louisrosenfeld.com logotype

Home > Bloug Archive

Mar 23, 2004: Multilingual, Multinational Information Architecture Design

Have you worked on a multilingual, multinational web site's information architecture? If so, there's a future for you; designing such architectures is one of the two frontiers of IA design, along with (naturally) enterprise IA.

And if you've worked on multilingual, multinational, enterprise architectures, well then, hats off to you. You've not only come up with ways to make those enterprise silos work together as one information environment, but you've made it happen with content owners and user audiences that literally don't speak the same language.

I'm working with a Fortune 500 client that's already gotten off to an excellent start on cracking the enterprise nut. But they operate in dozens of countries and do business in enough languages to make the folks at Berlitz drool. I've been charged with kicking off the research on what's been variously called internationalization, globalization, and localization. (Or should it be localisation?)

I'm hoping we can gather some collective notes on developing a multilingual, multinational enterprise information architecture right here on Bloug. It'd be wonderful to compile a list of the issues and challenges involved in designing such an architecture; I've started one below. Care to add to it?

Also, if anyone has any pointers to ideas on what might constitute variant models of international information architectures, might as well list them here too.

OK, here goes: the start of a list of issues and challenges for multilingual, multinational information architecture design:

  • Languages and countries don't necessarily correspond one-to-one. So an architecture may have to concurrently support languages spoken in many countries (e.g., Portuguese), one-to-one language/country pairs (e.g., Japan/Japanese), and languages spoken in a part of a country (e.g., Tagalog). Conversely, an architecture may have to support countries which speak multiple languages that are also spoken elsewhere (e.g., Switzerland, Hong Kong). These issues mean that an architecture may have to support a heck of a lot of different combinations of language and country.
  • Can you simply translate an information architecture? Short English navigation labels might be menu-busters when translated into German. Even if semantic translation works, structural problems may come up; for example, alphabetized architectural components, like an A-Z index, may not make sense when translated into a non-alphabetic language like Chinese.
  • An organization may have varying degrees of presence in different countries, ranging from selling its fullest array of its products and services to no presence at all. For example, FedEx maintains very different sites for Canada and San Marino. Can these different levels of presence be sufficiently anticipated to support a few template-based designs that accommodate all possibilities?
  • Can we assume that users in all countries expect similar degrees of service from their respective localized sites? Face-to-face customer service values surely vary substantially even regionally; I imagine expectations regarding Web-based services vary similarly. What kind of user research should information architects conduct to establish and design for these cultural differences?
  • And can our budgets support the level of research necessary to design these multilingual, multinational architectures? For example, might we find South Africans and Pakistanis to be search-dominant, while Turks and Singaporeans prefer to browse? Assuming it's a bit pricey to find stuff like this out for 150-200 countries, can we hope to make reasonable guesses without the data?

OK, that's a start. What would you add?

And, um, it should go without saying that the answers to these questions are very much of the "it depends" variety.

email this entry

Comment: Prentiss Riddle (Mar 23, 2004)

Boy is this a complicated and interesting topic! I've passed your message on to a couple of internationalization/localization pros of my acquaintance who may have something to contribute.

The many-to-one/one-to-one/one-to-many language-to-country model doesn't encompass all the possibilities. A friend tells me that his Fortune 500 company maintains multiple Spanish versions of its sites because of basic vocabulary differences among Spanish-speaking countries. The decision to do so is political and culturally nuanced: not all Spanish speakers agree on the importance of catering to particular varieties of the language, and it's hard to get hard data to support a particular decision. The process is also informed by budgetary constraints, as you suggest. Some Spanish-speaking markets may justify a full-blown localization of their own, while others may have to make do with the Spanish of a larger neighbor.

And then there's the even more ticklish question of how to support Spanish for the *US* market, where Spanish may or may not be considered more important as a marketing gesture than as an actual communications medium (just ask the Republican and Democratic political candidates).

Comment: Ken Westin (Mar 23, 2004)

-Depending on the language you are working with you may need to think about text goind right to left vs. left to right.

-Careful with icon usage, as they are generally metaphorical and may not be universally understood across cultures

-Decide if you are going localization (L10N) or internationalization ( L18N), there is a big difference particularly in the budget area. Do you have the resources to adapt each brand/site to a particular region and culture or do you want to generalize the presence as much as possible?

- Careful with the translation of any tag lines in logos

many many many more

Comment: Haydn Shaughnessy (Mar 24, 2004)

Multilingual I'm dealing with right now for European anti-discrimination
bodies. The language issues in this discreet group came down to selecting
two that everybody could live with or using one's own language plus one that
most people could deal with - no prizes for guessing which one.

