The Average service time engineering tableson page 112 list the number of ACD agents required to handle a given incoming call load. The top rows on each of these tables show the possible delay times for a given incoming call load (calls per hour or busy hour calls), and the Number of agents columns list the agents required to handle the incoming call load so that 90 percent the incoming calls will be answered by the agents before the specified delay has occurred.
Note:
Note: The entries in these tables are in busy-hour calls, which are the number of
calls received by the ACD during peak levels of caller activity.
To determine how many agents will be required to handle the incoming call load of an ACD split:
1. Use the BCMS split report (list bcms split) to determine the average talk time
(the time an agent spends processing a call, or talking to a caller). The Average service time engineering tableson page 112 contain data that describe 7, 15, 30, 45, 60, 90, 120, 180, 240, 300, and 600 second service times. Choose the appropriate table for the average talk time of the ACD split.
Note:
Note: Within this document the term average talk time is equivalent to the term average service time.
2. At the top of the table, choose the closest possible average speed of answer in seconds. Average speed of answer is actually a delay time that is defined as the elapsed time from when a call is routed to the ACD split until it is answered by an agent. The delay criterion states that 90 percent of the incoming calls will be answered by the agents before the specified delay has occurred.
3. If the calling volume, otherwise referred to as the busy-hour calls, is known, use the number indicated on the report. Otherwise, you must estimate this number. The value for busy-hour calls denotes the number of calls received by the ACD during peak levels of caller activity. A typical busy-hour calling rate might be 120, 130, or 160 calls per hour.
x1 Is the tabular value of the independent variable that immediately succeeds x
y0 Is the tabular value of the dependent variable that immediately precedes y
y1 Is the tabular value of the dependent variable that immediately succeeds y
Using reports for ACD planning
Note:
Note: The actual busy-hour calling rate depends on the number of agents staffed
and the particular application. Obviously, the numbers that are identified here as being typical would be much too high for five agent positions and too low for 30 agent positions. The numbers given are only for illustration.
4. After choosing the appropriate table and delay column, find the entry in the table for busy-hour calls that is greater than or equal to the number of busy-hour calls chosen. 5. The number of agent positions required is found in the Number of agents column of
the table.
6. You can interpolate between the tables (for different call service times), between the columns (for different delay times), and between the rows (for different number of calls per hour).
The Average service time engineering tableson page 112 were prepared by using a range of 1 to 1000 agents. For small service times, this yields high traffic rates, even for a small number of agents. The high traffic rates are presented in the tables for completeness only.
Agent engineering examples
This section provides agent engineering examples.
Example 1
The classified ads department of a newspaper receives 160 calls per hour. The average time an agent spends on each call is three minutes. If most of the calls should be answered in less than 30 seconds, how many agents should be employed in this department?
The engineering table 180 seconds average service timeon page 127 provides data for 180-second (3-minute) call durations. Under the 30-second column heading (Average speed of answer), find the first entry greater than 160 calls per hour (175). Follow this row left to the agents column and find 12 agents. The number of agents required to answer 160 calls (of 3-minute duration) per hour with 90 percent of the callers waiting less than 30 seconds is 12 agents.
For this example, consider the efficiency of the agents and the sensitivity of the parameters to changes in the call arrival rate. The efficiency of the agents is the ratio of the number of agent hours spent on the telephone to the number of agent hours in an hour. The number of agent hours spent on the telephone is 160 calls per hour times 0.05 hours (3 minutes), which equals 8 agent hours. Therefore, the efficiency is 8/12 (12 agents for 1 hour), which equals 0.67 or 67 percent.
Suppose the calls per hour increased to 185 calls per hour. The efficiency is now (185 x 0.05)/12 = 0.77 or 77 percent. The efficiency has increased, but this added
efficiency is not free of charge. The delay criterion has changed significantly from about 1.6 percent of all calls taking longer than 30 seconds for an agent to answer to about 15.0
Engineering ACD applications with report data
percent (175 calls per hour yield 10.0 percent, but 160 calls per hour were stated). To get the delay criterion back to 1.6 percent would require a delay time of about 55 seconds. Another measure of what is happening with the queue is the average time spent waiting for service in the queue. With 160 calls per hour, the mean time spent in the queue is 7.53 seconds. With 185 calls per hour, the mean time in the queue is 16.14 seconds. The point of this example is to emphasize the sensitivity of the time in the queue to the arrival rate. In other words, increasing the agent efficiency from 67 percent to 77 percent nearly doubles the various measures of queuing time.