 |
||||
|
|
|
||
|
Contents 
Introduction
Servcorp
Ausmelt
Suncorp
How to be an innovator
Executive Summary |
|
||
|
|
||||
|
|
||||
The implementation might have been regarded as low risk, but there was a risk to the organisation in terms of the large number of customers that would be touched by the new system. Suncorp's large range of products and services meant that traditional touchtone IVR ("for balance enquiries, press 1...") was frustrating for customers. Consequently, much attention was paid to usability and customer trials.
The first step was to collect the different utterances used by real customers by running the VeCommerce voice engine but having human operators in the background to route the calls. That information was applied to an initial implementation that was subjected to usability testing where it received a higher rating than the old IVR system. Performance Technologies Group carried out the testing and benchmarking for the project, which was managed by SMS Consulting.
Next came a 300-user pilot that took calls from a particular region, and after further adjustments a 5000-user pilot covering various locations including Melbourne and Sydney. By this stage, customer ratings had reached 3.9 out of five compared with 3.4 for the IVR system, but recognition accuracy hadn't reached the target. A seven-day all-customer trial provided sufficient calls to take the accuracy over 94 percent, and the system went live on 11 December 2002.
Instead of having to plough through a series of IVR menus, customers can now simply state what they're calling about, for example, "claim for a broken window, thanks."
The system incorporates more than 60,000 phrases derived from caller conversations, and uses skills-based routing to direct the calls to an appropriate agent.
One of Mulvogue's concerns was that the persona of the system would be consistent with Suncorp's values. "You can get it running so well that the customer thinks they are talking to a human being," he says, but adds that it is not desirable because it could lead to embarrassment for the customer.
To avoid discouraging customers from using NLSR, if the system twice fails to understand a request, the caller is transferred to a live operator and prioritised to reach the right queue.
The project was successful and met its goals: the average time to reach an agent was reduced from 70 to 30 seconds, and misdirected calls fell from six percent to 2.5 percent. Among other benefits, this led to a saving of over 2.6 million telco minutes a year.
Mulvogue believes an important factor was that the project was owned, led, and driven by the business as part of an overall vision to improve customer service.
Over the last year, most of the focus has been on developing agents' skills, knowledge, processes, and authority to deliver first-call resolution, rather than on the technology. For the cases where that isn't possible, Suncorp is planning to add the ability for callers to ask for the same agent when they call back regarding the same matter and they will be given the choice of waiting for that person or speaking to someone else based on the expected wait time.
1%
1%