Terminology management systems help manage recurrent phrases if the
information has a high repeat level and absolute accuracy are needed. If
that's a route you need to go get back to me and I will check back over my

Check also the EU's information sites: www.europa.eu.int - all in 11
languages. They very much run off templates that all 11 language versions
comply with - templates are essential.

Culture in my experience is an entirely different issue to language and the
research base is different. Again European specialists offer some hope - do
you follow Fons Trompenaars work? One of the points now emerging from
cultural research and experience is that ethical considerations are really
important. If you expect to be able to shortsell in one culture - watch out.

Comment: Haydn Shaughnessy (Mar 24, 2004)

I took this from a piece on terminology management - describes a product called TRADOS: www.translationzone.com; www.trados.com

Well-managed and up-to-date multi-lingual glossaries of terms are essential to ensure the consistency of key terms, i.e. terms that are of significance to the reader, of importance to the brand owner, and that are frequently used in communications materials and content. Well-maintained glossaries also ensure that translation memories, key to the translation of high volumes of content, can be effectively cleaned and updated.

A database is built for an individual client and structured to capture all the attributes of a term necessary to track the evolution, application and usage of each term.

In our experience TMs are most effective on:
· Content that has been amended slightly e.g. the change of a product name, or a specification or the insertion of a few new words (e.g. updating web templates).
· Content, which regurgitates paragraphs or sentences that have been used elsewhere (e.g. web summaries, abstracts, synopses).
· Content sourced consistently from one master language (i.e. the direction of the translation is consistent)
· Major updates of regularly issued high volume documents (e.g. manuals, guides, policy documents).

Comment: Stig Andersen (Mar 24, 2004)

It would be easier to discuss with a bit more information. I understand there are issues about confidentiality, but perhaps we shoot in all the wrong and not so constructive directions. Hints: Is it public sites or intra-/extranet? Main objectives; branding, e-commerce, costumer service?

I've some experiences a couple of cross-country/multicultural projects. Probably not so big. Here's my hints:

The issues you mention are good. An obvious thing to remember too are, that there are great cultural differences across regions. This surprised us in terms of humor, what people associate with colors etc. and – as already mentioned – iconography.

Regarding forms, you should remember, that in some regions people don't use ”Mr.” and ”Mrs.” (in translation of cause). Not to mention five digit ZIP codes. Also ”State” and ”Region” does not apply in many regions. The list goes on...

We were surprised to learn, that in Denmark users preferred few steps with more questions per screen in wizards. In other regions users preferred relatively more screens with fewer questions per screen.


Comment: Lou (Mar 24, 2004)

Keep'em coming, folks; thanks!

Stig, it's a public site with lots invested into strengthening its global brand, and a burning desire to expand into new markets (think two or three large and highly populated that dominate the Eurasian land mass). They need to give their local sales agents significant flexibility to push their products, but still have to conform to at least minimal brand and IA standards.

Yeah, I know, that sounds like any number of corporations, but it's the best I can do. And so far, no one's shooting in the wrong direction. :-)

Comment: PeterV (Mar 24, 2004)

Interesting stuff. The local IA that organizes information will probably differ significantly from the international one for two reasons. One, because the information to be organized is probably at least partially different (we don't sell exactly the same products in different markets). The other factor that could influence local IA's being different might be that local markets (even for the same products) are different, therefore users' expectations are different.

An example: look at how car websites work in the US versus the UK. The classification is drastically different because the markets are different, and the markets are different because the users are different (or the other way round, doesn't matter). (sorry for the plug, remove it if you want, here goes: this example is in my book)

I don't think investigating cultural differences like "search dominant", however interesting, would be worth your research dollars right now. I'd invest your local research dollars on teasing out ethical issues and differences in markets/users that'll change the taxonomies used.

And buy Trompenaars' books - they're fun to read and very informative.

Hope that helped a bit.

Comment: victor (Mar 24, 2004)

This is a great topic, and I think we're at the point where we've learned a critical mass of lessons and start to put stakes in the ground.

You might want to check the wiki:

I've found that small UI decisions that are to be replicated internationally can make all the difference between success and failure. For example, it's crazy hard to properly display different characters - especially double-byte - in pull-down menus. The lesson is to prototype any potential solutions.

Comment: haydn (Mar 24, 2004)

You may want to talk/write a friend of mine who does a lot of assignments in inter-cultural work. Mijnd Huijser www.businessdiplomacy.biz. He works with Trompenaars. Mijnd and I are working up an offer on culture in electronic networks.

