By Scott Carothers | Member
Technical writing professionals are amongst the most knowledgeable members of an enterprise, navigating through tasks that are typically scattered across numerous business units, silos, and departments. Often referred to as a “jack-of-all-trades,” being a technical writing professional means knowing every aspect of content delivery and process building. The job is a constant battle of simplifying the message, learning shortcuts, and maximizing content reuse. In today’s fast-paced world, deadlines never expand, they only truncate between the enterprise and the customer. For global enterprises, translations continue to be a growing aspect of a technical writer’s expertise. However, few have been truly educated in the complex process of translations, and few enterprises have a well-monitored, transparent process with real-time business intelligence available.
The following is a brief case study detailing how one Fortune 10 enterprise came to grips with their translation needs and made their process more manageable, efficient, and cost-effective. There are several important takeaways below that can make a major difference in your operations.
How Global Corporations Approach the Translation Process
One technical communications manager within the Fortune 10 company shared the following: “The translation process is like a black hole, I work really hard to get the content right, and then it goes to a translator and disappears for weeks. There is no transparency into the process, and no business intelligence to tell us if we are on schedule, on budget, or if quality is improving or deteriorating.”
Similar comments were made by an indirect sourcing manager regarding the enterprise as a whole, but with more specificity. He said, “I have four main concerns with my company’s translation process. First off, there is no uniform, enterprise-wide process that simplifies the management of this process for all of my users across business units, silos and departments. Secondly, our TM [translation memory] is scattered across several vendors, which is unacceptable. This should be centralized and maintained in-house to ensure proper use of 100% word matches, fuzzy matches, etc., before the projects are sent to a vendor. Also, we need to remain vendor-independent, instead of locked into the two or three vendors who work with each business unit. We need to choose the best vendor for each job. Lastly, we need independent business intelligence on every step of the process, with real metrics to ensure that we are continually improving the process.”
These comments are an unveiled look at the way many global corporations are beginning to approach the management of the arduous translation process. Achieving these results isn’t as hard as it sounds, because they are all interrelated.
Combining the first three elements cited by the sourcing manager above—1) a uniform, enterprise-wide process, 2) a centralized and maintained translation memory (which is their intellectual property), and 3) being vendor independent—actually facilitated and enabled the business intelligence. Let’s dig deeper in to how this trifecta fell into place!
1. A Uniform Process
Providing the enterprise with one simple-to-use translation process meant eventually getting everyone at the Fortune 10 on board. They thought they had identified and converted about 80% of the translation buyers within the enterprise at that point. These buyers were found in human resources, legal, regulatory, e-commerce, Web, training, marketing, polices, consumer communications, consumer help desks, software, information technology, and technical writing.
Getting nearly everyone on board with reinventing this process was certainly the toughest piece of the puzzle. There were plenty of naysayers with the same argument: “Our preferred vendors have a process that works, why would we change it now?” The sourcing manager’s response was simple, “This is not working. There is no oversight, no validation of quality or cost. A Spanish document could end up in the Middle East!”
Once those obstacles were cleared, the improvements began. They now started to see real-time business intelligence for the first time, and that meant that they would eventually be able to negotiate lower translation rates with this information. While it is not commonly understood, translations have reached commodity status; there are 28,000 language service providers (LSPs), and quantity drives pricing. So the more they could aggregate the entire volume from the vendors, the more the costs decreased. This would be impossible to accomplish without a uniform, enterprise-wide process because translations are an indirect cost, often sporadic and always added to more expensive and tracked items, like website expenses, product documentation expenses, etc.
2. Centralized Translation Memory
Centralizing their TM became relatively easy with one uniform process. For those who haven’t yet crossed this bridge: translation memory is source language phrases, paired with the corresponding target language phrases. It’s all of your previously approved translation content, which can be reused in each new document with the same verbiage. As new content is submitted into the process, the system parses the content into phrases, then searches the TM for the same or similar phrase and automatically inserts it into the new translated document, color-coded by degree of the overall match. This reduces the work required by a human translator, lowering the overall cost and speeding up project completion, all while increasing the consistency of the translations.
