The kind of analytics we focus on in the upcoming February issue (see here for the preview we posted in December) is a method of sifting through and comparing data from sources that produce numerical information and reports, like the ACD.

The big difference with speech analytics tools is that the sources they draw from are recordings. Speech analytics is software that uses speech recognition technology to find certain words in recordings of calls. It can be used to look for specific words, or it can alert you to frequently occurring words. In a sense, this is a numerical process too: speech analytics takes unstructured information and finds patterns.

When I asked Kevin Hegebarth at Witness Systems about new trends in analytics, the first thing he mentioned was speech analytics, which is included as a part of Witness's Impact 360 suite.

"Every call that occurs between a customer and a call center agent contains a lot of really valuable information," Hegebarth said.

He continued:

"For example, an agent is on the phone with a particularly hostile customer. This customer is talking about how a financial service competitor is offering a better rate on home mortgages. The agent says there's nothing she can do about that, and the customer has their mortgage written by the competitor. There's some really valuable information in that interaction. One: the customer probably told the agent the competitor's name. Two: they told the agent the competitor had better rates, and maybe even told them the rate. That information is frequently lost. There's no easy way for the agent to capture that on the customer screen -- she may write some quick notes, but that's all she's able to do. So with speech analytics, with the use of software and speech recognition engines, that call can be mined. I could set up the speech recognition engine to look for specific competitor's names. Every time it 'hears' those names, it flags the call so I can go listen to it."

And if you want to make sure you don't miss anything, the software will look for trends, Hegebarth told me. "There are speech analytics products out there, ours among them, that can do proactive mining, so if you really don't know what you're looking for yet, you can have the software 'listen' to all these interactions and mine them and return a list of the most commonly used words and phrases to the user."

But simply looking for frequently occurring words doesn't always give you a full picture.
"Word spotting" is not enough, says Galit Belkind from NICE Systems. Belkind told me via e-mail about NICE's use of "stereo capture" in speech analytics. Stereo capture, Belkind says, is running separate processing on agent and caller, rather than one per conversation. This helps us find the context of the call.

"This kind of advanced analysis capabilities helps the supervisor understand whether 'buy again' signifies an up-sell opportunity ('with so many great features I will want to buy again'), or a customer at risk ('with such bad service I will never buy again')," Belkind wrote.

Tone and emotion are also important. Belkind: "Emotion detection is critical to identifying true customer intent and pre-empting defection. If a customer expresses dissatisfaction, this needs to be flagged in real time. And once this call is flagged, it should be queued and routed to a member of the management staff. The issue can then be reviewed, and the caller may get a call back, almost instantaneously. This would result in unprecedented responsiveness and customer loyalty."

In an article called The Call Is Your Most Valuable Asset in our September 2005 issue, Ingrid Spencer wrote extensively about using call recordings to help the enterprise. In one section, Spencer wrote about speech analytics being used to watch for angry callers. Here's a segment from that article:

The emergence of speech analytics tools, which mine recordings for anger and stress, is another key trend in call monitoring.

“Technologies that power voice-analytic applications are viewed by many to be experimental and in the early stages of adoption,” says Richard Parton, CEO of V Worldwide (Washington, DC), which distributes systems from Nemesysco, an Israeli company, that uses a technique known as layered voice analysis (LVA) to categorize calls. “However, those corporations pioneering the deployment of LVA are experiencing significant financial benefits, reduced risk, and in many cases competitive advantage.”

Parton cites the insurance industry as one that is benefiting from speech analytics. “In the insurance industry, initial trials of LVA-based voice analytic software have effectively reduced the payout of fraudulent claims by more than 30% when used in a call center environment,” he says.

“Our five-month experience with LVA has shown us that the technology is sound and effective in detecting and analyzing the truthfulness or untruthfulness of a speaker and the risk that a claim involves fraud,” said Chris Kvochak, general manager of Safeway Insurance.

Technological developments like speech analytics notwithstanding, call centers recognize that the difference between surveillance and evaluation. LVA for example, helps call centers pinpoint potential for customers to behave in a way that’s harmful to a company. But, in the end, the primary purpose of call monitoring is to enable the company to provide more value to, and derive more value from, its customers.