Satisfied Customer? Dissatisfied Customer?

happy sad

Unmasking Customer Attitudes with Sentiment Detection

“Customer sentiment” – a customer’s attitude toward a company, its products and services – has long been a highly valued indicator of business health. It’s not difficult to understand why: satisfied customers buy more and stay longer; they give good recommendations about products; and they refer their friends to your company. Dissatisfied customers talk about your company too, but you won’t like what they’re saying.

When business was largely conducted in person, sentiment was easier to recognize from visual clues like body language. As more sales and customer support are delivered online and via phone in contact centers, rather than in retail storefronts and shops, it becomes a little more difficult for a representative to pick up on those clues and respond to them.  More importantly, it becomes even more difficult for the managers and executives to understand the sentiment of their customers and how the company’s representatives respond to it.

Sure, survey programs can get you part of the way there, since they can give you a summary view of the overall trends, but customers will stand for only so many surveys.  So businesses need to find ways to measure sentiment directly on every interaction.  They need to understand as much as they can about each customer’s thought process, and each representative’s response – not just the opinions of customers that agreed to complete a survey.

Sentiment detection provides the means to measure certain basic speech components to unmask customers’ underlying attitudes during all of the business’s non-face-to-face interactions.

So how to go about measuring customer sentiment?  The first thing to do is break it down by the aspects of sentiment that we can read and hear in customer interactions. For audio interactions, there are three dimensions of acoustic analysis possible:

  1. Customer’s emotional state – gauged by voice tone, pitch, volume and speed; often referred to as “emotion detection”
  2. Cross talk – callers interrupting and talking over one another; typically suggests negative sentiment
  3. Laughter – typically indicates positive sentiment, good feelings of a caller (and the representative, too!)

For both audio and text-based interactions, there is the linguistic dimension:

  1. Words and Phrases – what the caller actually says

Some interaction analytics tools focus solely on the acoustic speech components to indicate a customer’s emotional state, but those features by themselves don’t reveal a complete picture of customer attitudes. A caller’s observed emotional state can stem more from personal temperament or immediate life situation than attitude toward your company. This is not to say that emotional state is not relevant or not useful in some cases, rather, it is not enough by itself (and sometimes can actually be misleading).

This is where the linguistic analysis saves the day. The customer’s words and phrases hold the mother lode of insight into the purpose of the call. The most effective sentiment detection approach looks at both the linguistic features and acoustic features in context to deliver the most revealing results across the widest swaths of customer and representative populations.  It provides the granularity of detail that enables the business to control for personal temperament and external factors.

So now that we can measure sentiment, what can we do with it?

How Sentiment Detection Can Enhance Early Discovery
Sentiment detection provides a valuable search tool to use in the early discovery phases of the analysis of your customer interactions. Sentiment can be tracked and visually trended within each and every call (plus chat and email).  Summarize the sentiment across the entirety of each interaction, and now you have the means to sort those interactions based on positive or negative sentiment.

The ability to associate trending words and phrases in context with sentiment shows relationships between topics, and can reveal events and phrase associations you may not have known to search for, including those topics that cause the positive or negative sentiment. This surfacing of trends provides a first key step for mapping and analyzing the root causes of customer contacts.

For example, you may not be surprised that calls focusing on the topic of billing  typically trend negative, but within that call universe, the unexpected phrases “office hours” or “closed” seem to drive negative sentiment scores even higher. Discovering how a billing department that is only available from 9 to 5 affects the experience of customers with bill payment issues enables you to take proactive steps to address this situation. Likewise, if a particular phrase such as “free upgrade” tends to correlate to calls whose scores trend from negative to positive, you may have discovered a valuable insight to share as a best practice among your team.

The Value of Real Time Sentiment Detection
Within real time analytics, sentiment detection can be a powerful cue that a call is “going off the rails.” During a particularly contentious call, a real time alert based on sentiment scores can be triggered to request supervisor intervention. Or an agent may receive an alert to remind them to strictly follow established corporate and legal guidelines in executing the call.

Likewise, sentiment indicators from positively trending calls may offer upsell or cross sell messages to agents when a customer is identified as a receptive candidate for specific offers. The customer’s words combined with contextual emotional cues make such automated insight and action possible.

Sentiment‘s Strategic Benefits to Your Enterprise
Sentiment detection’s value can be felt far beyond the contact center. Customer sentiment is a key component in measuring a company’s brand value. The marketing department can use sentiment clues derived from interaction analytics to develop campaigns that address negative customer opinion. Or they might launch a program to expand customer engagement to support a new product launch. Research and Development might respond to unfavorable sentiment scores on a product with new features that address customers’ past disappointment. Technical support might learn that unpopular verification procedures required to access assistance are contributing to churn, and take steps to reduce customer effort.

No matter which departments benefit, armed with the sentiment insights provided by a total interaction analytics solution, your entire organization will be better equipped to make strategic decisions to power business transformation.

Check back for our next blog to learn more about the breakthrough technology that powers this balanced sentiment detection approach.

{Image courtesy of Pakron/Free Digital Photos.net}

Categories: Choosing The Best