Supporting Your Multilingual Customer Base: Why Google Translate Isn’t Enough

Providing excellent customer service is paramount for any brand looking to retain its hard-earned customers and, ideally, expand its relationships with them. Not only do 96% of customers say that they will end a relationship with a brand after receiving poor customer service (Forbes), but customers are also increasingly expecting more personalized support from companies (Zendesk).

The heightened demand for personal support experiences extends to language; 3 out of 4 buyers are more likely to purchase from a brand again if after-sales support is offered in their own language (CSA Research).

With data like this, it’s clear that providing multilingual support should be a major initiative for any business with a global customer base. But how brands choose to go about accommodating these language preferences can make or break their success. Free machine translation engines like Google Translate are widely available and provide instant translations, but the quality of these translations, while they can be quite strong depending on the use case, may not meet the customized needs of a global organization. For a brand to rely exclusively on a tool like Google Translate would be to risk communication issues in conversations with customers – a risk that most brands aren’t willing to take.

Here are some of the top drawbacks to relying on Google Translate:

Translation Accuracy

How accurate is Google Translate? It depends on which version of Google Translate you’re referring to. When the tool was first released back in 2006, it functioned by translating inputted text into English before translating it into the desired outcome language. Unsurprisingly, this resulted in a plethora of issues with the translation quality, particularly related to grammatical errors. It also translated word-by-word rather than looking at the context of a sentence as a whole, resulting in patchwork translations that, simply put, made no sense.

In 2016, Google Translate improved by switching to a neural machine translation (NMT) model, which vastly improved its ability to interpret whole sentences. This shift also improved the perceived quality of its translations–as long as the inputted text is well-structured, made of simple sentences, and grammatically accurate.

These qualifiers might not seem like such a problem until we think about how customers interact with businesses. Your customer support team probably regularly sees incoming chats and tickets rife with spelling errors, acronyms, slang, and industry-specific jargon. Though Google Translate is presently one of the best machine translation engines out there, relying on it to accurately translate strings of text that require a business-specific translation will likely result in translations that don’t fully capture the true meaning of the original text. This can lead to confusion or, at worst, completely distort the meaning of the original question or request.

Language Limitations

Google Translate supports translation of 109 languages, but not all language pairs are of equal quality or reliability. Studies have shown that the accuracy in Google Translate heavily favors European languages, such as German, Dutch, and Italian. When human translators reviewed certain other languages like Bengali, Urdu, and Punjabi, for instance, they ranked the translations completely inaccurate between 80-100% of the time.

If a business’s customer base speaks primarily European languages, then Google Translate will be more reliable–though translations for its most accurate languages are still lacking as far as slang, jargon, acronyms, and misspellings are concerned. For businesses who may ever have to interface with someone from South Asia, where languages like Bengali and Urdu are more common, Google Translate is simply not an option, as the translation quality is near nonexistent. Add complications like slang or acronyms into the mix, and the challenge only worsens.

Practical Constraints

While quality of translation is the most important element to any machine translation solution, there are a number of limitations associated with Google Translate that can make it even more of a frustrating experience for businesses to rely on. One of these is character limit: Google Translate allows users to input up to 5,000 characters to be translated at a time. For individual strings of text, such as incoming chats from customers, this doesn’t present a problem, but when trying to translate resources such as FAQs, knowledge base articles, and whole web pages, 5,000 characters will rarely be enough. 

Of course, there is always the option to translate an entire article in different chunks of text. Depending on the volume of content requiring translation, however, this can present a problem of inefficiency, making what should be a simple task take up far more time than it should.

Another constraint is that free translation engines like Google Translate don’t provide the option to request a retranslation. What happens if a support agent submits text to Google Translate and doesn’t like the outcome of the translation? The agent has the ability to rate the translation or suggest an alternative, but there is nothing they can do to trigger a different translation to come through. This means that whatever translation Google first provides is the one that agents are stuck with. If the quality of that translation is clearly poor, then the agent may find themselves conflicted on how to respond.

Lastly, when cutting and pasting support content into the Google Translate web portal, that content is subsequently stored by Google. It’s even available to be called back up in the “History” option. Because support chats and emails frequently contain personal data from customers, it’s critical that brands rely only on secure machine translation solutions to avoid subjecting that personal data to a breach.

Alternatives to Google Translate

With Google Translate not being a viable option for businesses to solely rely on, this naturally leads to the question: what is?

There are a number of free translation engines on the internet, including Microsoft Translator and DeepL. While the perceived accuracy of translations from these sources varies, they tend to have similar problems to Google Translate. Businesses looking to support their multilingual customer base shouldn’t rely on these engines for the same reason they shouldn’t rely on Google Translate.

To achieve the highest level of translation accuracy and quality, human translation will always be superior–but hiring a support agent or entire team to accommodate each language spoken by a company’s customers is, for the vast majority of businesses, completely unattainable. That’s where a solution like Language I/O comes into play.

New call-to-action

Language I/O’s technology doesn’t rely on a singular NMT model for its translations–it intelligently selects the engine that is best for the language being translated. Paired with this intelligent selection is the ability to impose a customized glossary of terms on all translations, meaning that tricky terms like slang and industry- or business-specific jargon will be translated accurately. From a security perspective, Language I/O scans each piece of content for personal data and encrypts it before it is passed on to any third-party NMT platform. Further, Language I/O only integrates with NMT engines that will agree in writing not to store the content that is passed to them for translation. 

For businesses who regularly communicate with multilingual customers, it’s critical to have technology in place that integrates with your CRM directly or via translation API, and provides reliable, accurate translations. If you’re interested in learning more about Language I/O’s proprietary technology and how it can support your organization’s multilingual support efforts, contact us