Text and Speech Analysis
Because every contact center collects agent notes, stores e-mail or Web chat threads, and records phone calls, there is a huge potential gold mine of “listening data” already available. A range of text and speech analysis tools (many of which were developed in response to security agencies’ need to tap phone calls and monitor huge quantities of data) exist to enable the capture of valuable information. Text analysis tools enable large volumes of free-format text to be searched for themes and trends around key words, or producing from scratch these themes and trends—for example, for such sensitive topics as “I am going to cancel my service” or contact coding, covered in Chapters One and Two. Speech analysis does the same to voice recordings, but in this case the customer’s sentiments and intentions are much more clear than through merely reading and scoring text.
Before text analysis can be effective, agents need to capture “the right stuff ” in the text. Analyzing millions of megabytes of data where agents have recorded that they have completed ID checks won’t deliver much insight. Agents may need retraining on what they should put in the notes so as to capture valuable information. The following list shows some examples of the kinds of notes that yield insights.
Notes Area Source of Value
Competitor information—for example, “Company B offered me x ” Product and value proposition research
“Do you have x?” “Can you make me y?” “Can my product do z?” Product design, campaign design, sales process redesign
Retention campaigns
Process comments—for example, “It would be easier if I could do b instead of c ” Process redesign
Channel comments—for example, “I tried to do this on the Web, but it wasn’t available” Channel design and growth
Customer emotions, including Retention and follow-ups
“This really annoyed me” Correlation between process and reaction
With these insights in the text, the tools can be used to search for themes and correlate these themes with associated customer behavior. Text mining can show, for example, that a certain type of inquiry is more prevalent among customers who are preparing to take their business elsewhere. One company identified that the apparently innocuous “When’s my next payment due?” inquiry was in many cases a warning signal that customers were considering offers from competitors. Prior to changing their notes strategy, the company had grouped all inquiries together as one type of note and could not glean this insight.
One of the major manufacturers of DVD players used text mining to discover that one of its players could not play a popular kids’ title from one of the big studios, after they had steadily told customers, “Sorry, your DVD is bad, so you’d better return it to the store,” and the store had to send them back to the studio. This manufacturer went to the studio with its tail between its legs apologizing for its error. (Despite this problem, the company in the end won considerable goodwill from the studio.)
Speech analytics can perform the same type of thematic correlation analysis on recordings of calls, highlighting emerging trends and themes and determining how important they are. Imagine being able to analyze how many times customers mention a specific competitor. Some of these tools are now being used in real time to spot a key word or pattern during a call and prompt the agent to ask for something he may have missed—in effect, these tools are acting as additional ears for the company during the calls.
No comments:
Post a Comment