What's Next In Monitoring Technology? Data
Mining Finds A Calling In Call Centers
BY LINDA DILAURO, DICTAPHONE CORPORATION
With over 70 percent of business transactions taking place over the
phone, call centers are important cogs in the customer retention wheel.
Consider also that customer interactions are becoming increasingly
complex. Today, companies are likely to segment their customers into more
categories than there are flavors of Ben and Jerry's ice cream, and route
customers to the ideal agent based on virtually any conceivable criteria.
Given the growing complexity of customer interactions, how can a call
center get to the root of the problems that can cost it money and
customers?
Typical monitoring systems that employ either a random or rules-based
approach to monitoring increasingly do not seem to be the answer. Such
systems, by design, either miss problems or force the manager to magically
guess what the problems are.
Let's explore the evolution of monitoring systems and see how call
centers can leverage customer-centric monitoring technology and data
mining tools to automatically identify hidden patterns in call center data
that can point to opportunities for improving customer service and
operational effectiveness.
The Evolution Of Monitoring Technology
In the early days, monitoring took the form of over-the-shoulder agent
observation, and then evolved to live monitoring of agents' calls using
the service observance feature of the call center's ACD.
While this "service observe" approach was more efficient, it
had its own set of drawbacks for call center supervisors, who still had
difficulty finding the time just to "listen in" on a handful of
calls during the course of a busy day. The supervisor was often forced to
stay on the line through extended "dead time," waiting for an
agent to begin a call. Then, once the call began, frequent interruptions
could force the supervisor to start the process over again.
The introduction of the automated monitoring system -- complete with
voice and data screen recording, built-in evaluation forms and reporting
packages -- made monitoring more efficient and eliminated many of the
drawbacks of live monitoring. Automated monitoring systems recorded agent
conversations so supervisors could go back and "QA" calls at
their leisure. A CTI link to the call center's switch let supervisors
program the system to record a percentage of each agent's complete calls.
Second-generation automated systems even allowed call centers to use
CTI-driven record commands to intelligently target which calls to sample
(e.g., calls associated with a specific DNIS).
In addition to saving time, this technology also helped supervisors be
better coaches, because they could now use recorded calls to show agents
what worked best, rather than just telling them. It also removed the
biases inherent in monitoring and evaluation because multiple supervisors
could now listen to and score the same recorded calls, and then compare
their results -- a process called calibration.
But this technology had limitations, too. It only recorded individual
agent conversations -- not complete customer experiences. If a customer
was transferred several times, the first segment of the call might be
recorded, but the customer's experience as he passed from one agent to
another, then possibly another, would be lost. By monitoring the first
conversation, the supervisor might conclude that agent Number 1 did a good
job -- but what he couldn't know was that the customer's overall
experience was dismal. Absent this information, the call center would be
doomed to repeat the mistakes of the past.
Last, because these systems were designed to record a random sample of
agent calls for monitoring (and throw away the rest), it was easy to miss
many of the problems that could lead to customer dissatisfaction.
Customer-Centric Monitoring Technology
Next in the evolution of monitoring technology was the
customer-centric monitoring system. The underlying premise of this
technology was to create a customer experience recording and analysis
tool.
The customer-centric monitoring system captured the various touch
points of the agent/customer interaction -- IVR menu selections, hold
times, transfers, multiple conversations and the like -- and then tied all
of the pieces together for a complete "cradle-to-grave" customer
experience.
This approach viewed recorded agent/customer interactions not as random
events, but as a pool of information to be mined. Instead of randomly
recording calls, the system recorded all interactions. Then, using the CTI
data captured along with the recordings, supervisors could identify the
calls that were likely to result in a bad customer experience -- for
example, calls with long hold times or durations, or excessive transfers.
By mining customer interactions intelligently, call centers could directly
address the root causes of customer dissatisfaction. For example, by
listening to calls with excessive transfers, a supervisor might learn that
an agent was not equipped to handle the call (a training issue) or that
the customer didn't reach the right agent in the first place (a call
routing issue).
The only limitation of this approach was that it relied on the
supervisor to use his or her intuition to determine what types of
interactions were likely to be problematic.
It's true that what you don't know can hurt you, but the opposite is
also true. What you don't know can help you, but you have to know how to
spot it. Surely, there are problems that the call center supervisor would
never uncover relying solely on his or her intuition.
How to solve this perplexing problem? The answer can be found in the
marriage of customer-centric monitoring and data mining.
Data Mining: What's That?
To understand how data mining techniques can be effectively applied to
call center monitoring, let's first understand what data mining is.
In its simplest form, data mining is a way of giving meaning to data, a
method of transforming bits and bytes of data into knowledge, which, when
acted on, can produce strategic business results. Data mining tools can
uncover hidden patterns, trends, relationships and predictive indicators
in any type of data. In the call center, there are many potential sources
of data that could provide rich insights into customer experiences and
operational effectiveness.
While data mining is now emerging as a popular technology, the
widespread use of data mining was slow to take off, primarily because the
technology had been geared toward analytic researchers and experts, not
toward mainstream business.
A recent report by GartnerGroup predicts explosive growth for data
mining. "In the next decade, the number of data mining projects will
grow dramatically (over 300 percent) to 'improve customer relationships'
and help enterprises 'listen to their customers.'"
