Mining for hidden truths: Everything you need to know about text analytics

There’s a lot of free-form content about these days. Survey responses, call centre records, and social media comments are just some of the contributors to the growing mountain of data that organisations across all sectors have to grapple with.

Extracting insight is akin to panning for gold - there are little nuggets to be found that are incredibly precious and will add value to your Voice of the Customer programme, but they’re hidden among a vast amount of less valuable content.

Sifting through all this free-form content manually is both very time consuming and resource intensive, so if you’re going to make full use of the gold nuggets, you need to find a scalable way to find them – and that’s where text analytics comes into play.

Here, we’ll look at some of the fundamental principles around text analytics and how integrating it with your Voice of the Customer (VoC) programme can help you to take meaningful action to enhance the customer experience and boost the bottom line.

First things first… does it actually work?

This may seem a bit basic, but as with most relatively new technologies, it’s still well worth asking. The idea of being able to take tens of thousands (or more) of customer comments and get a genuine steer on what customers think so you can take action does sound rather too good to be true.

Some argue that technology can never pick up the nuances of language as well as a real person. In fact, people can have such vastly differing interpretations of the same statement, that the right software can produce results almost exactly on a par with a human analyst. Furthermore, in a test which compared free-form product reviews on Amazon with the number of stars the reviewer provided for the product, some solutions can achieve an 89% accuracy rating, which is certainly enough to warrant a statement as bold as “text analytics works”!

Ok, it works, but what does it do?

Basically, it takes the complexity out of understanding what customers and the wider market are saying about you. Text analytics solutions are trained against the right types of language for your business, and then categorise comments by sentiment. The sophistication of these solutions now means that they can also understand complex comments which talk about several factors at once. For example, “My hotel room was fine but we were very disappointed with the restaurant, it was slow and the food wasn’t as good as we’d expected. The concierge was wonderful, though, and recommended plenty of great restaurants nearby.”

Sophisticated solutions will be able to take this comment and split it into multiple parts and allocate a sentiment score to each part. So the hotel’s restaurant will receive a negative score, while the concierge service will receive a positive one.  Multiply this by several thousand and add it to your VoC programme, and you can develop an incredibly rich picture of every facet of your company.

Do I need it, though?

The lifeblood of any Voice of the Customer programme is feedback. Surveys, in particular, bring in a huge amount of data, and much of it is neatly packaged into ranking scores, tick boxes and sliding scales. Insight like this can take you a long way down the road of streamlining processes, responding to dissatisfied customers and understanding some of the key drivers behind customer behaviour. But once your VoC programme has matured and some of the quick wins are behind you, adding text analytics can offer a new lease of life.

For example, you can gain insight into what is being said across different categories of your business, such as stores, call centre, and your website. Rather than having a simple picture that shows how customers have rated their experiences, and whether they intend to use your company again, you can get to the critical “why?”. Most importantly, you can access the “why?” at an aggregate level through sentiment analysis.

How does that help me?

It enables you to take action that will have a significant impact. For example, if you can identify from your business metrics that you’ve had a drop in your satisfaction or recommend scores, you can harness text analytics to understand why that’s happened. An important point here is that you’re not necessarily dependent on survey responses telling you why customers have become less satisfied.

If a significant group of customers have started providing lower satisfaction scores, but failed to provide verbatim comments as to why they’ve done that, you can bring in free-form data from other sources. When you add comments from contact centre records, CRM, social media and other unsolicited channels, and combine it with your survey data, the detail you uncover can provide clear causes. And when you have clear reasons for shifts in your metrics, you can take clear action to rectify the situation.

Voice of the Customer programmes are incredibly powerful, and can generate huge insight. The addition of text analytics and sentiment analysis provides the next steps in ensuring that not only are you able to hear the customer, but you’re really listening. 

Lauren Azulay is senior product manager of Confirmit