A voice of the customer (VOC) programme allows you to determine which questions to ask your customers. This has been and remains an important principle in managing the VOC, because if you just said to your customers “tell me whatever you want to tell me” they will give you a mass of unrelated information that is not possible for you to action. At least by structuring the feedback and by using good research practice, you can limit the responses you get back to the things that are controllable in your world.

Until recently, VOC was therefore mostly quantitative and largely structured according to the terms of the organisation, apart from that one customary verbatim question at the end of most surveys that represented qualitative data. The bias for quantitative data made things simple, because it required very simple database structures and it presented you with numbers you could take averages of to draw a pie chart. As your VOC information became more sophisticated, you could start asking questions like “how does A predict B?” Most CX (client experience) players are now very good at this.

But what about the world of unstructured data? This type of data is comprised of two types:

  • Unstructured data where your customer is talking about you. This is opinion pieces you might find on Twitter, Facebook, other social platforms and sites like HelloPeter.
  • Then there’s the unstructured data in emails and conversations. This is underpinned by behavioural information that looks at what people are actually doing and how are they doing it. For example, how often they phone an organisation.

Paradoxically, a big problem with unstructured data, is how to structure it in a way that makes sense.

CX used to be so simple because it was structured on your terms. Now you want to use information that you don’t want your customer to structure in any way. However, you want to bring unstructured information into a structure framework to see if it’s meaningful.

One of the most obvious ways to bring structure in, is to link the unstructured data to the customer journey map. We can also use time as a common structure element to which we link the unstructured data. This means we plot the unstructured data (Tweets and comments, for example) to a timeline and plot the structured data (VOC scores) to the same timeline. We can then use the concept of sentiment to see how that influences the scores. Once you have a view of the timeline, you can ask what sentiment’s role is in influencing CX scores: is sentiment predicting it, or is sentiment not predicting it? And is there a lag?

Given the growth in the availability of unstructured data, we have to consider the possibility that the voice of the customer will become entirely unstructured. In fact, in ten years’ time, we will probably not be doing surveys any more. Instead those of us who are adept at mining the information that is published for free by your customers, through how they behave and how they talk on social media, will be first to find new insights that help you stay ahead of the curve.