Voice Analytics = VoIP's Big Data Play

A lire sur: http://it.toolbox.com/blogs/voip-news/voice-analytics-voips-big-data-play-60950

Large call centers conduct tens of thousands of calls per day. Although a typical customer likely views those calls as a necessary evil when opening an account, getting a problem fixed, or ending the relationship with a company, every call is usable business information. Using Voice over IP (VoIP) and analytics tools, enterprises are turning a flood of raw data into valuable intelligence to reduce costs, improve business processes, gain near–real-time insights into competitive threats, and increase customer satisfaction.

VoIP provides the framework for capturing voice for recording, with the ability to take calls from any standard call center setup and essentially pipe it into another vendor’s solution for more detailed processing.

Voice Analytics 

Voice analytics, also called speech analytics, dissects the spoken content of conversations. The now-standard “Calls may be monitored or recorded for quality assurance purposes” notification indicates that the call center is recording the conversation between a customer and an agent. Take that conversation, multiply it by the number of calls per day that the call center processes, and you get a hefty amount of raw audio data.

After calls have been recorded, raw audio can be processed in a couple of different ways. Speech can be broken down into phonemes, the small pieces of speech, and then indexed to provide a quick text search for specific words or phrases. Managers can quickly skim through the index, playing back particular call instances.

Calls are also translated from speech to text to provide a detailed transcript for more sophisticated data mining to find frequently mentioned topics and the context in which topics are mentioned. For example, a company can look through its calls to find the word cancellation or return, and then conduct a deeper search for the mention of a specific competitor or product.

Emotion Detection 

Emotion detection can be a part of voice analytics through the use of key words (think profanity) and by analyzing the pitch and tone of a speaker’s voice. Rapid changes and spikes in tone and volume typically indicate an unhappy customer. A group of emotional calls associated with a particular agent might indicate the need for more mentoring or other corrective action, while an uptick in emotional calls spread across agents could indicate a breakdown in a product or service offering.

Speech Analytics at Avaya 

Avaya is “drinking its own champagne,” using its Speech Analytics suite at its U.S. Global Support Services (GSS) center in Highlands Ranch, Colorado. By using automation, 100 percent of all U.S. inbound calls—more than 3,400 hours per month—are now monitored, with supervisors flagging specific calls for review. Previously, only 2 percent of calls—about 64 hours a month—were manually screened.

After implementation, the Avaya GSS center showed a 40 percent increase in appropriate discussion for the use of the company’s self-service website and a 22 percent increase in the use of the proper greeting. The organization maintained a 99.9–100 percent rating per month for zero negative feedback on agents’ handling of calls.

Speech Analytics at American Airlines

American Airlines uses several NICE Systems voice analytics packages to process the input of 140,000 calls per day made to nearly 5,000 agents. The airline uses NICE Interaction Analytics in a variety of ways, such as gauging customers’ reception of a new product and being able to spot potential scheduling problems as customers call in trying to rebook flights after weather issues. Interaction Analysis is also used to measure how well the airline’s customer service organization and processes are performing to ensure that agents have the tools and information they need.

In addition, American is adding dollars to its bottom line. The company said that it increased revenues from its hotel partners by 25 percent by identifying the best words and phrases used by agents when conducting call transfers to those partners, then revising call scripts to incorporate those best practices.

Big Data Potential 

Because of the sheer volume of data and the expense of storing and processing speech, voice analytics has primarily been focused on dedicated call center implementations. As cloud-based VoIP solutions proliferate, it is likely that voice analytics will become an a la carte option, enabling more companies to tap into thebig data potential of their inbound phone calls.