Along our journey, we have discussed with a large number of companies with global customer bases. These companies have customers in many continents and therefore receive (or should receive) customer feedback in several languages. For many of them, this provides a challenge: they don’t have the personnel to read or understand the feedback in a language the headquarters doesn’t speak. It is interesting to learn how they deal with this challenge. These are some of the solutions we have met.
1. Feedback is asked only in the home market
This might sound hard to believe, but we have come across major global US companies, who – because of their lack of language skills – only ask customer feedback in the US (and, perhaps, internationally in the UK). This provides them with heavily skewed results.
2. Only the feedback in familiar languages is utilized
One company we talked to had a global policy of asking NPS feedback from all its customers. All feedback also contributed to the global NPS score of the company. But as there was no-one in the team to understand some major languages, part of the feedback was simply left unread.
3. No text feedback is used
Numbers are (nearly) universal. If you ask feedback with only questions where the consumer can give you numeric grades, you should be able to deal with feedback from anywhere. Too bad the numeric only feedback is seldom very insightful. And if you want to dig out what the customer really thinks with number responses only, you need to ask a lot of questions.
4. Text feedback is only utilized in local offices
Some companies do ask text feedback also in local languages but aggregate only the numbers to the corporate level. The text is handled in the local sales offices. This provides means for the local team to do some often-tactical improvement actions, but the global leadership is left in the dark, and company-wide strategic improvement actions are never going to be done.
5. Only Sample of the feedback is translated
This is a widely used method as well. Even though feedback is provided by thousands of customers, some companies decide to translate only a sample of the feedback. The translation can then be managed by a professional translation agency. If the sample is big enough and done well, the overall result can be statistically reliable. The challenge with this approach is that it is still expensive, often slow, and the sample size is hardly big enough to provide more detailed insights for operational decisions. E.g. the sample can be large enough to reveal to the leadership that there is something wrong with the product reliability. But if your product manager wants to study the issue deeper, he won’t find enough insightful comments by just reading through the sample.
As you can see, there are many ways to deal with the challenge of feedback in multiple languages, but they all have their weaknesses.
How do we solve this challenge?
Our solution at Lumoa is to get all the translation done with automated translation engine. Both Microsoft's and Google's automated translation services are quickly improving quality wise and they provide a cost effective way to get a large amount of text feedback translated without any delay.
The immediate worry our clients have is always the same: what about the quality? Everyone has experienced automated translation to produce some funny results - especially when translating from small and difficult languages (such as my native language Finnish).
It is true that automated translation will not be replacing professional translation services anytime soon. But we are not translating a book here, not even content for marketing materials! We just want to capture what the person talks about and the strength of the sentiment (how strongly positively or negatively the topic is talked about). And for that purpose, the automated translation works well enough. It enables you and your company’s leadership to understand what people say. And it enables aggregating and categorizing the feedback into topics across languages. In our analytics, it enables also assessing, how impactful the categories are. This is enough to support fact based decision making and start making improvements based on the customer feedback.
In customer feedback analytics the end-result matters most: if the translations are good enough to empower the right improvement actions, don’t worry about the quality too much. What your customer cares about is the improvement: the fact that they are being heard and the issues are being fixed. They don’t care if something they wrote sounded a bit funny in your regular monthly management team report. It is also great that this method allows us to take into account text feedback in dozens of languages. We don’t need to choose to focus the voice-of-the-customer efforts only in a few large markets but can listen to all of the customers. Whether you listen to your customers should not depend on the language they happen to write in. All customers matter.