Incoming Call Management: Past, Present and Future

Many US businesses derive their main revenues from managing and servicing incoming calls. These calls may deal with customers ordering products, obtaining services, making reservations, lodging complaints, inquiring about individual bank accounts or loans, and a host of other revenue-producing activities. Special agents or attendants are employed to handle these revenue-producing calls. Those business calls personally answered by the regular employees of a company are not considered in this article.

Depending upon the size of the business, revenue-producing calls may be distributed to the special agents via an attendant console (for example, a multi-button key telephone system), or a call sequencer, or a uniform call distributor (UCD), or an automatic all distributor (ACD). Incoming calls tend to be characterized by random arrivals when observed over short intervals of time (short compared to a call holding time). Call volumes observed over periods of a half/full hour (a long period compared to a call holding time) do obey some predictable patterns. (Figure 1 illustrates a daily traffic profile at an ACD center.)

Whereas the communication facilities (such as FX and WATS lines) may be sized according to average peak-hour traffic volumes observed over all the business days of a typical month, the agent force is generally sized by the traffic volumes of each half/full-hour period. Approximately 70 percent of the ocst of a call distribution center is generally, accounted for by the agent force. As labor costs rise, so will this percentage.

The key issue behind incoming-call management is simply how to provide an appropriate grade-of-service to customers at all working hours at minimal cost.

The decision as to the method of managing incoming calls is generally determined by a combination of the following factors: management perception of customer needs; the grade-of-service required by the customer; the demand for cost savings from better management/utilization of communication resources (for example, FX, WATS lines) and call-answering agents; the availability of intelligent switching nodes; the need for better visibility of the entire operation; and the need for better control of the entire operation. It is the purpose of this article to delineate the past history, the present state-of-the-art and some future trends in incoming-call management.

The first attempt to distribute evenue-producing calls to appropriate agents was done by a telephone operator assigned to a manual cord switchboard. Such a technique is as old as the telephone.

The next vehicle for call distribution was a multi-button 1A key telephone system (KTS), first developed in 1938. The 1A KTS was a special assembly of lamps and keys, which was subsequently standardized and referred to as the wiring plan. An improved version, the 1A1 KTS, was developed in 1953 to reduce the amount of field labor required to install and maintain the 1A. The 1A2 KTS, yet another enhancement, was introduced in 1963. It employed solid-state components, printed wiring boards and miniature relays to reduce system size. From the user's viewpoint, there was little difference from the older version. Packaged Key Systems Introduced

Two new packaged KTSs, called Com Key systems, were introduced in 1973, and a third in 1975. The Com Key systems provided a number of new service features, such as tone signaling, multi-line conferencing, and a provision for incoming callers placed on hold to hear music rather than dead silence.

Many key telephone systems are still being used to manage incoming revenue-producing calls in small companies. The major disadvantages of these KTSs are (1) unequal distribution of calls to all the agents and (2) lack of information regarding the incoming-call management operation during all the periods of its life cycle. In order to appreciate the KTS environment, consider this scenario: An agent line keeps on ringing because the agent has gone to the restroom, some agents just sit while others curse their heavy loads--the resulting confusion continues with an outward appearance of meaningful activity. Auto Call Distributor Debuted

A device to automatically distribute the incoming calls to serving agents according to the first-in-first-out (FIFO) basis is illustrated in Figure 2. AT&T introduce the 2A ACD (Automatic Call Distributor) in 1963. It could handle a maximum of 56 incoming trunks and 60 agent positions using a standard crossbar switch. The 2B ACD was introduced in 1973. It could accommodate a maximum of 68 trunk and 70 agent positions using a miniature version of the crossbar switch. One could connect up to three 2B ACDs to achieve a local network for load balancing among 180 agent positions. The 3A ACD, introduced in 1963, employs a step-by-step switch to handle 198 trunks and 200 agent positions.

