A checklist for multilingual customer support

7 min read

Nowadays, online services target customers in hundreds of countries at once. And thanks to a fully localised customer experience which includes payments on credit and debit cards all in the customer’s natural language, consumers are becoming ever more comfortable with using services domiciled in another country.

This will only continue: online services are built to scale, and the international payments ecosystem is rapidly becoming seamless, too (think businesses like Transferwise).

But delivering support in multiple languages is much harder than delivering a product in multiple languages. Localisation is pretty much a one-off process, and easily quantified: you know exactly how many pages, screens and experiences you’ll have to translate. Support is all about exceptions to the norm: it’s the constant challenge of dealing with the unexpected on a timescale dictated by the customer.

Clearly, keeping a call centre stocked with an international community of CSAs to rival the Eurovision Song Contest is not going to be economically viable. Maybe you’ll want to support some top languages and ignore the rest. It demands some research – and even if you want to deploy Unbabel’s instant international support service, these are the sort of things we’ll want to know…

 Know what languages your customers speak

Obvious. But still usually a surprise. IP data, emails and other contacts will give you a good picture of your language needs, but technology won’t give you all the answers:

Include second languages. In some territories, a second language is so pervasive that the problem is dramatically mitigated: in the Netherlands and Israel, for example, English is very widely spoken. 95% of Latvians speak a second languages.

Get the product right first. If you receive a disproportionately high number of support requests in a language compared to service users in that language, this points to a more fundamental problem with the localisation of the product (or possibly a cultural misalignment). In either case, don’t sweat about translations, sweat about the product itself!

Questions to ask:

> What are the main languages spoken by our customers?

> At what point do they tail off (expect 3-5 main languages and then a dramatic fall-off)?

> Is there a difference between the languages used in support calls and the languages used by non-exceptional customers?

> What are the secondary languages in our territories?

 Give customers as much information online as possible – and the tools to self-serve

Most of us have called a customer service centre, already mildly irate, and heard something like this to make their blood boil:

“Did you know, you can find the answer to most problems online. See our website, at www.patronisingthecustomer.com”

As we transition towards the majority of people being online-savvy (even 28% of elderly people will buy online), callers are either people who don’t like going online at all or people who have tried to find their answers online – and found the information lacking.

At the very least, ensure that all FAQs are localised into as many languages as necessary; and that both original and translated versions are revisited regularly (quarterly or every six months is ideal).
And where possible, localise self-service tools: you almost certainly run password resets without human intervention, so see what else could be automated.

Questions to ask:

> Is all our customer service information localised?

> Do customer contact reports feed into the production of future FAQs?

> Can we automate anything?



 Understand the service expectations of a local culture

Cultural aspects of customer support can vary widely. Some countries have different expectations of the sort of service they should receive. For example, the US travel services business, Priceline, found that South American consumers spent more time on the phone — which would require a larger support staff.

Perhaps more importantly, some countries have rigid rules on hierarchy and/or manners, such that spending more time crafting a culturally appropriate reply will yield a better level of satisfaction than racing to get a faster reply out. The same Priceline spokesperson notes that extra training would be given to CSAs servicing the Japanese market to account for the important honorifics (titles and forms of respectful address) in Japanese culture.

Questions to ask:

> What are the local expectations of service in each territory?

> What is the cultural role of communication in each territory?

> What do people in each territory value most? And what would turn them off?

 Time zones are cultural assets too

Today, speed matters. We have culturally ingrained expectations about how long a reply should take. For example, we now expect “verify your email address” emails to be delivered within seconds of signing up to a new service. We want a reply to a tweet within a few hours – especially if it’s a complaint. We expect a customer support email response within 2-24 hours (and we give extra credit if it’s faster than that). In short, we’re time-sensitive.

Research from our partner, Zendesk , found that in most territories, speed of response is the greatest single factor in customer satisfaction; and that means time zones take on a new importance.

If you require 2 hours to deal with a query (which is an impressive response!) but your service hours are 8am – 8pm, then what you call ‘night-time’ will, on the other side of the world, be up to 14 miserable hours of waiting. Many online businesses find themselves opening (or buying in) 24/7 support centres for exactly this reason.

Questions to ask:

> What does our world look like? Are there ‘hotspot’ time-zones to consider?

> What are our best and worst possible outcomes for response times with current staffing? Is this adequate?

 Assess your linguistic long tail

You’ll almost certainly have heard the concept of the ‘long tail’. It’s a diagram something like this:


…and it describes the many situations where a small number of items account for the majority of circumstances and a very large number of items account for the remainder. For example:

Library books: in most libraries, a small number of books are hugely popular and lent out all the time; but the majority of books are much less popular, and some are barely ever lent out at all.

Google’s most searched: Google’s most searched ‘how tos’ include ‘How to tie a tie’, ‘how to kiss’, ‘how to get pregnant’ and ‘how to lose weight’ (that was one hell of a night!). But after these very popular questions, the rate at which questions are asked drops off rapidly, with millions of questions only asked once.

Your demand for support languages is almost certainly a long-tail graph (our client, Skyscanner, saw exactly this type of demand for languages). A few languages will be worth local or in-house support from native speakers. The rest – the long tail – are where you must make a call as to how much support will impact profitability.

Questions to ask:

> What shape is our long tail? Where does in-house support end and on-demand support begin?

> At what point is support no longer economical to provide?

> What is the trend in product usage in long-tail territories – where is usage surging, which would warrant extra attention?

How can you scale customer service across multiple languages?

We recently did a webinar about just this, showing how companies like Skyscanner, Daniel Wellington, Pinterest and Under Armour get it done with Unbabel with our integrations for Salesforce, Zendesk and Freshdesk.



 If you have any questions or if you want to discuss your current challenges with us feel free to reach out to us at raquel@unbabel.com or at ricardo@unbabel.com.

ArtboardFacebook iconInstagram iconLinkedIn iconUnbabel BlogTwitter iconYouTube icon