Startups face different challenges than those of established companies, at least when it comes to localization. For startups, going global means creating content fast, in multiple languages and on a shoestring.
Startups develop incrementally and in a fast-paced manner applying an agile approach to content creation and localization. Traditional high-level design is replaced by frequent redesign. And if you are a localization vendor delivering to these businesses, you need to speak agile, because when you speak agile, you embrace efficiency. There are multiple ways to become efficient in content creation, but how do you measure efficiency? How do you prevent glitches in your localization process? Gathering actual hard data while your project is running is the only practical way to move forward — and agile processes are all about collecting and charting objective data, making it visible to the entire organization (even to the entire supply chain) so there is never any question what the status of a project is and what caused a certain error.
Why are startups different?
Startups explore unknown or innovative business models in order to disrupt existing markets, while more established, larger companies borrow an existing idea and take it to the limit to beat competition. Steve Blank and Bob Dorf, authors of The Startup Owner’s Manual, add that startups are not smaller versions of larger companies, but instead, a startup is a temporary organization designed to search for a product fit and business model.
A startup essentially goes from failure to failure in order to learn from each failure in its search for a repeatable, high-growth business model based on a Big Idea. But the execution of that Big Idea is challenging in multiple ways. Very often, these companies have small teams with no marketing, human resource or sales departments — let alone localization teams. Teams learn by trial and error. Incremental growth is of paramount importance to these companies and speed is essential for beating the competition and for establishing their businesses.
To survive, many startups need to be prepared for international expansion. The backyard is a safe place but you cannot stay there forever. This will have implications on the packaging of the Big Idea — be it a service, a product or a solution. Content becomes key. And the more content the better, but also the more challenging things become. Content produced by startups is by nature nonconventional, requiring new technologies, localization methods and evaluation types. By the way, going global doesn’t mean that you must have all your content signed, sealed and delivered from day one into 50+ locales. But be localization-ready and think localization at an early stage.
Of course, professional localization can be a costly exercise. And there is one way you can save on localization: do it efficiently. What is efficiency? Efficiency is the right mix of productivity and quality. The right localization vendor will deliver the right quality for the available budget at the right speed.
Start by setting up an efficient localization workflow. In the beginning you can do localization in-house, as long as you have the right resources, but a better option might be to outsource it. There are many options out there, from established language service providers to small startups disrupting the whole notion of translation — such as AltLang, which provides translations into different varieties of the same language — to crowdsourcing platforms offering “professional translation services at scale.” Make sure you track the productivity and quality of your vendor to avoid purchasing extras you don’t need.
Buyers of localization services should make agreements before localization projects start. A service-level agreement defined in the beginning of a relationship can cover high-level requirements, but is usually not enough in the long run. Each project will have some peculiarities and will need some fine-tuning. The best approach is to be over-explicit and to leave nothing to fate. Specifications are the king on the localization chessboard: every other piece may fall but these should stand fast. That’s because today there isn’t one single type of translation, there are multiple sorts. When you send something out for translation, there is a large diversity of products and services you can get back. You need to specify what you want and how you want it. Welcome to the world of requirements and specifications.
This is where content profiling comes into the picture. You need to make sure your content is of high quality when it is brand-image-sensitive and highly visible, and good enough quality when good enough is good enough. Why spend a fortune on content that’s meant for gisting purposes with very low anticipated pageviews, or rush projects needing special care?
To elaborate on this last point, evaluating quality of content can be a real challenge when it comes to nonconventional content. Nonconventional content is everything from social media to user reviews and multimedia content such as games and movie abstracts. In a session entitled “Quality Evaluation of non-conventional content” at the last TAUS QE Summit in San Jose, panelists discussed the challenges of localizing and evaluating new content types. One of the aims of the session was to define how to evaluate quality for content with a short life cycle versus content that is permanent and has more visibility.
During the session, Sonia Oliveira (Zynga) and Shaun Newcomer (R2Games) explained that the gaming industry is a dynamic and competitive industry where agile content creation and localization are a must. Prioritization is key when it comes to quality evaluation, maybe more so than in any other industry. Content that disappears fast receives less attention as opposed to quality-critical content that can affect player experience. The aim is to engage the players, to entertain them and to give a bit more each time they return. Game localization very often involves transcreation that is impossible to evaluate using traditional adequacy or error-typology-based methods. Besides which, games are about experiences and game content is only one of the building blocks of that experience. Quite often, evaluation comes down to monitoring the expectations of users: the accessibility, the payment methods, the ease of buying the game and measuring whether a game resonates with a given culture. All these factors are equally important and will have localization consequences.
Lupe Gervás Pabón (Netflix) continued by giving some insights into the dilemmas of film content localization. Needless to say that Netflix has nonconventional content as the norm. Pabón started her presentation with the motto: “Every sentence is a story,” and illustrated this with an example of localizing the title of the new Netflix series Orange Is the New Black into different languages. At the end of the day, these five words cost $1.40 per word due to the complexities of transferring the meaning into different cultures. Traditional measurements for productivity and quality simply don’t make sense when localizing this type of content. What’s more, content types at Netflix encompass hundreds of genres, thousands of show and movie titles, taglines, synopses and marketing content, each having its own challenges when it comes to localization. As for quality evaluation, some content types are still evaluated using traditional metrics (scorecards and quality reports), but nonconventional methods such as marketing campaign success or social impact on Facebook or Twitter, are also gaining importance.