The small 4A Call Distributor was introduced in 1973 to serve only 20 trunk and 15 agent positions. The 4A is not an ACD, since calls are not automatically distributed but are selected by the agents on the basis of flashing lights.

Although the above-mentioned ACDs did represent an improvement in the way that calls were distributed to the agents according to the FIFO criterion, there was still a lack of management information, without which no one could evaluate the performance of the revenue-producing system. There were some features on the 2A, 2B and 3A ACDs--namely, the ability to provide delay announcements to callers and the ability of the 3A to handle multiple agent groups and traffic overflow to a distant ACD--that probably resulted in better service to customers, but there was no way to quatify the improvement.

Some temporary solutions, in the form of overlayed monitoring systems, were employed to extract a bare minimum of traffic data. However, high operating costs and minimal management data continued to plague the early ACDs.

The next set of improvements came from the interconnect companies that owe their existence to the landmark Carterfone Decision of 1968. Of course, these improvements were also spurred by certain farsighted customers, such as Mike Huntley, then the communication manager for Continental Airlines. He was absolutely sure that the next set of enhancements in the areas of call-handling and a management information system (MIS) could only come from a digital switching technology.

To be specific, it is possible to achieve all the enhancements from any stored program controlled (SPC) switching node. All one needs is a computer, and--given enough memory--one can do almost anything. Analog-Digital Debate Raged

There were many papers presented during the '70s that either exalted digital switching or defended analog switching. When one looks in retrospect, both sides had a point and both made a lot of sales. Defenders of digital technology had a slight edge however--the future was on their side. Today, it would be very hard to find anyone who dares to defend either the analog switching or analog transmission technology.

Collins Radio Group of Rockwell International was the first interconnect company that developed a digital switch in the form of a research product and then harnessed it to achieve most of the enhancements specified by Mike Huntley. The first installation of an intelligent ACD was accomplished in July 1973 at a Continental Airlines site in Houston. The rest is history.

Many other vendors followed with smart ACDs, some with analog switches, some with nonstandard digital interfaces, and some with cheaper ACDs. Even AT&T introduced an improved ACD service from its 1ESS switches in 1975, and several enhanced versions (with better MIS packages) followed later. That's known as free enterprise. They all made money. Most of them still do. Call-Handling Was Enhanced

Enhancements in call-handling can be listed as follows:

* Automatic distribution of calls to all agents instead of only uniform call distribution (according to which each new call is sent to the first free agent on the basis of a fixed-list search).

* Direct outward dialing without operator assistance.

* The ability to dial anyone in the ACD center without operator assistance.

* Supervisor capability to transfer a call to another position, hold the call, or conference or monitor the call.

* Long-term queuing is allowed to achieve an economy of scale.

* Multiple queues--one for each gate (defined as a trunk group and an agent group association)--are allowed and inter-gate overflows are permitted to achieve additional economies of scale.

* Automatic configuration of the system through the supervisory console, which implies the ability to add, move and drop agents and trunks.

* An emergency alert is provided to guard against a bomb threat or an obscene call.

* Incoming call IDs (for example, originating city) are provided to aid the agent in delivering more-courteous and meaningful service.

* A night service feature that either gives a recorded message to the caller or diverts the call to another ACD center.

* A delay announcement followed by music and/or additional delay announcements, encouraging callers to wait.

* Call-waiting lights to prod the agents to complete their calls without being discourteous.

The above enhancements in call-handling could have been achieved through a SPC analog switch, but experience has shown that a digital switch solution provides many hidden advantages, such as higher reliability, quiet operation with no ear fatigue on the agents, and future adaptability to evolving digital transmission facilities. Reports Pakcage Proved Payoff

The reports package that came with the smart ACD was probably the most-important reason for buying the new ACD, since it was the only tool to help communication or reservation managers to defend their calculations for the payoff period in front of management. Furthermore, this was also the only tool to maximize the efficiency of an ACD operation. Those were then the two highly interrelated bottom lines.

