The basic idea is   put in a phrase and see how Google’s translation tool butchers it… or will it

I borrowed TechChrunch’s example below!

Have fun and let me know your scores!

If you are one of these organizations, please know that you are not alone. However, finding out how much your company spends can be extremely difficult, depending largely on how localization is currently managed. Most organizations who struggle with this question run localization in a highly de-centralized manner. Each region or business unit is responsible for managing the cost of their translation needs, and there is little documented process or symmetry in how each performs the task. This blog will focus on these types of organizations, since a majority of them are the ones struggling to find out their annual spend. For example, one region may send work out to a localization vendor that is contracted, while another may request translation ad-hoc when the need arises. Still further, some regions may have in-house staff perform translation as part of their normal duties, so not to incur any cost for their translation needs. If this sounds like your company, then here are three things you can do to go about collecting an idea of the “spend” your organization does as a whole.

First, ask the regions/divisions/business units how they are currently handling translation. Find the main translation stakeholder in each and ask them for their process. Specifically, ask them:

• How do they bill for localization?
• Do they have written processes?
• Do they maintain existing contracts with translation vendors?
• Do they use a specific code on POs for translation?
• Do they use existing staff for translation, or are their specific FTEs designated?
• Do they maintain any assets?

What these questions will help you discover is how each group operates, and where to start looking for the money they spend on translation. In some cases, you might find a relatively well-managed process and determining the specific spend will be much easier. In others, especially where there are no processes, or dedicated FTE count the “spend” will be much more difficult to determine.

Second, look at the products that are from each group, and in what markets they are currently engaged with. Ask these questions:
• Do you translate written material for each market your products are in?
• What written material do you put in to each market? Do you have a document/page count of what each group in your division creates and sends to the markets you currently engage?
• If you are not translating content currently, do you have plans to start in the next FY, and if so, for what markets and which content.

What these questions will help you determine is where the translation is coming from for each group. This is helpful when a group has no documented process and/or uses existing FTEs for their translation services. Additionally, it gives you an idea of what content is going out the door to end-users in a localized format.

Finally, you can just do the math. While each company’s localization “spend” will vary significantly there are constants in localization that can be used to determine a rough idea of localization “spend,” such as:
• Typical translators can translate 1500-2000 words per day, not including leverage
• Quality Review is normal accomplished at a rate of 10,000 words per day
• Desktop Publishing is normal performed at 5-6 pages per hour
• Web content normally has 200-250 words per page
• User documents average 250 words per page

If you have done your due diligence in the first two steps, you have an idea of what documents were translated in which groups. Knowing this will let you find the amount of pages and the number of languages for each. For groups that use existing FTE count to translate content on the side, the first two metrics will give you and idea of how much time is truly spent translating content. Now you have a better idea of what your company translates, ask the localization vendors for the average cost to localize each language on a per word basis. Put the numbers together and you have a rough estimate to deliver to the executives. While this process is not the most efficient, it is effective in providing numbers and equally effective in showing the executive team members exactly how inefficient the localization process really is at your company.

A strategy based on individual translators and Single Language Vendors (SLVs) are most appropriate for organizations with minimal translation needs, limited language requirements or who have the wherewithal to create a large global localization vendor management group – essentially an internal MLV. Testing individual translators or SLVs is really about testing how well the individual or group of individual can take words and translate them into the language concerned. A traditional translation test of a few hundred to a few thousand words is appropriate here – as long as you have someone available to check the output!

Moving up to a localization strategy based on MLVs or using a complete outsourcing provider changes the game though. The common error most organizations make when testing these vendors is assuming that the quality of translation is most important thing you can be testing. Beyond testing actual translation quality it is important to assess how well the partner can translate simultaneously to 5, 10, 25 or 50 languages, with the most attention paid to how well their processes will stand up to the rigors of your schedule. Due diligence at this level goes far beyond whether the prospective partner can find someone to translate a few hundred words!

For MLVs and localization outsourcing providers, testing is most beneficial in a “real-world” style piloting a live project with a small number o pre-qualified vendors following due diligence. This live pilot would typically come near the end of the vendor selection process, after the short-list of viable candidates – the one (ideally – although in some cases the right approach is to pilot with a small number) who look like ideal partners based on the rest of the due-diligence. A pilot should be a real-life project for both organizations and should be representative of how your organization works on a day to day basis as well as how you want to partner over the long haul. Specifically, Pilot approach must address these questions:

• Does this vendor have documented repeatable processes for delivering content in a consistent, high quality format?
• Can this vendor scale to meet the overall needs with their current processes?
• Can this vendor make “value-add” recommendations in the context of the PLP to help optimize the overall localization spend?
• If there is a technology aspect, can the vendor properly assess the best technology and implement it seamlessly into an organization?

