Wondering what secrets lie ahead? A global conspiracy where everyone's fibbing about translation? Not quite – let's just peek behind the curtain of how digital product translation actually works in the real world.
Maybe a better title would have been:
Website Translation Guide: How to Properly Localize Digital Products in 2024
When I first dove into this ecosystem almost a year ago, I was genuinely taken aback by several discoveries:
- What we commonly call translation in the digital world is actually termed 'localisation'
- Despite our AI-driven world, humans are still heavily involved in translating digital products
- Translation isn't processed in real-time as many might think
If you're reading this article, I reckon these revelations might resonate with you too. Let's explore these three key insights I've gathered from working with multilingual websites and apps.
What's the Difference Between Localisation and Translation? A Practical Guide
"What's the difference between localisation and translation?"
This is a brilliant question I get asked nearly every day. When I mention localisation and follow up with "it's essentially translation," I can practically predict the next question: "Well, why not just call it translation then?".
Most people associate localisation with map pins and sharing their location with mates. While that's not quite it, there's an interesting parallel here.
Localisation involves displaying the appropriate language based on your geographical location. So yes, there is a geographical component, but it goes much deeper than that.
What truly sets localisation apart from translation? I wrote about this extensively a few months back. The key distinction lies in cultural adaptation. It's not just about converting Spanish text; it's about adapting it specifically for Spain, Mexico, or Argentina. Similarly, English content needs different treatment for the UK, US, or Australia. While this might seem trivial initially (as it did to me), these subtle differences can significantly impact your message's effectiveness. We all have unconscious cultural biases that influence how we perceive content.
Take this example: A perfectly translated phrase might feel off simply because it doesn't align with local linguistic patterns.
"Keep your zen at the office" might be a literal translation, but it sounds rather awkward to British ears.
A more natural British English version would be "Stay calm at work" or "Keep your cool in the office".
You'll find two camps in this field: localisation enthusiasts and everyone else.Those unfamiliar with localisation often suggest: "Just call it translation."
Meanwhile, localisation experts might scoff: "Localisation is far more nuanced than mere translation!" It's a peculiar middle ground where you can't seem to please everyone.
How Does Professional Translation Actually Work?
Initially, like many product managers I've since spoken with, I assumed website translation was straightforward - just run everything through DeepL and you're sorted. Well, not quite. It's both more complex and more expensive than that.
First, you need to implement translation keys (if you're not familiar with these, I've written a detailed guide about them). Why? Because it's the only practical way to maintain multiple language versions of your website. You extract all text from the code, replace it with localisation keys that pull in the appropriate translation when the website loads.
After implementing these keys (which our Figma Plugin can automate for you, and for free), you're ready for translation. You might have JSON or CSV files containing your keys and source language text. For instance:
mobile.app.onboarding.welcome.title → "Welcome aboard"
Using DeepL or similar services for direct translation can be problematic. They translate line by line, missing crucial context between related elements.
Translation involves more technical aspects than most realise, particularly regarding translation memory. This is essentially a database of previous translations serving two crucial purposes:
- Cost Efficiency: Automatically reusing previously translated identical text
- Consistency: Ensuring uniform translation of specific terms throughout your content
Consider the confusion when "update profile" and "update settings" lead to the same page. Similarly, British English preferences like "basket" versus "cart" need consistent application.
While machine pre-translation (whether DeepL or Google Translate) can provide a starting point, don't expect perfection. The output still requires significant refinement to match your brand voice, legal requirements, and marketing objectives.
The Myth of Real-Time Translation
You might assume translation happens on the fly, like Google Translate's automatic webpage translation feature. If you've tried this, you know it's far from ideal.
Here's a recent example: Someone viewed our website through Google's automatic translation. Yes, I know the irony - we're working to revolutionize localisation, but our own website isn't localized yet. As they say, the cobbler's children go barefoot.
The results were... interesting. While technically comprehensible, the text felt unnatural and awkward. Even our company name, Gleef, was translated to "joy" (because "gleeful" means "full of exultant joy" - who knew?).
This demonstrates why human expertise remains essential in current translation workflows. Or rather, remained essential.
This week, we've launched the beta version of our translation tool (a new tab in our Figma Plugin), and the results have been remarkable. You should definitely give it a try. And if you’re not confident enough to try out AI translation, you should read this case study published in Le Monde (sorry, French speaker only).
Finally, there's another crucial reason why real-time translation isn't ideal: SEO impact. Local search optimisation requires carefully crafted, context-aware content - something automated translation simply can't provide effectively.