The major enhancements in the early reporting package were as follows:

* System reports that appear on the CRT screen and on printed pages every half/full hour by each gate. The report depicts number of offered calls, number of handled calls, number of agents assigned, number of agents that should have been assigned, average call holding time, average time to answer the call, average delay in the queue, and more.

* Reports for agent information groups within one gate or ACD split. The data is similar to that in system reports except for such details as after-call work time. These reports help management to evaluate the performance of newly trained agents.

* Delayed call profiles that depict the number of calls abandoned within several time windows. These profits were the first tools to describe the behavior of the callers by gates.

* Trunk performance reports that depict the number of calls actually offered, handled and abandoned; average call holding times on trunks; and, in some cases, the percent of times all trunks were found busy (a condition that's directly related to blocked calls within any given gate).

* Trouble reports that aid the ACD operator to run a center efficiently. To illustrate, an alarm signifying a failed trunk is a welcome sign both to management (implying a quick fix from the telco and hence 10 to 15 percent savings on monthly trunk-leasing costs) and to the telco whose monitoring equipment was incapable of discovering failed trunks on a real-time basis. Other alarms about hardware and software malfunctions ensure high availability through timely actions. These benefits are direct results of the evolving intelligent switching nodes (most probably, the digital variety).

* Records of call overflows and diversions, which are maintained and included in the timely reports. The subject of call overflows and diversion is rather controversial. One one hand, there's a desire to achieve an economy of scale and hence better grade-of-service. On the other hand, some local managers may wish to export their problems (resulting from bad scheduling decisions) to other managers.

* Recording of individual ACD and PBX (in case this facility is provided b y the ACD node) calls on a magnetic tape, which is sometimes very essential in planning future system enhancements. Some intelligent nodes record not only the call-holding and arrival times but also the associated trunk ID, number of rings, waiting time and after-call work time. Computer Printouts Overflowed

There was a lengthy learning period for deciding which reports were useful. During the early days, there was an attempt to get too much data to often. Only when offices began to fill with unexamined computer printouts did some managers decide to relax their demands. Nowadays, most reports can be disabled, as desired.

It may be worthwhile to relate an incident that reflects upon the complex relationship between human nature and evolving technologies. While at Collins/Rockwell, I was instrumental in implementing an early reports package for smart ACDs. Sent to several ACD sites to assess customer responses, I expected much praise of the software package. Instead, there were high marks from every customer for the software package. Instead, there were high marks from every customer for the way the agents liked the new ACD operation. Absenteeism was down, morale was high and there was pleasantness at both check-in and check-out times. The new utopia, I was told, resulted from (1) no ear fatigue and (2) the user friendliness of the system. Nevertheless, I was happy no one criticized the software package. Such are the surprises a person must be prepared for at all times while engaged in the field of high technology.

It's been said that if one does not know the past, then one is doomed to repeat it. I may add that if one doesn't know the present, then nothing can be known about the past and the future. The present can be understood in terms of the marketplace and the activities taking place in the areas of call processing, new report packages, traffic engineering and ACD network planning. The ACD Market Estimated

The total market for all ACD sizes during 1983 was estimated at $175 million. The market share for ACDs that serve centers with 150 or more agents was $50 million. Approximately 70 percent of all ACD installations have fewer than 50 agent positions and 90 percent have fewer than 100 agent positions.

It is expected that the ACD market will grow at the annual compounded rate of 12 percent. Accordingly, the ACD market during 1988 will grow to about $310 million. A word of caution is in order. The basic assumption made here is that an ACD is based on pure voice-switching Technologies. The discussion of the future at the end of this article may render these projections invalid.

Most of the current interest in state-of-the-art call processing deals with automating many of the tasks an agent performs. To achieve that goal, a computer-to-ACD interface data link has been developed; and several tasks, such as after-call work, retrieving a list of clients to be called, and out-dialing, are accomplished with the aid of the computer. Several applications that require not only servicing incoming calls but also initiating outgoing calls for producing revenue are now being implemented.