At the end of the project, each vendor should present a retrospective review, including concerns, lessons learned and potential gains from best practices. These “post-mortem” reviews are where the key differentiators between vendors will come to the surface.  A company will have the data necessary to answer the four questions above, as well as many others allowing a controlled and educated go/no go decision to be made.

The Pilot approach is the only way to truly gauge the strength of a vendor and their ability to meet your needs.  Typically it takes longer to complete, but it allows the vendor to demonstrate more of their specific capabilities and allows both parties to determine if they are a good fit.

The problems for these organizations are localizing content for the target audience and the intention of the survey or research getting “lost in translation.” This is because the nuances of translation are far-ranging. A literal word in one language, for example, may have no equivalent in another language, or could have a completely different “meaning” or effect in the translated language. This is why translation is an part-art not all science. No literal translation can be expected to convey perfectly the “meaning” of what consumers meant to convey in their own language. All the good work of a focus group moderator in not “interpreting” verbatim comments can be wiped out by a careless translator.

Back translation is instituted by organizations to improve the reliability and validity of research in different languages by requiring that the quality of a translation is verified by an independent translator translating back into the original language. Original and back translated documents can then be compared. If there are discrepancies, then the translated content has to be re-reviewed and necessary changes are put in to the document.

Due to its high cost, back translation is not overly common, but in organizations where there is a high risk-high reward component to target market success it is considered another “cost of doing business.”

The question is, “Is there a more efficient way of managing these risks than have back translation performed?” To answer the question, we need to understand the main driving point of the back translation, which is to ensure the quality of the initial translation.

Are their alternatives to back translation?

Translation quality can be impacted by a number of factors.

First, the content you are creating has to be managed properly. Authoring is essential. Especially in fields where there is heavy regulatory compliance, such as the pharmaceutical or biotech industries where the penalties for failure can be far greater than monetary costs.

For heavily regulated industries the assumption is that all source content created has enough rules around it from the compliance requirements, so additional scrutiny is unnecessary. That is far from true. An overall translation and localization management strategy has to take every part of the supply chain in to consideration and incorporate active practices for managing each of them. Source content creation (Authoring) is the first step in localization, and mistakes here can exponentially drive up costs as the process continues.

The second factor is asset management. Global organizations have long used Translation Memories to store previously translated content for re-use. However, it is more than just having assets behind your translation. It is about managing your assets intelligently and incorporating their care and upkeep as part of your overall strategy.

For TMs, that means more than just using them. An organization should invest time in a TM strategy that includes the maintenance, collection and “scrubbing” of all language assets on a regular basis. Whether you use “in-house” translators, freelancers or language service companies, you must have a strategy.

However, just as important as TMs are style and terminology guides for the target languages. Earlier in the blog, I mentioned there are times when a literal word in one language does not have an equivalent in the target language. Even worse, literal translation of a word or product name has negative connotations in the target language. A perfect example was the Chevy “Nova” when it was sold in Spanish markets.

Style guides and terminology have to be part of a asset management strategy to works in conjunction with you production (translators) strategy.

A third factor is the training and skills of the initial translator. The translator needs to not only speak English and the target language fluently, but implicitly understand the target market. The best way to ensure this is by having in-country, native speakers handle the initial translation of the content. No argument can be made that a person residing anywhere but in the target market can localize properly.

Finally, having a quality control process that ensure each of these three items is fully checked will further ensure the quality, accuracy and consistency of localized content.

Incorporating all four of these into a single over-arching strategy for approaching local markets will provide better, more cost-effective results than hiring “independent” translators to back translate content for review. When hiring an independent translator to back translate your content, at a minimum you have doubled the time and cost of your translation. The actual translation or content is the most expensive part of the localization process by far. It is also the most time-consuming part. Putting in a corporate strategy, opposed to using back translation will drive efficiencies in reducing cost, improving quality and speeding up time-to-market. More importantly though, it will guarantee that those surveyed in the local market are having the information provided to them with all appropriate local nuance, thus driving up their comprehension and minimizing their risks.

Conclusion

Back translation was created in the absence of a full localization strategy. Many compliance organizations, such as ISPOR (International Society for Pharmacoeconomics and Outcomes Research) instituted back translations as a requirement in the mid-to-late 1990’s. Here we are almost 15 years later and it has not changed. However, the localization industry has grown exponentially more advanced and in both technology and the understanding of local market nuances.

As professionals in the localization world we tend to deal with the top five languages most of the time, the top 50 some of the time and occasionally the top 100. Languages outside that are only tackled on an occasional basis. Dhivehi falls into that latter category (I believe we only have a couple of other clients with that language requirement) but there is another category of languages as well – those are the so called Endangered Languages. Estimates of how languages are in danger of extinction vary but the commonly accepted number is between 2500 and 6000 depending on which expert you ask. Beyond that there is a much less talked about category of languages that aren’t endangered but also aren’t commercially or globally viable.

