When QueueMetrics has to compute the Real-Time report, it tries to build the current situation based on what happened recently - normally since the last midnight. This usually works fine in the common case where all agents stop working at night and start working by logging on in the morning. On the other side, this produces incorrect results if your call-center works over the 24 hours, because at midnight by default all agents appear to be logged off.

This choice was made in order to minimize database searching for QueueMetrics, and it works just fine in the majority of call centers. There are a few cases where this choice may have to be changed:

• If your call center stops daily, but not at midnight, you should tell QM the time when you expect all agents to be logged off. If e.g. at 4 AM you are sure all agents are logged off, you should set:
  realtime.startHour=4
  

• If your call-center never stops, but agents work in shift of 6 hours, you should use a "sliding window" instead; this determines the maximum lookback based on a number of hours before now (make it 1.5x to 2x the length of the shift).
  realtime.startHour=s8 Of course, the larger the sliding window, the more data QM has to evaluate every time it produces a realtime report, so it will use more RAM and CPU.
• If your agents are always logged on, you should ignore the statuses as reported by QM, as QM has no way to reliably know if they are available or not. QM will generally log them on as they take calls within the given time window. See below for a detailed explanation of why it is not a good idea to have permanent agents.

See also:

• Why agents that do not log-on are not a good idea