So You Got Yourself a Speech Analytics Solution. Now What?

Speech analytics is among the most useful innovations in call center technology, ever. If you already have speech analytics software in place, you don’t need me to tell you all the cool things it can do. But there are several things you’ll want to keep in mind in order to get the biggest bang for your buck. Choose your words carefully.

Choose your words carefully. Speech analytics software can analyze every call that comes into the contact center, identifying words and phrases said during calls, and automatically categorizing calls into topics. Having your system find whole phrases can be very effective, but remember you need to cover all variations. “How may I help you?” means the same as “How can I help you,” but because it doesn’t sound the same, won’t necessarily be categorized together. Similarly, your speech analytics software should be smart enough to know that “agree,” “agreed,” “agreement” and so on have similar meanings. Avoid common phrases like “I want,” “very much,” etc., as these would yield too many false positives. And keep in mind that many words sound the same. You may be looking for “confirmation” and find references to “information” instead. Fine-tuning the lexicons of each call category is crucial to achieving high-quality analytic results.

Whose line is it anyway? It’s not enough to know what was said during a call. It’s imperative to also take into account who said what–agent or customer. Let’s take the word “cancel” as an example. You could set up your speech analytics solution so that all calls in which the word “cancel” is said are categorized for further analysis and follow up. Sure, it will find all the calls where the customer said, “I want to cancel my service.” But unless you implement speaker separation, it will also find calls where the agent said, “Thanks for signing up. Remember, you can cancel any time.”

Read between the lines. Any decent speech analytics package can uncover dissatisfaction by identifying negative words and phrases such as “this is unacceptable.” Still, many calls will fall through the cracks—and the insights they could yield with them. Some people can be very polite and choose their words carefully even when they are extremely dissatisfied. In such scenarios, you need to augment word-spotting analysis with emotion detection. Take advantage of speech analytics to identify when emotions are running high, not only by spoken words, but also by subtle variations of pitch and tone.

Finally, let me fill you in on a little secret: speech analytics will not improve your operational efficiency. It will not improve your customer satisfaction, nor will it increase your revenues. Surprised? You shouldn’t be. Speech analytics can only show you the way. It will tell you what drives your repeat calls. It will identify agent knowledge gaps that, once addressed, will optimize average handle time and increase customer satisfaction. It will predict which of your customers are at high risk to churn. But it is up to you to take the next step. Leverage your speech analytics insights and then take action.

What other speech analytics best practices are you using at your contact center? Share your comments.

Share this:
Twitter LinkedIn Facebook Email