December 2000

BY ART SCHOELLER

These days you can't read a single trade publication or go to a conference without hearing the word "CRM." But when you get past the buzz and get practical, the bottom line in customer relationships for most companies are the experiences that customers have with the company over time. And most of these experiences take place on the front lines...right in your call center. Still, customer experience information remains untapped and unexplored in call centers today. Why? Because in most cases, it simply isn't captured in the first place. Traditionally, call center monitoring systems have been agent-centric, recording random samples of agents' calls -- mere snippets of the customer experience.

These automated random monitoring systems served their purpose in their day. They eliminated many of the drawbacks to live monitoring and allowed supervisors to monitor and evaluate agents more easily, conveniently, and thoroughly.

But today, with the erosion of traditional sources of competitive advantage, service is becoming an even more potent differentiator. And service isn't just what the agent does. It's all of the many touchpoints -- both human and artificial -- that customers connect within their journey through the call center.

MANAGING CUSTOMER EXPERIENCES
Customer interactions are more customized and complex than ever. Consider that intelligent call center routing systems can now prioritize and direct different types of customers to the ideal agent based on virtually any conceivable criteria, ranging from the customer's specific need or inquiry to specific agent skill sets to the customer's corporate "value."

Increasingly, call centers are realizing that randomly monitoring calls reveals very little about these experiences. It does not help explain how to manage them better and be operationally effective at the same time. The question they are asking is no longer simply, "Did the agent do a good job?" The question has become, "How can I better manage customer experiences and be efficient and profitable at the same time?"

Dan Hawkins, research analyst for Datamonitor, recognized this trend in a recent report, noting "Call recording vendors must continue to develop applications that give real value to contact centers above and beyond simply recording calls for compliance reasons. Since recording calls provides businesses with a wealth of information about their customer interactions, one way to provide value is to help companies extract information from their recorded calls for business advantage. These efforts to increase the actionable insight extractable from the deadwood of unanalyzed calls should be applauded."

While agent-centric monitoring systems are ill-equipped for this new mission, complimentary technologies -- like recording and data mining -- are coming together to address this emerging need.

DATA MINING -- WHAT'S THAT?
Before we talk about how monitoring and recording can help call centers tap the vast wealth of customer interaction information and other data in the call center, let's review what data mining is, and isn't.

Data mining is the automated process of extracting unknown, comprehensible, and actionable information from data to make business decisions. Data mining is different from database querying in that the latter requires that the analyst or end-user typically frames a question and then queries the data to get answers to that question. Since database querying is a hypothesis-driven approach, if you don't know what to ask, you'll never find the answer.

Data mining, on the other hand, automates the process of knowledge discovery by analyzing large datasets and automatically extracting interesting trends and patterns in the data that a user wouldn't intuitively know to ask about. With data mining, you don't need to know what questions to ask to get answers.

Data mining techniques have been applied successfully in applications ranging from credit risk and fraud detection to customer churn analysis, market-basket analysis, and market segmentation. In fact, data mining tools can uncover hidden patterns, trends, relationships, and predictive indicators in any type of data.

One perhaps apocryphal but commonly cited example is the "beer and diapers" pattern. As the story goes, a supermarket mined their purchase data and found, much to its surprise, that these two products tended to be bought together. One interpretation of the finding was that new fathers probably tended to stay home more once a child is born -- and so picked up beer rather than going out for a night on the town. Why would the supermarket care? Because this knowledge can be applied to the store's merchandising efforts for more effective cross selling by simply putting the beer in closer proximity to the diapers.

Thanks to application-specific data mining tools that are easy to use and cost-effective, data mining (once the sole domain of scientists and artificial intelligence experts) is finding its way onto the desktops of line-of-business managers, and even call center managers.

In the call center, there are many potential data sources to tap. For example, today ACD reports provide a wealth of statistics and single dimensional data that help call center managers to analyze call patterns.

But perhaps the most valuable -- and to date, vastly untapped -- source of information is the customer experiences themselves. Customer-centric recording systems capture and preserve these experiences from start to finish. They don't just record the customers' conversations with various agents, but also capture the various metadata associated with the call, like hold times, transfers, call durations, agent IDs, and so on.

Capturing this customer experience information, along with the CTI data related to the calls, lays the foundation for leveraging powerful data mining tools to analyze customer experiences and other important operational trends in the call center.

DATA MINING AND MONITORING AT WORK
While the combined technologies of data mining and recording are opening up new opportunities to monitor and manage customer experiences, there will always be a desire to monitor and manage agents as well. In this area, monitoring and data mining has the potential to inject a new level of fairness into agent evaluation and scoring.

Research shows that multiple transfers and long hold times can create customer dissatisfaction. An agent who receives a caller who has already been transferred twice and put on hold for 20 minutes might not be scored on a level playing field with an agent who had an easier call. By mining the CTI data that describes the relative level of difficulty of the call (e.g., hold times, transfers, etc.), along with the agent's evaluation scores for that specific call, the data mining tool can automatically propose an adjustment to the agent's score based on the call's relative difficulty.

With turnover in call centers skyrocketing to 31 percent annually, and the estimated cost to replace an agent running close to $7,000 (source: Sibson & Company, a global management consulting firm), call centers need to do everything in their power to retain good agents. CTI-driven agent performance analysis has the potential to inject a new level of fairness into the monitoring and evaluation process. This level of fairness could potentially influence an agent's decision to stay or leave. And of course, holding on to good agents gives you a better shot at holding on to your good customers too.

CAPTURE THE EXPERIENCE
Monitoring and data mining can also point out areas where call centers can improve customer experiences and operational effectiveness. A case in point: using these combined technologies, a call center was able to identify that its "one-and-done" call transfer policy was actually having detrimental effects -- increasing call durations, costing money, and creating customer dissatisfaction.

When the data mining tool analyzed the center's calls, it found an unusual pattern. Calls that were transferred from one agent to another agent lasted three times as long as calls that were handled solely by the first agent. But when calls were transferred to a third agent, the call duration dropped dramatically. This was an example of a data pattern that a human analyst would be unlikely to detect since one would normally assume that the call durations would continue to increase as the number of transfers increased.

So what happened to the second agent? Because the data mining tool provided the ability to link directly to the recordings of the calls associated with the unusual pattern, the manager was able to listen to those calls and determine that calls were not being escalated fast enough by the second agent because of the "one-and-done" policy. Although the goal of the directive had been to quickly move calls through the center, it actually had the opposite effect: prolonging call times, costing the center money and resulting in a frustrating experience for customers.

Had the center not been able to use data mining to discover these relationships between transfers and call durations, and then perhaps even more importantly link to the recorded customer experiences, it is likely that this problem would have remained obscured forever, along with the opportunity to solve it.

And that's the very beauty of data mining... for the first time, call centers can identify hidden problems and opportunities in what was once simply "a deadwood of unanalyzed calls" without even having to know the right questions to ask.

Art Schoeller is vice president of marketing for Dictaphone, a Lernout & Hauspie company, and leading provider of recording and monitoring solutions for call centers of all types. Call 800.886.4908 for more information. 

[ Return To The December 2000 Table Of Contents ]