IBM Japan Business Services: Improving agent effectiveness by 88 percent

IBM Japan Business Services Co., Ltd. (IJBS) operates multiple contact centers on behalf of several clients, including IBM Japan. The organization’s PC Software Help Center handles inquiries from clients via phone calls, web and email. The PC Software Help Center in Ginoza, Okinawa has long faced a challenge of growing and maintaining expert customer service professional talent. Training and skill development programs had not kept pace with staff departures, resulting in problems with service quality. Additionally, the PC Software Help Center in Okinawa could not always find the highly skilled IT engineers it needed as service agents. Therefore, it was vital for the organization to develop a training program that would allow it to efficiently hire and grow inexperienced staff and contractors into skilled resources.

“At first we put together a list of responses to questions and a training curriculum based on the experiences and instincts of veteran agents,” says Hisashi Hasegawa, general manager of the IBM Japan Business Services Okinawa Center that oversees the Ginoza Contact Center. “But when the veteran agents were no longer working there, there was a lack of balance in customer responsiveness and in personnel training.”

To complicate the situation, the preexisting frequently asked questions (FAQs) based training program was not sufficient to train primary agents to effectively handle customer problems. As a result, complex inquires were frequently escalated to secondary agents—and the organization did not have enough secondary agents to handle demand.

“Training based on FAQs was completed in about three months, but after that there was no effective training in the handling of the difficult and wide-ranging questions that are received in real life,” says Kazuyoshi Miyazato, a group leader at the IBM Japan PC Software Help Center.

Using content mining to develop training curriculum

For staff, the question was: How can we effectively train personnel? To solve this problem, the decision was made to use IBM® Content Analytics. This content mining software is a product of machine translation technology that was developed at the IBM Japan Tokyo Basic Research Center (Yamato City, Kanagawa Prefecture), and it has advanced Japanese language processing capabilities. These include functions for effectively analyzing questions that come to contact centers and are stored in text and audio.

A large volume of data is stored at contact centers. This includes numerical data, such as handling times, which is the time required for an agent to handle a telephone call from a customer, and textual data in the form of call logs. For a single call it takes an agent about five minutes to enter such data.

“For 20 years, IBM database specialists had worked as IT architects, always looking for ways to use data effectively,” says center general manager Hasegawa. “Then they were assigned to contact centers to see what kind of data there was, and since so much time was being spent entering the data, they wanted to use it in a more positive way.”

However, database systems don’t usually have functions for effectively analyzing content. Therefore, in the past, the IBM Japan PC Help Center made the decision to use IBM Content Analytics software to greatly improve effectiveness. A project to introduce content mining was started in October 2007, and it was implemented in January 2008.

“Ease-of-use is important because instead of using specialized analysts, it uses the agents who deal with customers daily,” says group leader Miyazato. “It also incorporates Japanese language analysis technology that is the product of research and development in Japan.”

A focus on quality improvement

At the Ginoza Contact Center, in order to address an increase in calls and shortage of agents, a support system was set up with primary response agents to handle simple questions and secondary support agents to handle comparatively difficult questions. However, the primary support agents ended up handing over most questions to the secondary support agents, so the workload of the secondary support agents increased.

Then content mining was used to analyze the trends in the issues that were being handed over to the secondary support agents. A training curriculum was created to address the weak points that were identified and to better train the primary support agents. In October 2008, before the training, there were 202 issues that were handed over to the secondary support agents, while in May 2009, after the training had occurred, the number had gone down to 16.

Also, while the rate at which the primary response agents made responses on their own was no more than 40 percent in September 2008, after the training in May 2009 it was at least 60 percent, and it improved to a high of at least 75 percent.

“Insight into customer interaction logs is an information gold mine for us,” says center general manager Hasegawa. “Until now, content mining at contact centers usually meant analysis that was related to product questions, so analysis that is focused on agent training appears to be a new way of working.”

Adds agent Kubota of the IBM Japan PC Software Help Center: “By analyzing the activities of primary response agents, individual weak areas were understood and a training program was created to truly address these weak areas. The results were dramatic.”

IBM Japan personnel also analyzed trends in questions related to product bugs and were successful in identifying specific bug items. Analysis results concerning problems and expectations related to bug fixing were fed back to development departments. Until then, the number of bugs about which there were questions was about 66. After the feedback the number decreased to eight. “That result may be just in terms of the number of bugs about which questions came in, but if we think about the number of users involved, then we can be very satisfied with what a big effect it had,” says agent Kubota.

Additional content analysis planned

At the Ginoza Contact Center, where the introduction of IBM Content Analytics made personnel training more effective, additional uses of content mining are being considered.

“The workers at Ginoza have become accustomed to content mining, so I’d like to do other types of content mining in the future,” says center general manager Hasegawa.

Agent Kubota adds, “We may be a technical help desk, but after the product introduction we are also doing analysis for marketing. In this way, even in fields other than responding to requests from customers, we will probably use content mining more and more.”

Finally, group leader Miyazato expressed his hopes for the future: “In two years we have used data to analyze agent weaknesses and improve their skills, and we have used analysis to feed back the voice of the customer to development departments. At the contact center we will probably find new advantages in doing this, and use content mining to respond better to the voice of the customer.”

Business benefits

• 92 percent reduction in number of issues needed to be handed over to secondary agents
• 88 percent improvement in rate of calls handles by primary agent
• 88 percent decrease in number of calls on product bugs