Superhero

K. David Harrison is Assistant Professor of Linguistics at Swarthmore College but as one of the founders of Living Tongues perhaps he is better known as the world’s superhero for endangered languages: He has written books, has appeared in almost every newspaper in the world, and has even reached the pinnacle of media – he was on The Colbert Report - which would be compulsory TV in my house if we had cable! His most recent book is a bit of a mouthful “When Languages Die: The Extinction of the World’s Languages and the Erosion of Human Knowledge” doesn’t concentrate on proving that languages are dieing - that battle has been won and most researchers agree that outside the top languages  there is a steady and rapid decline in language use. Harrison concentrates on my favorite question – “So what, who cares?”. This isn’t a translation or a localization book and even goes beyond anthropology. It is a profile of humanity.

So what, who cares

Harrison has spent many years working with communities whose languages are dieing or have died. The Tuvans (as obsessed over by Richard Feynman) have a vibrant culture but their language is dying – being replaced by Russian. Using the Tuvans and a hundred other examples, Harrison helps us understand why languages are vanishing (urbanization rather than design) and what we lose when those languages disappear.

Back to Dhivehi (and Welsh)

Obviously Dhivehi is nowhere near endangered status – after all, it is the official language of a Sovereign State (the Maldives). That said, is is not a commercially viable language for most translation companies because not enough people outside the Maldives want to translate to or from it. So the question I asked was what role does the localization industry play in supporting the many thousands of languages, like Dhivehi, that lie in-between commercial/global and endangered. As a natural born Welshman, my own country’s language once fell (and arguably still does fall) into this category so perhaps this is personal.

As a given languages becomes less and less popular it moves towards the tipping point – the point at which it becomes impossible to find people willing to base a career in translating back and fore between it and the more commercial languages. Vendors, especially the larger ones, unwittingly play a role here - they are the connectors between clients with need and the ever-decreasing number of translators to that language.

Don’t get me wrong: I am not ascribing a noble nature to the language service business. I certainly don’t think the big-thinkers at the larger vendors (myself included, I guess) spend much time thinking about this type of thing… but perhaps we should.

This year there were 37 countries represented on the G500 up from 35 in 2008 and 33 in 2007. Hungary, Israel and Venezuela are all newcomers. The EU and US dominate of course (163 and 140 respectively) but Japan, China, Switzerland, Canada and South Korea are all in the top ten. As Zia Daniell Wigder of Forrester put it in her news item

While there are still twice as many Fortune 500 companies based in the US as in the next country on the list (Japan), the number of US companies declined from 162 in 2007 to 140 in 2009. The number of Fortune 500 companies in Britain, Canada and the Netherlands also fell. By contrast, all other countries in the top 15 saw their numbers grow or remain flat. Out of the top 15, Russia, China and Spain showed the highest percentage growth in the number of companies listed.

This shows that although there is still a strong bias towards the EU and US, there is opportunities for companies wherever they are. This all looks good for localization – as more countries baptize global companies their populations will be more willing to accept brands created elsewhere – which means more localization requirements. Even in the depth of a recession many of the businesses here continue to grow – which again is good news both for the global economy and for us.

Question: So does Google Tool Kit spell doom for the localization industry?

I think my answer to Nic (and everyone else) is, “don’t panic and don’t be afraid” – much as it was last week when everyone thought the world of professional translation was dead due to crowdsourcing. In the right place the toolkit might be useful – but a systematic replacement for professional translators it isn’t.

Renato Beninatto had a pretty measured response in his own blog earlier this week.

Google Translator Toolkit is not perfect. There are valid concerns about using it, along with the predictable resistance to change by those tied to the existing model. However, Google has already changed our behavior in the way we look for information. Now, it is launching a platform that has the potential to revolutionize the translation process, especially if combined with Google Wave, which is expected to be launched soon.

Google has done a great job of making the world rethink its approach to information and if they continue to develop the workbench and resolve some of the not-insignificant issues around the intellectual property of TMs it might someday be a force to be reckoned with.

Of course, for a generalist to work better the translation management process should include at least a reference glossary, a specialist reviewer and a minimum review process to make sure that the final translation is completely accurate in terms of subject matter.

The problem is that many translation agencies or companies do not consider investing in reviewing a translation worth it, since they usually base their competitiveness only on the quick turnaround/deadlines and profitability/prices.

I couldn’t agree more with Ivete here. Translation companies have to take responsibility not just for the project management of localization but for ensuring that translators have everything they need to perform their tasks.

