Erlang c calculator
Estimate simultaneous agents, waiting time and occupancy for a queued customer service workload.
Modelled staffing requirement
Agents for your service target
Available at once
5
Schedule with shrinkage
8
5 available agents meet an 80% target within 20 seconds. Schedule 8 with 30% shrinkage.
Estimated queue performance
| Measure | Result |
|---|---|
| Offered load | 3 erlangs |
| Average handling time | 180 seconds |
| Answered within target | 81.09% |
| Probability of waiting | 23.62% |
| Average queue wait | 21.25 seconds |
| Agent occupancy | 60% |
Queue estimates use available agents handling one contact at a time. Shrinkage increases scheduled headcount without changing queue performance.
One-agent comparison
| Available agents | Wait probability | Average wait | Within target | Occupancy |
|---|---|---|---|---|
| 4 | 50.94% | 91.7 seconds | 54.41% | 75% |
| 5 | 23.62% | 21.25 seconds | 81.09% | 60% |
| 6 | 9.91% | 5.95 seconds | 92.9% | 50% |
Turn contact work into offered load
Offered load, measured in erlangs, is total handling work divided by interval length. Use the same time unit for both.
Include the time an agent spends talking and completing follow-up or after-contact work. Exclude the customer’s time waiting in the queue. Average handling time is total handling minutes divided by contacts.
In the hypothetical starting example, 60 contacts take 180 handling minutes over a 60-minute interval. Average handling time is 3 minutes, or 180 seconds. Offered load is 180 / 60 = 3 erlangs.
Choose an interval that describes the period you need to staff. A whole-day average can hide a busy hour. Use contacts and handling work from the same period.
Compare staffing scenarios
The calculator finds the smallest whole number of available agents that meets your response target. It then compares neighbouring staffing levels for the same workload.
With the starting example, 4 agents answer an estimated 54.41% within 20 seconds at 75% occupancy. Five agents reach 81.09% at 60% occupancy. Six agents reach 92.90% at 50% occupancy.
Five available agents meet the 80% target, with an average queue wait of 21.25 seconds. That average includes every contact. The service level measures the share answered within 20 seconds.
For 30% shrinkage, scheduled headcount is 5 / (1 - 0.30), rounded up to 8. Plan coverage across the interval so 5 agents remain available.
Know the queue model’s limits
Erlang C assumes independent random arrivals at a steady rate and exponentially distributed handling times. Agents have the same service rate and handle one contact at a time.
The queue is first come, first served with unlimited waiting space and no abandonment. Canadian University Dubai’s Erlang C queue explanation describes these assumptions and the need for capacity to exceed demand.
The traffic engineering reference by Richard Parkinson explains how the queue formula estimates waiting probability and response times.
Calls can fit this model when one agent handles one call. Asynchronous chat includes customer pauses and agents working across conversations, so chat duration alone is not a suitable handling-time input.
Average queue wait includes contacts answered immediately. Occupancy is the share of available agent time spent handling contacts. When available service capacity is at or below demand, a queue cannot settle and its waiting estimate is marked unstable.
The model accepts up to 100,000 contacts per interval and 10,000 erlangs of offered load. A 100% service target has no finite staffing answer for positive demand.
Use the estimates to compare staffing under the stated assumptions. Check your actual arrival peaks, abandonment and skill routing before turning the result into a roster.