Agile Localization: More Questions than Answers. Still.
By Ultan O'Broin-Oracle on Aug 22, 2011
The best reference I can find on the
interwebs is Tex Texin's (@textexin) document Agile
Localization Practices. Recommendations (October
2010). I exchanged some email with Tex about this piece before it
was published, so I should disclose I have more than a passing
interest in it. I would agree with Tex that the
localization industry appears to be reacting to the agile framework
adoption rather than help determining it. The evidence from the interwebs
is that LSPs have a stock response in the affirmative, “Yes we do
agile”, while being very low on 'how-to' specifics or showing any real thought leadership
Here are my own observations on the general seeming lack of 'hard' information.
One thing is very clear to me: If you development processes and tools aren't up to the job for agile-based operations, you will find out pretty quickly. Same for translation tools and processes. Whither the role of the crowd and disruptive innovation in agile localization.
Internationalization (I18n) is a sine qua non for any agile-based localization, not just the basics of character set handling, code conversion, multilingual tables in databases, and country, language, or regional-neutral code that can handle variabilized settings for currencies, dates, times, sort orders and the rest, but translatability. Make sure your product uses technology that is internationalized, but pay attention too to areas of composite messaging, string externalization, the building in of context, and of unique and persistent Ids on strings and content or other devices to aid reuse.
Agile has key advantages in the areas of shortening innovation cycles, exposing technical and project risk early, and shortening time-to-market for competitive content. Getting real, working content out there to a global audience in multiple languages, early, means a better global UX too. The idea behind the framework is that all functions are performed together within sprints, rather than preceding or bolting on particular functions to the entire process – usually the latter, waterfall style. Parts of the traditional localization (and UX) process are immediately pressured by this agility.
How is terminology developed (for source and target)? Using locked-down, frozen terms in advance of development is hardly agile, yet it's common. Terminology takes time to research and develop in a source language, and getting the language equivalents agreed and onto some kind of usable system isn't an overnight task either. Terminology is also going to change during sprints, in response to customer feedback, like it or not.
Possible solutions here are to start the agile process with the best terminology available, using possibly crowdsourced validation, and revise terminology as customer feedback comes in, in context from real user. Or allow users to personalize their own user experience. Why try to anticipate the ever-changing mindset of say, teen gamers in Korea when you can accommodate their opinions another way. An term submission and approval workflow and ability to reapply terms to code rather than a spreadsheet of opinion is the way to go here. A terminology-only sprint isn't agile.
Testing localized versions should be done in parallel. A build environment that can produce localized product or site versions automatically at regular intervals and publish them to customers for evaluation can be used. In the enterprise applications space, where complexity of functionality, and the need to integrate with other products, databases, and technology stacks, agile translation testing requires automation, coordination, and the setting of clear expectations as to what is in each build is a requirement, and isn't easy. However, a testing-only sprint ins't agile. That's the waterfall we all know and love so much.
Communications across all stakeholders in the localization process also needs to be considered. How is that done between development, internal client localization management, outsourced LSP project management, in-country translators, localization QA and linguistic experts working globally, across time zones? Does the daily stand up meeting scale in cyberspace? Video-conferencing? Plasma screens of daily objectives? Web-based reports? Scrum of scrums? Best of luck.
Some content is probably better suited to the agile localization process than others, and the range is broad: social media, web sites, mobile apps, games, anything where UX and language decisions are heavily dependent on user opinion or more complex UX-critical deliverables that don't require a ton of technical integration or the provision of dedicated build environments for each language version for sure. Content where terminology and style isn't mission critical works well with repositories full of stodgy content, but that's hardly competitive.
In some of these areas, the traditional notions of language quality are challenged anyway, and the decisions of the users are the deal-breakers on what works in the market, not professional linguists. Such an approach only works in the enterprise applications space if its moderated carefully. Some middle ground, based on iterative process of getting good enough fast enough would work, but not if you're implementation cycle takes months, and you're a public sector organization behind a firewalls with very strict demands on what can and cannot be said. Tough calls for product owners.
Consider, of course, localizing user assistance (PDF, online help, multimedia). Very easy to say DITA, topic-based content is the solution, but how does that really help? What of the information quality? How is that content rendered? How much translatability has been considered? And what of the budget for this? If users don't want documentation, fine, but how do you know? It may be required to help users configure their application and actually use it. And in some markets, translated doc may be a legal requirement. How fast can you write doc or create videos or demos, and localize them to go with the software?
As for acceptance criteria and review meetings, well, how is feedback from global users and localization teams accommodated? What is the process for improving translatability and translation tools? How are development processes influences to improve the process next time around? How is the feedback from international customers made to support, to development, to product development? What about real-time feedback from users? Bug databases online? Focus groups? Private Twitter stream or Blackberry IMs with real users? Does your organization have any track record of doing any of this really well, anyway? If not, why would agile be any different?
Finally, why assume the source code and content is in English? How would an agile framework work for a Japanese game localized into English (or Korean for example). Do tools and processes work for those scenarios?
I don't know of any organization providing meaningful agile or scrum training aimed at integrating a full localization process into a product life-cycle. If you know of any, let me know using the comments. Other observations on agile and scrum are welcome too. Lots of questions, not too many answers. “Yes, we can” is a fine declaration, but let's hear some “Here's how we can do it...” statements too!