Comment: Walter Underwood (Mar 30, 2004)

Some portions of your taxonomy may be shaped by local regulations. The classic example is "401(k)", which is widely understood in the US, but useless in other (English-speaking) countries, because it is a section of the US tax code. You may want standard and localizable regions of the hierarchy.

One strong suggestion: use UTF-8 for all pages, no exceptions. It is possible to use multiple encodings, but it makes your QA just ridiculous. Be assured that someone will want to use their local encoding because some font is only available for that encoding. Tell them "no".

Note: "Tahoma" is a Unicode font, but it is Microsoft-only. Don't depend on it.

Comment: Noreen Whysel (Mar 30, 2004)

Another thing to watch out for is backend processes that utilize the same tool or database across the global enterprise - what data is collected? How does the localized site collect the data? How is it processed for global use (marketing, lead management, strategy, etc.)? How is it processed for local use (transaction fulfillment, inquiry follow up, referrals, FAQs). Some of the areas where I have faced multilanguage puzzles include feedback/webmail forms, transaction processing, search and the metrics culled from these processes. It is a big issue and an interesting one at that. I'm looking forward to hearing more on the topic.

Comment: Guilherme Szundy (Apr 4, 2004)

As long as IA is concerned there is another issue other then just language and culture differences.

How will navigation take place between the various localized versions? Will you have many parallel versions of the site, which are only a replication of an structure for each locale or will there be crossed information. For example, if a visitor is looking at a product detail in English can he/she go from there to the same information in French or he/she will have to go the French site homepage and find the information from there?

Another important issue is maintenance. Will it be centralized of spread across the globe? How information will be produced and translated?

Comment: Livia Labate (Apr 13, 2004)

Designing multilingual, multinational information architectures is the natural complication of an ever-expanding global economy. The business world has much to teach us about this topic. It is far from being a novelty, it has just been poorly addressed, specially when it comes to interaction design.

First, there is the difference between multinational (involving more than two nations) and transnational (extending or going beyond national borders). Here lies one of the origins of much grief in discussions about i18n and l10n. My perspective is that *a multinational information architecture product is the result of a transnational information architecture effort.*

To go beyond national borders means having a i18n vision and shaping your design to reflect that desire. It relates to strategy and direction. Having a multinational solution, means you were capable to enforce that vision and create adequate solutions for each of the locales [1] you intended to serve. This is the successful result of i18n/l10n in my opinion. In translation lingo globalization = internationalization + localization.

I had written a good deal in response to the original post, mentioning nitty-gritty topics like (mis)use of flag-icons to represent locales and the labeling of links in the target language, but it is easy to drift off when you start to talk about what can go wrong and what you should not do when addressing i18n/l10n. I think this approach is what makes the topic so hard to deal with. It is sort of the international IA equivalent to 'standard usability rules', it limits your perspectives and keeps you from asking the real questions.

Back to Lou's original question, if you're going to tackle a business model that already operates in various countries and languages (external business perspective), this whole thing could very well be called Transcultural Information Architecture (this way there is no mistake about borders or languages being limiting factors). If you're going to integrate the aspects pertinent to Enterprise Information Architecture across nations and languages where the company operates (internal business perspective), maybe we should call it Transnational EIA or something like it (Transcultural EIA wouldn't make sense as it would mix with notions of Corporate Culture -- note: I'm talking about a company's mission and practices, there will be obvious differences in management and organizational structure in different environments).

I wonder if it's harder to try and understand them separately or in conjunction (by the way, I'm only giving them distinct names because I can't talk about different things calling both 'IA'). Since I'm not focused on an EIA per se, I'd keep them separated, but the differences shouldn't be many. The main dissemblance should be how you understand the vision: you're providing an information solution for external or internal agents?

For both the responses I feel that the key is to have a strong and consistent vision (where do you want to go) to drive the initiative (getting the work done), not the other way around. You don't set out to design localized versions of a Web site, you set out with a internationalization goal (i.e: reducing translation costs for product collaterals worldwide). The localization concerns are objectives which make up that goal (i.e: how each locale will develop their product collaterals).

And if I get any more abstract I'll change my name to Picasso, so...

The operational concerns of creating i18n and l10n solutions are many - great suggestions so far, I loved reading this whole discussion so far. I would just like to emphasize how you don't have l10n without thinking of i18n first, otherwise you're building separate solutions for each locale (or providing translated content in an inadequate structure). Again, in translation lingo g11n = i18n + l10n.

I18n is not just setting a vision though, there are core issues to address like addressing technology standards - languages (the computing, not human kind), database models, and most of the things linked to technology which you will want to have in place throughout your solutions.

