A Crash Course in Speech Analytics: How It Works, Why You Should Use It

Modern-day speech analytics tools are invalidating the one safeguard we had against robots taking over the world: the notion that machines can’t understand feelings. Emotionally intelligent AI analyzes customer sentiment to tell you whether you’re providing the best customer experience, but its use cases extend far beyond flagging calls as ‘good’ or ‘bad.’

Speech analytics tools can give you an indication of why your customers are calling, provide real-time analytics for agents and even monitor conversions. We spoke to two leading providers in this space, CallMiner and CallFinder, to find out more about how speech analytics work and why they’re worth considering.

Maximizing desired outcomes

As early adopters of speech analytics, debt collection agencies use the software not only to ensure agents are following strict compliance regulations, but to provide a better customer experience. Yes, even debt collectors care about how you feel - as it pertains to persuading you to pay up.

“We have customers that do debt collection on medical bills, which is a very sensitive topic, so what they’ve found is that customer experience matters,” said Scott Kendrick, VP of marketing at CallMiner, whose Eureka platform provides speech analytics. “If you treat the customer well they’re going to be more willing to pay the debt.”

In some companies, agents are required to deliver a compliance statement at the start of the call, such as a healthcare provider explaining patient confidentiality. Speech analytics software can alert the agent with a reminder if they haven’t done so within a reasonable time. “That’s based on the absence of language - so if the agent hasn’t said something yet that they should have said,” explained Kendrick.

The ability to monitor calls in real-time is even more important to organizations that provide legal advice or, say, guidance on insurance benefits. Sam Wilkerson, a quality assurance manager at Benefit Communication Insourcing, which provides businesses with HR benefits consulting for their employees, was tasked with overseeing agents who have to deliver information very precisely or be liable for damages if a client tries to file an insurance claim and there’s an issue with enrolment or compliance.

“One of the challenges is we really have complex conversations with the people we talk to,” said Wilkerson. “We have 19 different scripts given the scenarios we face with clients and employees.”

Most speech analytics tools provide full call transcripts and recordings, alongside post-call analysis that scores the quality of the call. Leading platforms provide real-time analytics that can send alerts to a manager or supervisor, or pull up a specific knowledge base article for an agent to view if a specific trigger word is detected in the conversation.

For example, if an agent is trying to upsell a product and a customer references a competitor’s product or promotion, the software can pull up a corresponding offer within the CRM system so the agent can counter with: “Well, we can do you one better. Here’s our latest promotion…”

Beyond helping call centers gauge typical KPIs like first call resolution, analytics software provides an empirical deep-dive into the question we’re all dying to ask our customers: How am I doing? After using CallMiner, one consumer goods manufacturer discovered a major design flaw in one of its products after the software flagged a recurring usability complaint as the main reason for a spike in call volume.

“They also had CallMiner analyze to quantify how big of an issue it is,” said Kendrick. “Is this something that somebody’s being noisy about on social media or are they getting a lot of calls and returns from customers?”

When it comes to sales, companies are constantly tinkering to come up with a winning strategy. By using CallMiner to review calls that ended in a successful sale, one of the largest sellers of security systems in the US created a blueprint to help its salespeople drive better sales. “They figured out what the structure of those calls were, when certain compliance statements were stated, what order did certain things happen,” explained Kendrick.  

Analytics tools can also help a business root out internal deficiencies within its customer service operation, such as a gap in training or excessive call transfers. “By employing analytics you can measure what the issues are, why customers are being transferred, why agents are taking more time to solve a transaction,” said Cliff LaCoursiere, VP at CallFinder.

A human-centric approach to CS

Speech analytics not only records and analyzes data but also suggests improvements and provides guidance to agents and quality assurance staff. The software can give agents automated suggestions and prompts in real-time and even help them coach themselves.

But at the same time, forward-thinking organizations are gunning for a more human-centric customer service approach by providing their agents with empathy training and prioritizing “soft” metrics like emotional connection with the customer informed by a constellation of data points. Doesn’t automating phone support processes run counter to this movement?

According to Kendrick, one of CallMiner’s first clients previously used manual call scoring, where a QA manager would tick items off a checklist. By discarding this rote system, agents had more freedom to express themselves authentically. “They found that when they moved to speech analytics it was actually easier for the agent to be less scripted and more natural because speech analytics is looking for concepts and topics.”

In fact, feedback from speech analytics can even cause organizations to pivot their CS strategy. “We’ve seen several customers make discoveries that they hadn’t anticipated along these lines and one example is where the call center measures average handle time as a metric but the data identifies that longer calls result in higher customer satisfaction or increased sales.”

Without the insights provided by the analytics tool, the business would have continued to assume that customers want shorter, faster calls just because it’s presumed commonsense in the industry.

Scaling quality assurance in the call center

For most businesses, their only claim to quality assurance at their contact center is an overburdened quality assurance manager who scores 5-10 calls per agent per month. Meanwhile, the industry average for supervisor to agent ratio is 20 to 1, and average handle time is 4 minutes.

“If you take those 5 calls per agent per month, that means that that supervisor is only spending 0.56 percent of their time listening to how the agent is performing,” said LaCoursiere. “Is that meaningful? Absolutely not.”

For businesses that provide consultative service, average handle time can be much longer, which makes it even more difficult for the QA team to keep up with call scoring. Wilkerson was scoring complex calls filled with legalese that often ran 25-50 minutes long.

After implementing CallFinder, BCI was able to score 100 percent of their calls. “When we were scoring only 10-15 percent of the calls it took a long time to identify what the trends are and what we need to address in training,” said Wilkerson. “Having the results at our fingertips has given us the ability to be more proactive rather than reactive in training new employees.”

Gamifying the call center for agent engagement

With an attrition rate of 20-35 percent, call centers churn at alarming rates because agents are bored in their jobs, are unclear about their objectives, or have no sense of progress or mobility. A speech analytics system helps gamify the process so that agents have a clear sense of how they’re doing from a dashboard that updates in real-time and provides post-call analytics for them to reflect on past performance.

The CallMiner platform translates those analytics into suggestions for improvement through the platform’s self-coaching tool. “It does gamify the process to a point where agents don’t want to be at the bottom of the leaderboard, they want to be at the top,” said Kendrick. Meanwhile, the ability to score calls in real time shortens the path to proficiency for new hires and allows them to get on the floor faster and take calls with real customers.