How the Erlang C calculation works
Erlang C turns your forecast into the number of agents needed. It works in four steps:
- Traffic. Traffic intensity in Erlangs = contacts in the interval × average handle time ÷ interval length. 100 calls of 3 minutes in 30 minutes = 10 Erlangs: on average, 10 agents are busy all the time.
- Chance of waiting. For a given number of agents above the traffic, the Erlang C formula gives the probability a contact waits in queue.
- Service level. The share answered within your target time = 1 − P(wait) × e−(agents − traffic) × target ÷ AHT. Average speed of answer = P(wait) × AHT ÷ (agents − traffic).
- Agents needed. The calculator adds agents until the service level (and your occupancy and ASA limits) are met, then grosses up for shrinkage to give scheduled headcount.
Using it with Amazon Connect
- Take contacts per interval and average handle time from Historical metrics (Contacts handled, Average handle time) or from the CTR export, and paste them into "Staff a whole day".
- Set the same service-level threshold you report on in Connect (for example 20 or 60 seconds).
- For chat, enter how many chats an agent handles at once; the result is an approximation.
- To turn the staffing into a bill, use the Amazon Connect pricing calculator.
Limits
Erlang C assumes callers never abandon, calls arrive randomly and agents are interchangeable. With high abandonment it overstates the agents needed; with skills-based routing across several queues, treat each skill group separately. Validate against real interval data before you change a roster.