Metrics

In the "service_in_time_counters" table, how is the value for "num_calls_answered" determined?

If the call is distributed to an agent, and that agent answers either manually or via auto answer, that call would count toward the num_calls_answered metric. For more information on this metric and others in the service_in_time_counters table, see the Reporting Database Specification.

How is EWT Calculated?

The real-time metric Estimated Wait Time (EWT) is calculated using a composite approach: it combines calculated EWT (the method of calculation is described below) with the actual Average Speed of Answer (ASA) on the service. This approach allows EWT to adjust better to complex configurations (e.g., teams serving multiple services, blended services, or scenarios that require specific skill sets in the Find Agent block).


	EWT is reset once every 24 hours at the configured reset time, along with other resettable real-time metrics. The first EWT value after a reset is 200 seconds.

To calculate EWT, we use a probabilistic approach: the conditional probability of an event occurring within a given time interval is calculated with the assumption that agents become Ready over time according to an exponential distribution.

Imagine there is a service queue with incoming interactions and a pool of agents assigned to this service. If an interaction comes into the service queue and there is an agent in the Ready state, then EWT is 0. If all agents are in a Busy state at this moment, we can calculate the probability that an agent becomes available within a certain amount of time.

In this case, the probability distribution can be approximated as an exponential distribution where the parameter of distribution is proportional to the total number of agents and inversely proportional to the sum of Average Handle Time and Average After Call Work (ACW) Time.

We can then do the following:

Select a confidence interval to find the probability distribution for when agents become available.
Estimate the probable handling time as the current handling time uniformly distributed among available agents.
Use the average abandon time to estimate how many interactions ahead of the current interaction in the queue will be abandoned before reaching an agent.

Solving with these values gives the estimated wait time.

What is the definition of "total_busy_time_in"? What about "total_busy_time_out" and "total_ready_time"?

These metrics and others are defined in the Reporting Database Specification. Those specific metrics are defined in section agent_performance:

total_busy_time_in - Specifies the sum of times the agent was busy with inbound calls; includes hold times, but does not include ringing time or after call work time
total_busy_time_out - Specifies the sum of times the agent was busy with outbound calls; includes hold times, but does not include dialing time or after call work time
total_ready_time - Total time the agent spent in the Ready state during the aggregation interval< Previous | Next >