"This call may be monitored or recorded for quality assurances purposes." It's a phrase that is so common it's become white noise to consumers calling contact centers. These recordings have resulted in so many miles of recorded transcripts it is impossible for even a fraction of them to be read and analyzed by human employees. This overwhelming feat catalyzed the now-popular tactic of running these transcripts through algorithms that turn them into word clouds. Word clouds are an excellent starting point when it comes to the the business of keeping customers happy, but they are just the beginning of the insights that can be extrapolated from customer feedback data. It's time to get your head out of the word clouds and dig a little bit deeper.
Word clouds serve a specific purpose and do it well: They help surface the topics about which customers are most frequently talking. If you're in the business of customer experience, you can probably rattle these topics off the top of your head. If not, by all means, get your head back in those clouds and start your exploration there!
Word clouds are only the tip of the iceberg, however, when it comes to exploring the ways in which your company can improve its relationship with customers. Today the technology that was originally used to create word clouds, natural language processing (NLP), has evolved to help us understand customer experience issues on a more granular level. What used to be the finish line is now the starting line. This more evolved approach is known as natural language understanding (NLU).
Tools that apply natural language understanding offer a more sophisticated look into what customers are discussing because they can understand nuance. NLP can identify individual words, label parts of speech, and can even pick up on the basic relationships between words. NLU, on the other hand, takes these inputs from NLP and uses them to interpret the intents, goals, meanings, and purposes within the body of the text. Essentially, we have taught machines to understand the nuances that humans pick up naturally.
Let's consider the difference between "The car allows us to crest the hill smoothly" and "Crest toothpaste is very smooth." NLP technologies would recognize the word "Crest," and some would even differentiate between the verb and the proper noun, respectively. But, I would bet that you've never found business insights from simply analyzing parts of speech! Tools focused on understanding go one step further by recognizing that Crest is a brand name within the consumer packaged goods industry. Looking beyond individual words, NLU techniques help us interpret customer feedback by identifying the most salient sentences (such as suggestions, requests, or cries for help) through analysis of structure and syntax. The power to look beyond specific words opens up a new universe of opportunity for understanding and analysis.
Being able to navigate these important nuances can help uncover customer emotions and effort that often lead to surprising discoveries. Take the case of a personal kitchen appliance company. Its product offerings includes slow cookers, but the company was curious to know about demand for pressure cookers - especially in the wake of several news reports of pressure cookers exploding. Using a tool that employed NLU principles to analyze the emotion of worry, the company started tracking how consumers felt online about pressure cookers. Surprisingly, despite an increase in the volume of feedback about the emotion of worry (typically considered negative), there was also an increase in positive sentiment. Upon closer inspection, customers were expressing a lack of worry about pressure cookers when discussing rival products. This insight gave the company the confidence to release a pressure cooker to the market and increase revenue.
Beyond this fairly low-stakes example, there are even more pressing reasons why the word cloud is limiting. The word cloud is a snapshot into topic frequency and how often certain words are being repeated within different channels. As such, word clouds mask individual voices and individual issues and, unfortunately, customer experience professionals can't always take such a laissez faire approach to customer feedback. Some situations require immediate attention. For example, if a customer calls a contact center once with the complaint of an injury sustained at a retailer's local store, this is something the legal department would likely need to know about right away.
For industries that have to be particularly cautious when it comes to compliance issues, understanding nuance is of utmost importance. The financial industry, for example, is an ideal candidate for employing NLU, as words like "fraud" or "false account" are couched in context that makes all the difference in the world. A customer saying "the teller assured me there are no false accounts in my name" versus "I'm calling to report a false account in my name" are two very different statements that require different actions. The ability to contextualize their use is critical, and missing these moments could lead to dire consequences.
I wish I was offering you an easy button, but like all analytics offerings, NLU-based features must be deployed thoughtfully, with consideration to data source and use case. Each channel's data has to be treated a little differently. The language your customers are using on Twitter is different than the language they are using over email, and missing this context could lead you astray when it comes to driving strategy and prioritizing.
Data sets themselves can also be misleading. Oftentimes contact center data is composed of a mix of word-for-word call transcripts and the notes agents took on calls. While the former is a gold mine of information for you, the latter could obscure customer voice. Even though something is better than nothing, you'll have stronger insights if you can hear from actual customers themselves.
Finally, NLU should be seen as a tool, not a consultant hired to fix problems. If you leverage NLU techniques and features and discover where your company has room to grow, it is up to you to map out your action plan. Like their word cloud predecessor, NLU techniques can point you in the right direction, but the next steps are yours to map out.
Ellen Loeshelle is senior product manager of natural language processing and enrichment at Clarabridge.