Staffing estimate
The Staffed / needed column of the queue table compares the agents working a queue now with the agents the queue needs to meet its service level target at the current load. The estimate uses Erlang C, the standard contact center staffing formula. It is an estimate, and the console labels it as one.
Inputs
| Input | Where it comes from |
|---|---|
| Arrival rate | Contacts offered to the queue in the last 30 minutes, doubled to give contacts per hour. |
| Average handle time (AHT) | Today's AHT for the queue once it has at least 10 handled contacts; before that, the last 7 days' AHT if that has at least 10; otherwise today's. |
| Service level threshold and target | The queue's settings, for example 80 % answered within 20 seconds. See Service level settings. |
| Agents working now | Agents who can take the queue's work and are available, on a task or wrapping up. Agents who are away or offline do not count. |
"Agents who can take the queue's work" are the queue's members as TaskRouter reports them. Until that member list has been read, a queue that routes on a single Flex skill (a target expression like routing.skills HAS 'billing') counts the agents with that skill. For other queues the column shows a dash until the members are known.
Outputs
Hover the cell to see the reasoning, for example: Erlang C estimate for 80 % within 20 s: 60 contacts an hour × AHT 4:00 = 4 erlangs (30 contacts in the last 30 min). Needed: 7. With 5 working now: projected 49 %. Six agents would reach about 76 %, just short of the target; seven reach about 90 %.
| Output | Meaning |
|---|---|
| Load (erlangs) | Arrivals per hour × AHT ÷ 3600: how many agents would be busy all the time just to keep up. |
| Needed | The smallest number of agents that reaches the target at this load. Always more than the load, because some slack is needed to answer quickly. |
| Projected service level | The service level the agents working now would reach at this load. |
| Short | Needed minus working now, when positive. |
The cell turns yellow when the queue is short of agents, and red when it is short by at least 2 agents and at least a quarter of the agents needed.
How to read it
- Short and service level falling: add agents. The queue's attention items offer Add agents, which gives selected agents the queue's skill.
- Short but service level fine: the last 30 minutes may have been unusually busy, or agents are covering several queues. Check the trend before moving people.
- No estimate ("Not enough data for an estimate"): the queue has no handle time yet.
Limits of the model
Erlang C assumes:
- contacts arrive at random, at a steady rate;
- nobody hangs up while waiting;
- agents work only this queue.
Real queues break all three: arrivals come in waves, callers abandon, and agents often serve several queues. On small queues the estimate jumps around. For this reason the estimate never raises an attention item on its own; the attention items look at the service level and waits that actually happened.