A number of factors can affect the accuracy of wait time predictions. Wait time predictions are best suited for medium-volume or high-volume call scenarios. The potential accuracy of a wait time predictor increases as the rate of removal from queue increases.
Under all conditions, EWT is the most accurate wait time predictor, but EWT is most accurate when the rate of removal from queue at a given split priority level is a minimum of one call every 30 seconds.
Predictions can be made for a split with multiple priority levels as long as the majority of calls are delivered to lower priority levels. If the majority of calls are queued at the higher-priority levels, any predictions made for the lower-priority levels not be accurate.
The following circumstances can limit the accuracy of the wait time predictions.
Factors limiting wait time prediction accuracy |
Description |
System restart or new split administration |
The EWT algorithm uses a combination of historical and real-time information to make predictions. When no historical information exists, such as when a new split is added or a reset system 3 or 4 is completed, there is the potential for inaccuracies. To prevent inaccurate predictions when there is no historical information, administer the Expected Call Handling Time field on the Hunt Group screen. The value in this field is then used in place of the missing historical data. If the value of this field does not accurately reflect the call handling times of the split, EWT predictions can be inaccurate until some call history is generated. The algorithm normally requires about 30 queued calls to be answered from a split priority level before it reaches its maximum accuracy. You can change the value in the Expected Call Handling Time field by executing a change hunt group command. Changing the value does not disrupt EWT predictions by overwriting EWT history. The value is stored and used the next time a reset system 3 or 4 is executed. |
Low call volume applications |
Split priority levels where the rate of removal from the queue is very low can only be predicted with limited accuracy. |
Sites with frequent staffing changes |
Although EWT immediately adjusts for all types of staffing changes, since predictions can already be made for calls that are waiting in queue, the past predictions were based on staffing information which is now out of date. Therefore, the EWT in scenarios where large staffing changes are continually happening can only be predicted with limited accuracy. |
Staffed agents who rarely answer calls to a split |
The EWT algorithm takes account of agents in multiple splits in its calculation. However, suppose there are many agents who are assigned to a split but spend most of their time answering calls in their other splits. If a large number of these agents are moved to or from the split, the EWT for this split be temporarily inaccurate until it adjusts to those changes. |
Applications with widely varying call handling times |
If the majority of calls to a split are handled within a narrow range of times, the accuracy of any predictor will be much greater than that for a split where call handling times are widely different. |
