When HealthCare.gov debuted in October of 2013, the fact that the website interface was terrible became a hot topic. The fact that so much money had been spent on what was clearly a sub-par website had users and newscasters up in arms. The user experience was terrible, and in many cases, the system didn’t work. It was so bad, that even SNL got in on the fun:
But nobody had gave the Spanish version of the website a second thought until the AP released a report talking about the poor job that was done on www.CuidadoDeSalud.gov (The Spanish version of the health care portal). A couple of websites have already addressed some of the translation issues, but today I’d like to take a deeper dive into the issues that this project probably faced, and how they can be fixed.
Decide Who You Are Targeting
Being a government project, the organization that built this website should have access to more data about their target market than anybody. This should inform the way that content is written, and the essential points that need to be covered. And even if they didn’t, getting started with Spanish persona modeling is not that hard!
Doing so will help ensure that you considering the needs of the user, every step of the way.
Finding the Right Domain
For organizations that want to start approaching bilingual or Spanish speaking audiences, especially if they are targeting users who are familiar with their brand, then we usually suggest that they go with a subfolder or subdomain structure. This way they can leverage any branding benefits from things like TV commercials, physical placements, etc.
For this website that strategy would have made a lot of sense, especially if you are concerned about “English Only” groups being upset that you are doing something special for Spanish speakers.
However, if the stakeholders insisted on a translated URL, then the first thing to do is look at the intent that searchers have when looking for health care services. Unfortunately it appears that this was not taken into consideration, even a little bit. As you can see in the graph below, “Seguro Medico” easily has more searches over the past ten years.
And yes, there is an uptick in searches using the chosen term, but those are most likely a reflection of the advertising posh that was made on the website’s behalf. The fact is that the searcher intent of “Seguro Medico” is a much better match than “Cuidado de Salud”. I’ll write a bit about how to avoid translation issues a little later, but suffice it to say that this could have been avoided with some market research.
One of the main critiques of the website according to the AP report is that it appeared to be written in “Spanglish”. This led to the assertion that machine translation must have been involved. And one of the main reasons for confusion is how the developer chose to build the URL’s:
Their first mistake was to redirect the homepage of the site to an /es subfolder:
Think about that for a sec: You’ve gone to the trouble of using an entirely new domain, specifically tailored to a Spanish speaking audience, and then you redirect the whole site to a subfolder? Forget the SEO implications (basically your site is not as flat as it could be), but from the perspective of a user, this is really confusing! This mistake is then compounded when you start looking at the actual site pages:
Yes, they decided not to translate the names of the articles.
From an organizational perspective, this is completely understandable, since it is always easier to keep a handle on your URL’s if all that changes is the subfolder. But for users to see English language URL’s on a Spanish language site can be confusing (hence the “Spanglish” accusation”).
The fact that the content on each page is virtually identical to the corresponding URL in English could be detrimental, as Google could conceivably see it as duplicate, and penalize the site.
And let’s say for the sake of argument that they won’t budge on using the same URL’s for both sites. Then implementing the Hreflang markup to ensure that the content is correctly understood by Google’s spider is a must.
Set Rules and Follow Them
For every website, we have a set of guiding rules are compiled. They consist of limitations that client has given us and strategic decisions that we have made. For example, if a website gets most of its traffic from mobile devices, the meta data will have to be shorter.
Unfortunately, the metadata on CuidadoDeSalud.gov is not only inconsistent, it is also inconsistent, and unusual:
Using quotation marks around the metadata is just strange…why would you do that?
Another common rule that we have to take into consideration on Spanish SEO projects is whether to address users with as Tu or Usted (formal or informal). And once a decision has been made on that point, we consistently use the same form of address. Unfortunately, when you have several translators working on a project, and the rules aren’t clear, there is inevitably confusion.
While this particular mistake was not made, they did capitalize all the words in every title, which is not a part of the way Spanish is written. There are many other typos and grammatical errors that suggest a lack of review. In fact, there are even pages where you can see that the translation is almost 100% literal, and this is a problem because Spanish and English are not perfectly compatible. For anyone who has seriously attempted to learn a second language, this should not come as a surprise, but when you are undertaking a major translation project, the conversation isn’t how can we do this as well as possible, and becomes “what is the most cost effective way of doing this”.
