January 8, 1997
First, I would like to thank the IRS for their invitation to present this report. It is an honor to be invited to meet with you and I hope that my remarks help you in your important task.
Customer Service is an enormous challenge that is being faced by every enterprise in America today. While it seems that common sense would provide for easy design principles, it turns out that most organizations, even the largest and those with the deepest resources, often struggle to define an approach. They wonder how good their service must be to make a difference.
They puzzle as to whether the large investments that are often required will lead to greater customer satisfaction and, in turn, to greater sales and profits. In doing so, they sometimes lose track of the basic principles of customer service and get trapped by slogans, the latest hot technology, or, if it does not seem to be in conflict with the title of my presentation, "Best Practices".
Rather than start off with a list of Best Practices, then, I thought it would be critical to define what really is important in customer support. From this perspective, Best Practices are simply the tools that organizations use to reach these levels of performance. The important thing to note is that you should not design customer support centers with a goal to include Best Practices, but, rather, to provide effective customer service. Best Practices are a means to an end, not the end themselves.
What are the keys to effective customer service? The following is a short list. If these principles are met, your callers will be satisfied and you can be assured that you are meeting your responsibilities:
BASIC CUSTOMER SERVICE PRINCIPLES
PRINCIPLE DESCRIPTION
1. Near-Immediate Access Callers should be able to get through to the customer support system the first time they call and in less than 45 seconds to a minute. The most often-quoted goal in industry is that 80% of calls should be answered in 20 seconds or less. The least supportive service level objectives we have seen in the last two years anywhere have been 80% in two minutes.
2. One and Done Callers should have their questions answered on the very first call if they have all of the information needed by the support center to address their questions. Calls should not have to be referred for research or later follow-up for any reason. Referrals to more knowledgeable agents should only occur occasionally. When these referrals must be made, the transfer should be made to another agent immediately, often with the first agent still on the line. In order for this to work, agents must be properly trained and technology must be available to provide the agents with access to any information that will be required to answer the callers' questions immediately.
3. Immediate Follow-Up The support center should be capable of making customer record changes immediately, changing rulings or judgments or ratings etc. without any needed additional steps and should be able to order the dispatch of requested documents, brochures, instruction manuals, etc. while still on the phone.
In short, callers should be able to reach the support center almostimmediately, receive accurate and informative answers to their questions and, when necessary, obtain follow-up materials and changes to their records as soon as possible.
The IRS does not seem close to meeting these goals. The April, 1995 report directed to the Oversight Committee on Ways and Means indicated that as few as 54% of callers get through to the IRS after numerous calls, and much less on the first call. Data was not available that directly address One and Done and Immediate Follow-up. However, we understand that not all agents have computerized access to caller records. If this is true, it would be impossible to consistently meet these goals.
Much of the research and development being done in the customer support field is directed at developing technologies, processes and people management procedures to attain these goals. However, not all Best Practices are appropriate for all circumstances. Rather, call center architects must identify the needs that their callers have and the technical and resource constraints under which their organizations work. They must then select the optimization strategies that are most aligned with these drivers.
In this light, call centers can often be classified by the degree of caller-specific information that an agent must have to answer the callers questions. This can be referred to as personalization. Non-personalized calls typically involve questions about general rules or procedures that, for the most part, do not change regardless of the callers situation. An example of this is the filing deadline. It is the same for virtually everyone and can be quoted by the agent even if he or she has very little information about the caller. The same could be said about most of the procedural calls (How do I do...) that the IRS receives.
Call centers that answer non-personalized calls should utilize the following Best Practice components:
1. Agents should have rapid (i.e. computerized) access to descriptions and examples of the rules, procedures and facts that are necessary to answer these kinds of calls.
2. Artificial intelligence-based search engines, Frequently-Asked Questions (FAQ) lists and agent-directing scripts should be available to assist the agents to identify the information that the callers need.
3. Given the above, it is imperative that every agent be equipped with an intelligent terminal that can support these functionalities.
4. If callers can generally identify the nature of their questions, then an
Interactive Voice Response Unit (IVR) should be used to ask the callers to
identify their needs. For example, a simple system might ask them if they are
calling to ask about IRS procedures, specific subject matter issues such as
deductions, etc (There are specific rules about the number of levels and number
of alternatives that should be included in the IVR script). However, the
overall purpose of the IVR should be clear - to obtain information about the
caller~s needs and pass that information to computerized systems. These
systems can pre-warn the agents, possibly pull up appropriate scripts or FAQ
lists, and/or direct the calls to the right subject matter experts.
5. Expert routing, a component of the process that was described above, is
also important. It directs calls to agents who have received extra training
in specific areas or who have access to specialized information. This gives
the agents the best chance of answering the calls quickly and accurately.
The overall purpose of these strategies is to shorten the length of time that
agents take to satisfy callers as well as make their answers more accurate.
This will result in agents being able to answer more calls per day and will
therefore increase the access level experienced by your callers.
Highly-personalized calls are those that address specific aspects of the
caller~s situation. For example, callers may be inquiring about the status of
an appeal or wishing to know whether they had properly depreciated a business
property on their last return. These types of questions require all of the
technologies listed above. However, they also require that the agent can
retrieve relevant portions of callers~ personal records. This might be a copy
of last year~s return, images of correspondence between the taxpayer and the
IRS, or copies of previous rulings on the caller~s returns.