Most of my work with is with global companies who routinely translate our to 10, 25, 50 or more languages. The problems Ivete highlights are exacerbated in global-scale localization where the same terms are to be translated by so many translator teams spread around the world. For this type of localization the translation company should be providing complete documentation on all the terms and concepts for each of the translators involved along with all the other supporting materials. If they don’t, the likely result is inconsistency between languages since each translator will need to perform their own research and come to their own conclusions.

Obviously there are software products that can help – SDL’s MultiTerm is a good example but for anyone who is in the process of assessing vendors – especially if the candidate pool includes some of the smaller players – I always recommend questioning the capabilities for central research and distribution.

I will certainly be looking out for Ivete’s next Sacred Cows entry!

Companies in advanced economies were shocked to see suppliers in Asia and elsewhere simply disappear as orders from abroad contracted. In China some 67,000 factories went bankrupt, according to that country’s authorities, in the first half of 2008—even before the U.S. financial crisis sent the global economy into a tailspin.

Outsourcing and subcontracting are rife in localization at a number of levels.

Typical practice for global scale MLV’s is to subcontract to SLV’s on a language-by-language or region-by-region basis. The problem for the company procuring the localization services is that they are disconnected from those SLVs and have no way to predict the stability of the companies that are actually going to be performing the work. Procurement departments and selection committees will often spend a great deal of time  reviewing financial stability of a potential localization partner – work which is rendered useless if the actual production is performed by a completely different organization.

I would recommend anyone tackling a vendor consolidation to start begin asking questions that will identify stability issues not just at the MLV / master contractor level but at each point in the production supply chain. Here are some key questions:

• request third-party validation of each sub-contractors financial health and its sales or operational history,
• ask to see references from other or past partners,
• ask to see ongoing processes for tracking sub-contractor risks (a buyer, for example, should want to see a map of the partner’s supply chain including contingency), and
• develop backup options for each important sub-contractor.

Having confidence in a trading partner means having confidence that the partner knows what to do in the event of a breakdown in their subcontractor supply chain. Where subcontractors are part of the equation, as Josh says,

These practices will give both sides greater confidence that a crisis won’t wipe out their trading partners and leave them scrambling to find new ones.

How to define the total cost of ownership for localization?

I define Total Cost of Ownership, or TCO, as the complete cost of translation based on the current strategy weighted by ratio’s quantifying overall quality and time-to-market targets. This can be expressed in a formula as shown below.

TCO=(c×qr×tr)

Total cost (c) multiplied by quality ratio (qr) multiplied by time-to-market ratio (tr). Once you know the total ownership you can begin to understand the efficiency (E) of your translation environment, and make decisions on how to make improvements. To caluclate this simply divide TCO by the total number of translated words (w):

E=TCO÷w           or           E=(c×qr×tr)÷w

We know how to calculate TCO and the efficiency of the environment, however with localization, to quote that old saw, the “devil is in the detail”; each of the component variables needs to be carefully and consistently calculated. This is not an easy task of we are to get an accurate answer for either TCO or E.

Calculating Total Cost (c)

c (cost) is the total cost of the strategy including internal, external (vendor), technology and overhead. This can be tricky since most organizations don’t track these items under an umbrella category (or even track them at all). Sometimes it can be easier to use cv (external spend to vendors) but this will provide misleading assumptions.

Calculating the quality ratio (qr)

qr (a single number representing quality for the time period being measured) you need to have a formal measurement of quality. There are a number of ways to measure quality however I favor a simple count of agreed “bugs” with severity weighting, measured against agreed quarterly target bands. This is shown in the chart where we can easily see that qr for the third quarter would be measured at 5 points since quality was rated “poor”. The heavy lifting here is in agreeing how to quantify, agree and weight each individual “bug” taking a pragmatic view of the preferential nature of translation without losing track of accuracy. There are a number of standards available that can be used or modified including SAE J2450 and a proprietary standards such as the one available from LISA but the amount of effort and the need for pragmatism here shouldn’t be underestimated.

Calculating the time ratio (tr)

tr (a single number representing time-to-market – the time taken from deciding to translate until the content is authorized for delivery) you simply need to choose when to start the count for any given translation job, when to end it and whether to count weekends! Weighting is very important since being one day late on a three month project is very different from being one day late on a security patch that has a 24 hour deadline. Calculating the number of words w (words) is a count of the net number of words translated including 100% matches and machine translation output. If you are using automated environments such as a Translation Management System or have a well-structured vendor strategy this should be fairly easy – otherwise quantifying the total number of words will take a great deal of effort.

Why go to all this effort?

Typically time, cost and quality are the three Key Performance Indicators that are measured in localization, however to be useful outside of simple benchmarking those KPI’s can be used to model and measure strategic and tactical changes to the environment. Our research and experience show that measured steps to reduce TCO and increase overall efficiency have a significant benefit over and above a simple-minded focus on driving costs down.