How To Approach Translations… Carefully:
For those of you who have never attempted a large scale translation, I want to show you how the process works in the best and worst case scenarios.
Option 1: Google Translate the entire website
This is a bad idea for several reasons: First, the translation itself will be terrible. GT has improved drastically since it was introduced, but it is still not even close to acceptable. When users find pages like this, they tend to either go away, or flip to the English version of the site, and try and slog through it. Second, Google has explicitly stated that machine translated pages are considered spam, and could be penalized. This is not a bluff. We have seen cases of this happening in the wild.
Altura Grade: F-
Option 2: Google Translate the website, then pay an editor (or editing team) to fix it for you
In our experience, these projects will in many cases take longer than actually translating the material to begin with! The challenge is that you have to understand what the website was trying to say, meaning that you have to flip back and forth between the source material (in English), and the translation, and then fix everything on a document level, then a paragraph level, and then at a sentence and word level.
Altura Grade: D
Option 2a: The same as above, but use “machine learning” as part of the editing process
Machine learning is a giant rabbit hole as it relates to translation, but most systems operate strictly on a sentence level. This means that you have a platform that contains all the text on a given website. You then edit content, changing words and sentences to make them grammatically correct. Then, the machine will scan all the text on the website, and whenever that exact sentence occurs, will replace the previous version with the updated edit. For smaller websites, this doesn’t make any sense at all, but when you have millions of pages of text, there are bound to be thousands of sentences that are repeated, and therefor can be corrected once, instead of doing so several times.
It sounds sophisticated, but really, it is just like hitting the “Replace” button in Excel.
But you still have the same problem as before. There is no context, and the process of editing the content is tedious and expensive.
Altura Grade: C-
Option 3: Crowdsource the translation
Again, we are talking about giant websites that need to translate massive amounts of content. Some of the biggest websites in the world (including Twitter) have used crowdsourcing to cheaply and effectively localize their content.
The two main companies in this space are Smartling and Duolingo, and they are both interesting options. Smartling offers crowdsourcing as one of a number of options (including agency translation), and position themselves as a software company for translation projects, rather than a translation company, which is smart.
Duolingo is a whole different animal. In fact, the first time I heard about them, I was angry that I hadn’t thought of it first! They have positioned their app as a way to learn a second language by translating and editing content in other languages. Then, they give these users text from their client websites as homework. As users learn, they are granted ever increasing access. Of course Duolingo also checks content before delivering it to clients.
The websites (including CNN and BuzzFeed) pay for the service, but the translation app is free to users. I’ve even had friends independently tell me how much they love this app, that they have made more progress there than on Rosetta Stone.
The organizations that work with them still hire native editors to check the work, but even so, this might be the most cost effective way to do large scale translations.
Altura Grade: B+
Option 4: Have every word translated by a capable, native marketing team, who has done research on your company goals, has built personas to target, and has a firm grip on your company culture
This is the most expensive solution, but also the best. It ensures that your messaging will be consistent with your brand, in the context that you intend. It is our policy that every piece of content (translated or original) goes through at least two filters before being delivered to the client.
Of course, this attention to detail costs money, a lot of it (relative to other methods), but it also means that it won’t have to be reworked, or redone.
Altura Grade: A+
The fact of the matter is that in translation, you get what you pay for. This is why having typos (“De el” instead of “del”, mis-capitalization of words, etc.), are unforgivable considering the millions of dollars that were spent on this website. If only a fraction of the budget had been used on editing and Quality Assurance, we wouldn’t even be talking about this right now.
Everything Is Marketing
When you build website, everything about it has to be considered from the perspective of the user. And marketers have never been in a better position than now to make these sorts of decisions. We have more ways of gathering and analyzing data than ever before. It means there are no excuses for building something that is clearly substandard. Let’s hope that Accenture does a better job when they take over.