A great deal of interest in state-of-the-art report packages exists in the agent-performance reports available from many of the interconnect ACDs. Each agent is assigned an individual identification number; and the system tracks the performance of each agent, no matter to which gate or information group the agent is assigned by the supervisor. This capability represents utmost flexibility in the way an ACD site may be configured.

The capability to customize the format of any report is another recent innovation in the report package. Column headers can be specified and columns can be interchanged. Since the raw data are already known, new functions can be computed and presented in the report.

Additional improvements have been made in the trend-and-forecast reports. The original version, as introduced in 1975, was too cumbersome for the ACD site personnel. Several recent versions produce time-consistent peak-hour traffic profiles on a running basis for predicting the trunk and agent requirements. No daily, weekly and monthly merging of magnetic tapes is required, as was the case with the original version. Cheap mass storage and random acess memories made all that possible.

To perform traffic engineering, the communications manager must first consult reference works to acquire all the theoretical and practical knowledge for dimensioning trunks and agent forces. For dimensioning trunks, one needs to evaluate the average value of peak-hour erlangs considering all business days of a typical month, and then use the Erlang B formula. If one must consider the effects of blockage and retrials, one could use the Poisson formula or Extended Erlang B formulas.

For dimensioning agent forces, one must first get time-consistent weekly traffic profiles for each half/full hour and then apply the Erlang C formula. One should note that most of the published tables assume infinite sources (trunks). Approximate methods are adequate considering the way ACD centers are staffed during each half/full hour. One must refer to published works to acquire a supply of rigorous analytical tools. The effects of call abandonments and retrials are still beyond rigorous solutions.

Some recent results obtained from ACD simulations suggest that each time an agent force is cahnged for a gate, it takes 50 to 60 hours to reach a statistical equilibrium. Consequently, any hope to find agreements between theory and observed results over half/full-hour periods should be dashed. But that should not be discouraging to pragmatic minds. The advantages gained from scheduling agents for each half/full hour far outweigh the slight discrepancy between theory and practice. Integrated Networks Implemented

At the present time, there's a great deal of interest in achieving economies of scale by integrating ACD and corporate communication applications on the same networks justify traffic overflows and diversion.

Network planning involves computing the optimum number of switching centers and their locations, optimum choice of communication facilities (especially the new WATS and TI offerings), an optimum strategy/methodology for scheduling of agents for all half/full hours of every business day, and the optimum interconnections of these centers. Figure 3 illustrates the final results of such an exercise. Repeating this process for all years of the life cycle, one can compute the payoff period. Only then can one convince management of the need for a network of ACDs.

As personal computers/workstations proliferate in our society at a tremendous rate, it is inevitable to see the need for direct interfaces to large data bases via packet switching or some form of fast circuit switching. It is interesting to note that while the customer is communicating to the agent on a circuit-switched path, the agent is communicating with the computerized data base most probably via packets sent on a virtual circuit (VC). By allowing the customer to establish a VC with the data base and to communicate directly after using some passwords and account numbers, it is possible to reduce the costs of an agent labor force to almost nothing. Direct Workstation Access Eyed

One can already retrieve data about plane landings and takeoffs via push-tone telephones or workstations without the intervention of agents. It is only a question of time when all services will be requested through a workstation.

The real question will be how to establish concurrent sessions with the data-base computer. Fast circuit switching or packet switching will be contending for victory. A new ACD based on fast circuit switching or packet switching should have a brighter future than that for the existing variety.

For that particular reason, the assessments made earlier about the ACD market may be too small. The large number of new data-base-oriented services will easily eclipse the services handled by the older ACDs. The current proliferation of personal computers and workstations is about to usher in the new Information Age. The new ACD is destined to play an important role.

COPYRIGHT 1985 Nelson Publishing
COPYRIGHT 2004 Gale Group