When dealing with multiple languages, controlled vocabularies are great friends (specially in addressing search), not to mention rich metadata structures. And when you need different characters sets, a database structure which can understand and translate them interchangeably will prevent many headaches. Machine Translation and Translation Memory are also very important for proper reuse and will keep your Content Management System finally tuned (assuming it was built with this in mind - yet another thing to think about).

Those are language-related, it's when it comes to culture and behavior that hell breaks lose. Stig mentioned great points and when you get to l10n (after you got your i18n strategy in place) you'll see that it is easier to treat the locales as 'regular' user groups, since you won't be trying to develop something for 300 different personas (every user group in every locale, at once). More than this, it allows you to better manage costs, because you will be able to develop one locale at a time, or hire a different team to work on each local solution, etc... Doing things in this order (i18n -> l10n) also provides a very powerful and rich feedback-loop for business intelligence: the more you know about each locale, stronger is your global goal attainment (globalization).

In the end it is a big company-wide balancing act.

My wish list for this topic:

* patterns for l10n (what have people being doing well?)
* deliverables for i18n strategies (how to enforce the vision?)


I was revisiting www.interactionary.com/files/disciplines_radial.gif and noticed it has been updated on the 2nd edition of 'IA for the WWW'. Maybe now it is time to add 'Translation and Versioning' as well as 'Diplomacy' as the established areas the 'Web Era' professionals can learn from as both disciplines have explored the complexity of multilingual and multinational environments. Having a Foreign Trade specialization helped me as much (if not more) in addressing these topics than my background in Business Administration. Oh how interdisciplinary is IA :)


Finally (before I run out of bytes), if I had to select one book that would best describe the needs for information architects to comprehend the implications of culture and language, it would be Scott McCloud's 'Understanding Comics'.

Probably a surprising choice but it provides a very good grasp on the impacts of culture and language in how it shapes an experience. If you manage to trace the parallels between information architecture and sequential art, you can extract really valuable considerations form this book, particularly when it comes to 'Transcultural Information Architecture' (whatever you want to call it).

Thanks for the brain gymnastics Lou!

[1] Defining locales as audiences within a common cultural and linguistic boundary. More on that here http://louisrosenfeld.com/home/bloug_archive/000249.html

[2] Not even half-way through and I already invented an acronym.

Comment: John O'Gorman (Oct 13, 2004)

Great discussion and a very challenging topic. I have a few observations from past experience that I would like to offer in the context of an information architecture that I'm working on. The multi-lingual implications of the architecture are mitigated by the separation of objects into four domains: Audience, Product, Content, and Delivery. There are a number of techniques that enterprise architects can use to simplify the delivery of multi-lingual content.

Find peristent objectives. First, find out how audience objectives connect with your product. Determine whether these objectives are independent of locale. Most product users are trying to accomplish something that is common to users of your product around the world. If that's the case, the common objective can be used as the parent to connect to equivalent child content.

Next, determine what drives divergence and specializtion. If your product line has been adapted for use in every different locale then, while the objective is the same, the methods for achieving the objectives may all be different. You can still connect the common objective to the content, but the content for en_us is no longer equivalent to the content for a different locale. The source of variance has been identified as the product, and can be managed as the product changes.

Eliminate divergence. If your user objective is identical and the method for achieving the objective is identical, then the content is by definition equivalent in every way except one: language; thereby making translations renditions of an original. Think of PDF renditions of a FrameMaker document: it is equivalent in every way except format. The point here is that renditions have the simplest rules: if the original changes the renditions must change as well.

Manage the divergence you can't get rid of. If on the other hand, you have locale-specific objectives, and product-specific tasks that require custom content and the associated translations, then you are in for a whole whack of translation costs, BUT: having identified the sources of variance you can post some fairly solid relationships between changes in the objectives, changes in the product, and/or changes in the source content.

Notwithstanding many of the other issues brought up in the forum, if you at least identify the persistent and divergent characteristics of your audience, product, content, and delivery you'll be off to a ggod start.

Add a Comment:



URL (optional, but must include http://)

Required: Name, email, and comment.
Want to mention a linked URL? Include http:// before the address.
Want to include bold or italics? Sorry; just use *asterisks* instead.

DAYENU ); } else { // so comments are closed on this entry... print(<<< I_SAID_DAYENU
Comments are now closed for this entry.

Comment spam has forced me to close comment functionality for older entries. However, if you have something vital to add concerning this entry (or its associated comments), please email your sage insights to me (lou [at] louisrosenfeld dot com). I'll make sure your comments are added to the conversation. Sorry for the inconvenience.