GartnerGroup classifies data mining tools into six key categories,
ranging from the generic, application-independent tools (the best-known
class) to application-specific tools. Generic tools offer a wide range of
data mining techniques and can address many different types of problems,
but they require skilled statisticians to prepare the data, determine what
analytical techniques to use and validate the results.
With the introduction of application-specific, user-friendly data
mining tools, data mining is no longer just the domain of PhDs,
statisticians and computer scientists. Application-specific tools are
designed -- as the name implies -- with specific applications in mind.
Less expensive and easier to deploy than generic data mining tools,
application-specific tools are customized, turnkey solutions that require
little expertise on the part of the user.
Practical Applications Of Data Mining And Monitoring
Embedding application-specific data mining tools within the
customer-centric monitoring system is an extension of the philosophy that
customer interactions, and all of the associated data surrounding them,
are not mere random events, but information to be mined.
How can call centers benefit from the marriage of these two
technologies? Through a querying interface, a call center supervisor with
no statistical expertise can ask "what if?" questions of call
center data to identify hidden patterns that can point to operational and
customer service problems. Once these patterns are identified, the
supervisor can immediately listen to the associated voice recordings to
drill down to the source of the problem.
Following are a few examples of applications of this technology in the
call center.
Example 1: Mining CTI data for operational
effectiveness.
CTI call pattern analysis can help the call center maximize productivity
while sustaining high-quality customer care by improving training and call
processing practices. For example, analysis of a test site's CTI data
revealed a counter-intuitive trend that caused the call center supervisor
to reexamine call escalation procedures. The data showed that calls that
were transferred three times were actually shorter in total duration than
calls that were transferred twice. Listening to the associated recordings
pointed out that although the goal of the "first call
resolution" had been to quickly move calls through the center, its
actual effect had been to prolong call duration, costing the call center a
great deal of money and potentially frustrating customers. A new
escalation policy or additional training of first-level support
representatives was needed to solve the problem.
Example 2: Mining CTI data and agent evaluation scores for a
true picture of agent performance.
Historically, call center supervisors have evaluated agents based on
theoretical responses associated with an ideal call. However, the path
that certain calls take through the call center (e.g., the number of times
the caller is transferred) can influence the relative difficulty in
handling the call for each successive agent.
By mining the CTI data that describes the path of the call -- hold
times, transfers, etc. -- along with the agent evaluation scores for
specific calls, the data mining tool can propose an adjustment to an
agent's score that would reflect that call's actual difficulty (based on
the customer's prior experience in the call).
CTI-driven agent performance evaluation can inject a new level of
fairness into the monitoring and evaluation process. It could also be used
by call centers for automated calibration analysis -- to see if calls of
similar types tend to be scored the same or differently. Last, call center
managers might use this information to create new incentive plans to
reward agents who consistently take and resolve calls of higher
difficulty, since this will now become visible.
Example 3: Mining customer experience evaluation data.
In many companies, 90 percent of the revenue is generated by 10
percent of the customers. The use of customer experience assessment
statistics (customer profile data, and data describing the quality of the
customer's experience in terms of routing, transfers, hold times and
stress levels) provides another layer of data for analysis. By mining
these data, call centers can determine if they are providing adequate
levels of services to customers based on the same criteria they use to
route calls. They can listen to actual recordings of interactions to
determine how to do a better job next time.
Future Applications Of Data Mining And Monitoring: As Limitless As
Everything You Don't Know
The future applications of these combined data mining and
customer-centric monitoring technologies are even more compelling.
Soon, call centers will be able to incorporate "outcome
variables" into their data mining analysis. This might include
capturing customer rating scores from a back-end IVR application. For the
first time, call centers would have a true, real-time assessment of a
customer's experience (provided by the customer), linked to an actual
"cradle-to-grave" record of that customer's experience,
including the voice recording. Call centers would have a tool for
understanding what types of customer experiences actually produce
satisfaction. Listening to the voice recordings of highly customer-rated
interactions (as well as poorly rated ones) would let them confirm or
refute their hypotheses on how to adjust their strategies to win and keep
customers.
This information could also be mined relative to other outcome or
customer segmentation data captured from the call center's various CRM
systems. These data could come from external sources such as account
databases or screen data field entries. By mining these data with other
data sources, it would be possible to analyze what types of customer
experiences produced desirable or undesirable business outcomes. For
example, was the customer's complaint resolved? Did the call result in a
sale or cross-sale? Or did the customer close his account? If so, why?
One of the most powerful planned uses of embedded data mining
technology will be to intelligently push information to upper-level
management within the call center for immediate review and response.
Imagine being informed by your monitoring system when key customers are
put at risk due to poor call center experiences. With this knowledge comes
a chance to repair the damage and retain key customers, a critical
capability given that it costs far less to retain a customer than to
acquire a new one.
So, what's the last frontier for this new combined technology? How can
it help call centers bridge the customer experience information gap and
improve operational effectiveness and customer satisfaction?
The answer lies at the very foundation of data mining itself. It's what
we cannot guess and what we do not know.
Linda DiLauro is a senior marketing manager for the CRS division of Dictaphone
Corporation, headquartered in Stratford, Connecticut.
|