The closing talk of the session was given by Mike Dillinger from LinkedIn. The company has 400 million members around the globe and localizes into 24 languages on a daily basis. Dillinger explained that LinkedIn covers both conventional and nonconventional content. For certain content types such as search queries, support chat sessions, social media and the like, machine translation (MT) is used even to the extent of raw MT. This is simply because there is no way a human translator could provide translations for the incredible volumes available. To still fulfill expectations of customers on delivered quality, Dillinger suggests that any agreed-upon specifications remain very important. Product, process and project parameters are of key importance and need to be clear to all participants of the localization workflow. Providing the right quality results in increased traffic, engagement and, of course, revenue. While products, services and content types are changing, these variables correlate well with traditional metrics.
As I mentioned earlier, startups are by definition nonconventional and agile. Agility is their core element. It’s in their DNA while they’re in quest of a raison d’être — incessantly reinventing themselves. Starting from scratch, making up something new that will end up in the dustbin — failing and starting over. An ode to all startups in the world!
In the ever-changing environment where startups operate, agile methodology is the only way to move forward. Agile localization will also have its challenges. And vendors working with startups know that an agile environment requires internal retooling to handle the steady stream of smaller projects that correspond to the client’s sprint. Traditional tools (emails, spreadsheets) can’t sustain the traffic of agile projects. Cloud-based translation tools and translation management systems are becoming the norm. Also important for agile localization is a good relationship between the buyer and the vendor. There are multiple benefits for both parties: higher quality, lower cost and quicker turnaround, as well as greater subject matter expertise among the translators. For continuous publishing and localization, you need to build relationships on trust and transparency.
Agile localization should follow the scrum model with approximately two-week cycles, where each cycle should produce an outcome that is measurable. This is opposed to the traditional “waterfall model” in which localized products are delivered after a long development and testing period. In agile localization, testing and evaluation of localized versions should be done in parallel, on the fly. In a startup and agile scenario, the traditional notion of quality is challenged anyway. As we have learned from the different content profiling pilots conducted by Dublin City University and TAUS, there is no one-size-fits-all quality because text type, purpose and target audience differ largely in various localization scenarios.
Your processes and tools need to fulfill the requirements of agile localization or otherwise you will lose time, in which case you’d be better off going back to the old waterfall model. The key, as I mentioned earlier, is measuring efficiency. Is there a good balance between productivity and quality? Even more than in established companies, efficiency and the constant monitoring of business intelligence data become extremely important.
The datafication of the translation industry is an ongoing trend that has been discussed extensively at conferences and industry events. Data and metrics have been collected for some time now and there is a considerable amount of data to be tapped into. However, there is currently no straightforward, generalized way to convert raw data into meaningful and useful information for business analysis purposes. As long as there is no transparency or agreement in the measurements, results may be subpar and lead to frustration and misunderstandings.
In order for data to be useful, it needs to be annotated using commonly acknowledged metrics. I have intentionally avoided the term standards, because metrics don’t have to reach that status to be useful. A wide agreement about them is enough. One example of such metrics is the harmonized DQF-MQM error-typology intended for error review. The typology has been developed in the European Union subsidy project QT21 and is being implemented in the TAUS Quality Dashboard along with other metrics. The goal is to seamlessly integrate quality and productivity measurement in a number of computer-assisted translation tools and enable reviewers to go a step further. Rather than just correcting or counting errors, assign error labels from a two-level hierarchy and specify severities on a preselected sample.
Startups are in a prime position to try and test new metrics and apply the best fit for their purpose, driving industry adoption. Of course, error typology might not be the best solution for their content, just as we have seen in the examples for nonconventional content. But there are so many evaluation methods out there, including usability, readability, adequacy, fluency, rank comparison, productivity — the sky’s the limit.
Localization is a data-rich industry. How can we leverage this rich seam of data to help make better decisions in our localization projects? First of all, data should offer real solutions focusing on how to target customers, provide real benefits and drive a better service.
Secondly, there is plenty of data available in our industry, but not all data is good or relevant to everyone. Before sharing business intelligence data, it is also important to verify what the impact will be when sharing this information: negative data doesn’t always bring us further. Positive data can be a good thing if it is shared because it can have a virtuous influence on the enterprise.
A new market segment
TAUS predicts a growing market for translation and consulting services targeting startups. These ventures are desperate for advice when jumping on the localization bandwagon. This is simply because their growth potential is huge, but they are starting from scratch and they need a knowledgeable companion to avoid mistakes down the road. There are many technologies to choose from, many approaches to localization and many tools to measure return on investment. One of the key elements for growth is localization readiness at an early stage. Appealing content should be created fast with a multilingual purpose in mind. Ideally, the localization should be outsourced to a vendor or multiple vendors that provide scalable solutions and different levels of quality.
Startups need a partner that can help in getting the word out there and support them in their early endeavors. I’m sure you can have it all in one: a consultancy with access to business intelligence tools, that can set up agile localization, speak the lingo of vendors and is familiar with the technology out there. With a mighty companion like that, achieving efficiency for startups should be a piece of cake.