A Word Is Worth a Thousand Pictures

Napoleon said that one picture is worth a thousand words. In the case of real-time speech analytics, one word can be worth a thousand pictures. So, picture this: Every day, your agents speak to scores of people on the phone. The purposes of the calls vary, but nearly all have some sort of customer service objective. You probably monitor your agents' calls, either a percentage of them for quality or for compliance, or for other reasons. Buried in those recorded calls is a gold mine of information about customers, information that not only could be used to rate agents' performance, but that could be used for sales, operations, or customer service. Here's an example. You are an insurance company and one of your clients calls in to report a change of address. Your agent dutifully takes the address change information and alters the database to ensure that the customer address change is made. During the information exchange with the agent your client also mentions that the reason he's moving is that he just got engaged and he is buying a house with his fiancee. If this call has been recorded and reviewed, what could we learn? Not only that our client moved, but also that his insurance needs will drastically change, as will those of his new partner. This information could be converted into a sales opportunity--one that wouldn't cost the sales department a penny. If this call has been recorded but not reviewed, what could we learn? Nothing more than that the customer has a new address. The practice of speech analytics, or analyzing recorded calls for nuggets of information, is not new to the contact center industry. It has been around for several years, but only recently has this emerging technology and business development strategy been streamlined for current business conditions. Traditionally, speech analytics systems needed to translate the recorded voice to text, and then scan the text for key words or phrases that are meaningful to the user. The accuracy of speech-to-text translation is typically around 50 percent, so half the potential value of the recorded information is lost in translation. The analytics engine then attempts to interpret what has happened in the call by looking at the contexts around certain words or phrases. Unfortunately, looking at the contexts can't easily identify the value of the information. Doron Aspitz, the new CEO of Utopy, says, "The real value of a recorded call is in the things customers tell us while handling the transaction. Those nuggets generally tell us why someone is calling, and with that knowledge the enterprise can better act on the information." Utopy has developed one of the first solutions that actually mines the voice in real time, rather than converting voice to a form that's easier to process. Using acoustical templates and proprietary algorithms, Utopy not only looks for key phrases that may have been mentioned, it also identifies the sequence in which they were mentioned and flags the recorded transaction as one with additional value. The real value of speech analytics lies not in finding information useful to the contact center, but in finding information of value to the entire enterprise. Just as a useful CRM strategy must touch all points in the enterprise, a useful speech analytics strategy has the potential to benefit all points in the enterprise. The bottom line is that most agents today are not equipped with either the time or the tools to accurately capture the nuances found within most customer contacts. To depend on the agent to analyze the value of each call as it is transacted is both unfair and impractical. Valuable and usable information often lies undiscovered in recorded customer calls. Paul Stockford is chief analyst of Saddletree Research, which specializes in contact centers and customer service. Contact him at pstockford@saddletreeresearch.com