Share this content

Voice of the Customer: How to handle unstructured data

19th Mar 2014
Share this content

Voice of the Customer programmes thrive on data. Increasingly advanced programmes gather customer feedback from both relationship and transactional surveys across the customer lifecycle. In many cases they also reach beyond direct customer feedback into data from other business systems, such as CRM platforms and employee feedback. Surely now companies have everything they need to create that elusive holistic view of the customer?

It’s a broadly accepted fact that around 80% of enterprise-relevant information originates from text-based, unstructured sources. Emails, social media, free-text survey questions and other sources hold huge, often untapped potential to dig deeper into our customers’ experiences. Keeping track of this, and extracting useful customer experience information from it, is a growing challenge for many businesses.

Do we really need unstructured data?

It’s a valid question, though one with a very clear answer. With so many businesses now operating effective Voice of the Customer programmes, and the best among them integrating data from a range of other sources, decision-makers are in a better position than ever to put their companies on the right track. 

Even so, as customer experience professionals, we need to be exploring other sources and that means unstructured data.

It’s understandable why many have been reluctant to do so. Unstructured data is not only messy, but when you begin to factor in unsolicited sources, such as social media and email, then it’s also vast. However, it is manageable if you take it one step at a time.

Automate and aggregate

This seems obvious, given the scale of the issue we’re addressing, but you need to automate as much of the process of capturing and analysing unstructured data as possible. For consumer businesses in particular, it doesn’t take long for data to build up to a point where it is otherwise completely unmanageable.

Text analytics solutions come in several forms, including standalone platforms and built into wider Customer Experience Management solutions. In essence, these products are able to ‘understand’ what customers have told you in free text and refine those messages to categorise each comment accordingly. This enables you to aggregate comments into a series of ‘buckets’ which provide much clearer meaning and help you take decisions about the action you need to take. If your bucket that says ‘Your delivery processes are a mess’ is filling up fast – you know where to focus! Without text analytics, you might miss the cause of customer dissatisfaction because customers have only told you that they’re not happy, but not why. Or they might have tweeted about you, instead of filling in your survey.

Integrate with structured data

A fundamental part of a good Voice of the Customer (VoC) programme is the avoidance of silos. Whether they’re created by using different feedback channels (e.g. web, mobile, phone etc.) or by different teams running their own ‘independent’ touchpoint surveys, silos hugely reduce the insight a programme can generate. Unstructured data poses a real problem here if it’s treated differently from the rest of your information. It’s critical to ensure that you’re able to capture and analyse verbatim text and integrate the results with, for example, your survey data, CRM and ERP data and employee engagement information.

When you’re looking for the solution that’s right for your business, make absolutely sure that you can move data from the collection, or ‘mining’ stage into the core of your VoC analysis. If VoC provides the ‘why?’ to CRM’s ‘what?’, then unstructured data offers the ‘what next?’ that you need to take the right action.

Start small

Like all elements of a VoC programme, it pays to start small and build up as you ensure you’re getting every stage right. So don’t start by trying to analyse everything that’s being said about you on social media. The scale is likely to be vast and you need to make sure your system is properly “tuned in” to your business and industry before you take it to that level.

A great place to start is with solicited and ‘owned’ feedback, such as free text answers to your survey questions and comments made on your own social media pages and corporate Twitter account. This will provide a great learning ground for your expanded programme and help you to iron out any problems early on. An additional challenge of unsolicited feedback is that you’re not easily able to tie it into a stage in the customer journey. This makes it harder to prioritise and again highlights the need for the ‘buckets’ which help you to understand trends. As you build up to analysing unsolicited feedback, you can use it to tweak your surveys and sense check what’s going on in the market, to what extent it impacts your customers, and identify ways of moving forward.

Take action

This is another seemingly obvious one, but as with all aspects of VoC, you must take action on what you’re told. While you can’t necessarily get back to every customer to respond to every comment, you can group customers by the ‘buckets’ into which their comments fitted and tell them what you’ve done to make things better. Unstructured data provides incredible insight into the underlying drivers of loyalty, and it’s vital that you make the most of the information at your fingertips.

There’s no doubt that the idea of mining potentially millions of customer comments is a daunting one. But with customers carrying on their own conversations about – but potentially without – you, it’s not something you can avoid forever. The companies that get it right, by listening in the right places, using this new insight to augment structured, solicited feedback data, and taking action accordingly, will have a huge advantage over their competitors who decide it’s a road best left untraveled. 

Karine Del Moro is vice president of marketing at Confirmit

Replies (0)

Please login or register to join the discussion.

There are currently no replies, be the first to post a reply.