3. Vendor Neutrality
By centralizing their TM and combining it with their new uniform process, translation quality and process efficiency began to skyrocket. First, everyone was working with the same translation memory, so they started to see greater consistency on their translations projects. Working with multiple vendors, each translator uploaded their finished project (including the source and target language pairs) so that every new project would leverage all previously approved translations. This was in stark contrast to simply farming assignments out to various vendors, each using a small percentage of the company’s intellectual property that they had accumulated from their previous assignments. Different translators may translate the same phrase differently, so the website translation starts to differ from the product sheet, etc. This inconsistency not only causes problems for your customers around the globe, but it also can be quite costly. Glossaries, term-bases, and style sheets were established across the Fortune 10 enterprise for staff and vendors. Subject matter experts and in-country reviewers who make the final translation edits from the vendor’s submission uploaded to the TM and shared the changes with the vendors. Utilizing the centralized translation memory eliminated inconsistencies and additional expenses. More importantly, it became evident that efficiency was increasing.
Translation Case Study
Let’s look at an example project, a 15,000 word technical document. The typical translator will translate approximately 1,500 words per day, so just the translation portion of the project will take 15,000/1,500 = 10 days. Using the centralized TM typically produces a 50% match rate, changing the formula to 7,500 new words/1,500 words per day = 5 days for the new content, plus 7,500 matched words/ 5,000 reviewed words per day, or 1.5 days. Now the turnaround time from the translator should be 6.5 days, meaning that delivery will occur 3.5 days sooner and with greater consistency! Financially, if they were paying $.20 per word, they should only be paying for 7,500 new words * $.20 or $1,500, plus 7,500 matched words at the review rate, which is typically 50% less, say $.10 per word, or $750. That totals $2,250 instead of the full rate of $3,000, a 25% financial savings, on top of 35% time savings.
4. Business Intelligence
With the first three elements in place and well adopted within the company, attention focused on the business intelligence for the process. The first step was to establish vendor performance ratings, which is standard for most vendors but completely absent for translation vendors. To make the best buying decision, three key elements had to be tracked: quality, delivery performance, and cost-competitiveness. And these elements had to be tracked all the way down to each language pair. Vendors are often better at certain languages than others and having that knowledge and leveraging it every day reduces performance issues. Quality could now be measured in defects per 1,000 words and captured in the review process. Delivery performance was automatically captured when the finished files and language pairs were uploaded to the system (they had to include the language pair uploads because LSPs are reluctant to send these to the customer for fear of losing control, but they all want the best ratings in order to earn more business so it has become automatic). Lastly, cost-competitiveness could now be calculated by the system in timed intervals based on each vendor’s calculated price, compared to all other approved vendors for each project.
Having these vendor performance ratings gave the client the upper hand in the relationship, changing the status quo in the vendor-enterprise relationship. While price is always a consideration, it is usually secondary to quality. With the vendor performance ratings, they could now reasonably choose the best vendor for the project, greatly lowering the possibility of complications and delays within the project.
In addition to the vendor performance ratings, they could use metrics to show management the overall increase in volume with the current staff, the specific volume by each language, the decreasing net word costs, and the translation memory growth—all tracked by language pair and comparable over time.
With substantive data, it is much easier to plan, forecast expenses, and even calculate the cost of adding new languages to the website, for instance. It is commonly understood that most technical writing teams typically see this type of metric relative to their daily work via content management systems; what has been described above simply extends the same metrics when applied to the translation process.
It is important to note that this Fortune 10 company spent $12 million annually on translations. In the first six months of implementing the process, they projected to save 42% with the steps described above. This presents one enterprise’s navigation to their optimized translation process. While changing the status quo is always difficult, understanding the big picture will offer insights for everchanging technologies that are now available, even for an enhanced translation process that expands the deadlines instead of truncating them.
Scott Carothers (Scott@theTechnologyAgency.com) is a senior globalization executive at Kinetic theTechnologyAgency. Kinetic provides Globalizor™ software, an enterprise-level, vendor-independent translation process management tool. Kinetic theTechnologyAgency is based in Louisville, KY.
