognizes that there is a diminishing marginal value for each additional unit of effort invested. 100% often does not make logical sense; using suggestion tools to bring the effort down to 80% and costs down to 60% would be possible. Translation companies doing the front page of their website and their contracts usually pick 100% effort with translation. When publishing their own blog content that isn’t so front and center, they go with 60%. The 1.5% human effort investment level is reserved solely for clients, and clients only get it because they demand it.
