Telecom New Zealand Gets Its Self-Service in Ship-Shape

Article Featured Image

With 35 self-service topics and more than 150 destinations, it would seem like Telecom New Zealand's goal of easing the customer experience is out of reach. Just the opposite; the company's new speech-enabled, self-service system is boosting customer satisfaction levels and saving customer service representatives time.

Telecom New Zealand's challenges stemmed from its touchtone system, which confused and frustrated callers. As the company's annual call volume hit 24 million, its products and services also expanded. Callers were not only experiencing inconsistencies with each new toll-free number they dialed, but also were hitting zero to escape the numerical maze of the touchtone menu. "The business behind the scenes was getting more and more complex with the services we were offering. It got to the point where the IVR that was the front to our customer care just got far too complex and we had messes of people zeroing out—above fifty-five percent," explains Hamish Stewart, manager of channel strategy at Telecom New Zealand (TNZ). "They were ignoring the IVR to get to an operator."

TNZ deployed TuVox's Perfect Router, a natural language call router, to handle calls about its landline, mobile, and broadband services. The solution is combined with Genesys Telecommunications Laboratories' skills-based routing technology that identifies the customer based on the phone number from which she is calling. "This makes the interaction shorter and more efficient," says Azita Martin, vice president of marketing at TuVox. Genesys also provided the CTI and screen-pop technology that transfers the information collected in the speech self-service module to the live agent if needed. The TNZ team, marketing, and contact center business members worked with TuVox to develop a persona that displayed the epitome of New Zealand-ness. The voice talents—both male and female— were narrowed and presented to C-level executives, staff, and stakeholders for review. The final selection was decided by TNZ's customers via focus groups.

The project was initiated in mid-2005, but was placed on hold while TNZ sorted out internal IT issues. "The thing with speech is that the quality of the call is so crucial," Stewart says. "You can't deteriorate the call; otherwise, you compromise the whole project."

While it wasn't exactly smooth sailing, TNZ expected a little turbulence as it integrated the new technologies with its existing system. "We did trip over a few difficulties, but we expected them and they were just delays in the project that had to be solved," Stewart affirms.

In late September 2006, TNZ initiated a pilot program where it recorded six weeks of live traffic and then had the agents listen to those calls. This helped train the agents to understand not only how the system works, but also how the caller interacts with it and how to help customers resolve common communication barriers when using the self-service system.

In conjunction with the pilot, TNZ also performed a usability study. That study enabled TuVox to measure the caller's perception of the TNZ brand, persona, and the speech application. The results assured TNZ of the persona's positive impact and revealed some error recovery changes that needed to be made to the system. This was the first of three phases of tuning that came about before the application was fully rolled out.

After gathering the six-week's-worth of live traffic, TNZ transcribed about 25,000 utterances and used that data to identify and integrate new destinations and router targets. The company used approximately 150 call recordings to analyze the callers' experiences of the self-service and to improve the overall call flow. The final phase is currently in development and is focused on improving recognition rates across the application. TNZ is once again using call recordings and transcribed utterances for the analysis.

The solution was fully rolled out in November across seven business units: residential, small business, mobile, faults (troubleshooting/repair), broadband help desk, credit and collections, and select advertising toll-free numbers. Customers are now greeted with How can I help you? The question works with open-ended dialogue recognizing what the caller says and transferring him to one of TuVox's 35 speech modules or to an applicable agent. "It is very open-ended and there are two thought processes on this. Some people think you always have to give directions, but we have a lot of customers like Telecom New Zealand that have so many products that listing all of the options is not the right design," Martin explains. "It is not going to provide the same level of efficiency that our open-ended dialogue provides."

The caller can control the prompts in the self-service application with commands such as wait, skip, stop, and so on. Being able to include all of these little concepts has given way to a sigh of relief for the TNZ team. "I am just impressed that we got what we set out to do," Stewart says. "On such a big project, you expect to have to compromise along the way, but it has delivered everything that I had hoped it would."


Usability testing was not TNZ's only measure for customer satisfaction with the system. The company also surveyed customers at the end of the calls. Live agents would request customer impressions of the speech applications from end to end. "We wanted to catch them when they went through the call so that they could reflect immediately on that experience," Stewart explains. "It has helped us ping some details of what the caller didn't like, then turn around and tweak the solution to fit that need."

TNZ also brought in a third-party research agency to perform weekly evaluations of customers interacting with TNZ's various contact channels. The percentage of customers jumping ship went from 55 percent zeroing-out with touchtone to 4.7 percent with speech self service. Customer satisfaction improved by 44 percent over the previous IVR. Speech also improved agent efficiency by increasing first-time call resolution from 76 percent to 90 percent.

"I'm finding that customers are much happier than they were using the old system. With the softpop we seem to be much more clued [in] because we already know what the customer wants," says Sandra Sangster, customer service representative for collections at TNZ.

Customer satisfaction isn't the only thing increasing with speech. There has been an 11 percent uptake in the number of customers who are completing their calls without ever speaking to an agent. "Telecom New Zealand believes that as more customers get used it, the adoption rate and the level of self-service is going to increase," Martin says. "One of the things they have felt very passionately about is ongoing tuning."

TNZ officials have been very impressed with the results that they have seen thus far with speech, so much so that they plan to automate the remaining DTMF destinations and other standalone applications for extended services such as broadband, mobile, and a customer information line.

"I fall off my chair sometimes with what speech science and recognition can achieve," Stewart asserts. "We've had calls where I have no idea what the customer said and the speech system recognizes what the customer wanted. It sometimes just defies belief."


SpeechTek Covers
for qualified subscribers
Subscribe Now Current Issue Past Issues