This retrieval of caller-unique information is enormously complicated and
involves highly-sophisticated Information Technology. Technologies required
include image systems, large databases and complex middleware to connect all of
the pieces. They must be connected by high-speed networks that can move
specific information elements to any of the IRSs service centers located
throughout the country. The careful combination of these technologies can
create a technology-enabled customer service center that both provides
excellent service to the caller as well as handles more calls because each call
can be handled more rapidly.
While the technologies listed above are critical to call handling, there are
another set of Best Practices that have been shown to improve call center
performance by 15% - 25% with a relatively small investment. These best
practices maximize agent productivity. They can ensure that a call center will
attain performance levels equal to private industry and, almost certainly, can
substantially increase the number of calls that IRS call centers can handle
without adding more agents.
The technology and procedures required to implement these processes include:
FORECASTING/SCHEDULING/ADHERENCE
TECHNOLOGIES AND PROCEDURES
TECHNOLOGY DESCRIPTION
Forecasting / Scheduling Packages Forecasting/Scheduling packages utilize past
call arrivals to predict future call volumes. Advanced packages provide
predictions by + hour for every day of the year. They print out the number of
agents required to meet any service level (i.e. 80% of calls answered in 20
seconds) requested. Scheduling subsystems then populate the schedules with the
names of specific agents who will be assigned to work those time periods.
These systems help call centers to schedule just the number of agents required.
Adherence Systems Adherence systems monitor agent compliance with their
schedules. They note whether agents are signing on and off when required and
also note whether they are answering as many calls as projected, taking an
expected (average) amount of time to handle the calls and no more than an
acceptable amount of time to complete after call work activities. Adherence
systems are essential to protect the integrity of the forecasting/scheduling
process so that service levels can be attained. Newly-evolving adherence
systems are beginning to be built into or offered as optional modules of
forecasting/scheduling systems as well as certain customer support systems.
Specialized Personnel Management Procedures The mere implementation of
technology-based management systems has almost no impact whatsoever on overall
call center performance unless management backs them up with carefully-designed
procedures. These include reviewing service levels, accuracy of forecasts,
adherence, call lengths and after call work lengths daily, with summaries
passed to senior management weekly, monthly, etc. Minimum levels of performance
must be established for all personnel and enforced through the review of data
generated by these systems. Senior management must totally dedicate itself to
establishing the appropriate service standards, training group leaders in their
use and backing them up when they are used. A good starting point would be the
following, assuming that little outbound calling is done:
z Required Sign-On Time: 7 Hours/Day
z Required Talk/After Call Work Time: 6.5 Hours/Day
z Number of Calls Handled: As predicted by the system
z Acceptable Caller Wait Times: 80% In 60 Seconds
z Acceptable Average Percent Abandons: 5% - 10%
The key thing to recognize about the Best Practices listed in the table above
is that they work together. The technology is used to generate forecasts and
schedules that will provide the best possible service and the highest possible
levels of agent utilization. The management routines communicate the
importance of these standards and protect them through performance review,
training and, if necessary, appropriate sanctions. At Ernst & Young, we call
this a Fused approach, since it combines technology improvements, proper
procedures and careful attention to people~s capabilities.
While no improvements come inexpensively in call centers,
forecasting/scheduling systems and their associated management improvements are
much cheaper than any others. They often can produce greater percentage
improvements. It appears from a very cursory review of the IRS~s call centers
that realistic performance standards, strict schedules and strict after-call
work expectations have not been fully analyzed and implemented. For example,
one document mentioned that agents are given a target number of calls to reach
each day and then can quit. The report implied that some agents may hit their
targets and then cease taking additional calls. Why wouldn~t the agents be
asked to continue work until the end of their shifts, given the number of calls
that are not being answered?
While it is impossible to determine from the documents that we reviewed, it
seems unlikely that some current procedures were based on real observations of
what agents can do. It also seemed unlikely that individual schedules tied to
historical call arrivals had been built. If so, the IRS has a tremendous
opportunity for improving performance with a relatively modest outlay. For
example, advanced forecasting/scheduling systems can be obtained for
approximately $100,000 per location, not including current state review,
planning, initial scheduling and implementation expenses. By contrast,
effective customer support systems, including new PC~s, artificial
intelligence packages, knowledge engineering, middleware and other required
outlays can add up to more than $10,000 per agent, not including similar
implementation costs.
While this report has been extremely brief and the amount of information
available to prepare it less than would be required to offer recommendations
with complete confidence, the information we have been able to gather suggests
a clear course of action. We suggest that the IRS carefully study its agent
efficiency levels to determine if additional calls could be answered by its
current force if more effective forecasting/scheduling technologies and
procedures were in place. If so, then the IRS should either procure advanced
forecasting/scheduling systems or review how they are using their current
systems. It should develop appropriate management procedures to ensure the
proper utilization of its agents and implement the new systems as rapidly as
possible.
Only a detailed review of current performance can ensure what degree of
improvement can be obtained. However, given the budgetary constraints that the
IRS currently faces, the use of this set of Best Practices might be the most
promising approach to increasing service at the least possible cost.