While working on a maintenance project for a large U.S. telecommunications firm, information technology company Infosys Technologies Ltd. discovered that an unacceptably high percentage of calls in the clients’ automated payment system were being routed to the call center agents. The team conducted a DMAIC project, outlined here, to reduce the number of misdirected calls.

Define

The scope of the Infosys maintenance project was to provide end-to-end support to maintain and enhance usage billing applications for this telecom client. During the maintenance project, the Infosys team discovered that about 120,000 calls, or 70 percent, in the client’s payment interactive voice response (IVR) application were being routed to the call center agents instead of being sent to the automated system. The goal established by the team was to reduce the misdirected payment IVR call transfers to the call center by at least 15 percent (around 20,000 calls). Addition or removal of self-service options in the automated system were considered out of scope for this improvement project.

The drivers for improvement in the project were to:

• Improve call routing effectiveness and reduce misdirected callers.
• Improve the customer experience.
• Improve the efficiency of the call center agents.
• Improve business metrics of the payment IVR application.

Measure

The team determined the following critical success factors for the project:

• The number of direct calls (i.e., calls using the IVR’s 800 number) made to the payment IVR getting misdirected to call centers
• The number of indirect calls (i.e., calls getting transferred from other IVR applications) made to the payment IVR getting misdirected to call centers
• The number of misdirected callers getting appropriate behavior after the change is implemented

As part of the project’s measurement system validation strategy, all data that was captured and used for analysis was verified with the client and validated with their in-house reporting system. Initially, data was collected for six months. Later, data was collected for another six months and was broken to fortnightly data points for post-improvement comparisons.
There was not much variation in the data trends, so this data was considered for analysis and calculating benefits in the Measure phase.

Analyze

Normality analysis for the input call data was conducted and data trends were analyzed for a one-year period. The results showed that the data belonged to a normal and homogeneous set. Additional analysis was performed to determine whether other misdirection issues (e.g., no account found, authentication failure, hang-ups, etc.) were the cause of the payment IVR problems.

The team also completed Pareto, root-cause and 5-why analyses to discover the exact root cause of the misdirects. Out of the multiple causes identified, following are the two main reasons leading to the misdirected calls:

1. Customer account information was not present in the back-end system. Due to a corporate merger, users were migrated to some other service or IVR but were not aware of this change. As a result, users were calling at wrong service/IVR where their accounts did not exist. Users also were unsure whether to enter a billing telephone number (BTN) or a working telephone number (WTN). User information was fetched using BTNs, but in cases where users entered WTNs, it might lead to no information or wrong information. When users called for a query related to their accounts after disconnection of service, this information did not always exist in the back-end system.

2. Customer authentication failure. Under the original authentication methodology, users were asked to submit their previous bill amount (including cents). This is a more error-prone system because it introduces human error into the system, with some users inadvertently entering incorrect information.

Improve

The solution that was implemented was the introduction of a check within the payment application for the unintended callers (callers with other services not serviced by this payment application) and a re-route to the appropriate IVR application. For direct callers, the check was placed in the payment application, and for indirect callers, the check was placed in another application directing calls to this payment application.

Control

Six Sigma improvement suggestions were implemented and the post-improvement phase clearly shows the results (see Tables 1 and 2).

Table 1: Indirect Caller Improvements

Customer Account Not Found	Indirect Calls
	Pre-improvement	Post-improvement
Number of tickets	496,881	29,179
DPMO	490,105.78	116,478.38
Percent defective	49.01%	11.65%
Yield	50.99%	88.35%
Z score	0.02	1.19
Median	22,209	3,646

Indirect call improvement highlights:

• The total number of indirect calls has been reduced. (There has been reduction in the customer base.)
• There is a significant reduction in the percentage of calls going to the call center pre- and post-improvement, as highlighted in the yield.
• The process capability has considerably improved as indicated by the improved Z score.

Table 2: Direct Caller Improvements

Customer Account Not Found	Direct Calls
	Pre-improvement	Post-improvement
Number of tickets	188,150	8,761
DPMO	154,392.66	68,521.33
Percent defective	15.44%	6.58%
Yield	84.56%	93.15%
Z score	1.02	1.49
Median	8,789	2,175

Direct-call improvement highlights:

• There was a slight reduction in the percentage of calls going to the call center pre- and post-improvement, as highlighted in the yield.
• The process capability has marginally improved as indicated by the improved Z score.

Comparison of pre- and post-improvement shows significant improvement in percentage of misdirected calls to the agent (see Table 3).

Table 3: Total Misdirected Call Improvements

Parameters	Pre-improvement	Goal	Post-improvement
Direct callers	15.44%	10%	6.85%
Indirect callers	49.01%	36%	11.65%