All free tools

Erlang C Calculator

Free Erlang C calculator: agents needed to hit your service level, with ASA, occupancy, shrinkage and AI containment. Formula and worked example.

Inputs

Interval length

Everything runs in your browser. The inputs are saved in the page link, so you can share the result.

Results

Updates live

Agents on phones

14

Erlang C minimum for your target

Agents to schedule

20

After 30% shrinkage

Answered in 20 s
88.8%
Average speed of answer
7.8 s
Occupancy
71.4%
Probability of waiting
17.4%
Answered immediately
82.6%
Traffic
10.00 Erlangs
Staffing table around the answer
AgentsService levelASAOccupancy
1264.0%40.4 s83.3%
1379.6%17.1 s76.9%
1488.8%7.8 s71.4%
1594.1%3.7 s66.7%
1697.1%1.7 s62.5%

Check occupancy at other headcounts or from your actual handle and idle time with the call center occupancy calculator.

How this is calculated

Converts your calls per interval and handle time into traffic in Erlangs, then adds agents one at a time until the Erlang C service level meets your target and occupancy stays under your cap. Shrinkage then grosses the result up to the number you need to schedule. An optional AI-containment share removes contacts that never reach the queue. Everything runs in your browser.

How to use the Erlang C calculator

Enter calls per interval, average handle time and your service-level target (for example 80% in 20 seconds). The calculator uses the Erlang C formula to find the fewest agents that meet it, then adds shrinkage to give the number to schedule. For example, 100 calls per 30 minutes at 3 minutes each needs 14 agents on phones and 20 scheduled.

Staffing email or ticket queues by monthly volume? Use the support staffing calculator. Ticket queues do not hold a caller in line, so a monthly workload model fits them better.

What Erlang C does not model

  • Abandonment. Erlang C assumes every caller waits and nobody hangs up, so with real abandonment it tends to overstate the agents you need. For queues where callers hang up often, use Erlang A or a workforce management tool.
  • Chat concurrency. It assumes each agent handles one contact at a time. If agents run two or three chats at once, the result overstates headcount. Email and back-office work do not queue in real time.
  • Averages. Run it for your busiest interval. A daily average hides the peak that sets your staffing.

Assumptions (Poisson arrivals, exponential handle times, unlimited queue, no hang-ups): Wikipedia, "Erlang (unit)". Limits for abandons, email and chat: Call Centre Helper, "An Introduction to Erlang Calculations".

The Erlang C formula

Notation: λ = contacts offered in the interval, I = interval length in seconds, h = average handle time in seconds, T = target answer time in seconds, N = agents on phones.

  1. Traffic intensity in Erlangs: A = λ × h / I
  2. Probability a contact waits: PW = [Aᴺ/N! × N/(N − A)] / [Σi=0…N−1 Aⁱ/i! + Aᴺ/N! × N/(N − A)], valid only when N > A. With N ≤ A the queue grows without limit.
  3. Service level: SL = 1 − PW × e−(N − A) × T / h
  4. Average speed of answer: ASA = PW × h / (N − A)
  5. Answered immediately: 1 − PW; occupancy: A / N
  6. Start at the first whole number above A and add one agent at a time until SL meets the target and occupancy is at or under the cap.
  7. Agents to schedule: N / (1 − shrinkage), rounded up

Factorials overflow quickly (171! is too large for a standard floating-point number), so the calculator uses the Erlang B recursion instead: B = 1, then for j = 1 … N, B = A × B / (j + A × B), and PW = N × B / (N − A × (1 − B)). It stays accurate for queues with hundreds of agents.

Sources: Call Centre Helper, "Erlang C Formula: Made Simple With an Easy Worked Example" (Jonty Pearce); Wikipedia, "Erlang (unit)" for the Erlang B recursion; Wikipedia, "M/M/c queue" for the same mean waiting time in queueing notation.

Worked example: 100 calls in 30 minutes

100 calls per 30 minutes, 180-second handle time, target 80% in 20 seconds, 85% occupancy cap, 30% shrinkage. Traffic A = 100 × 180 / 1,800 = 10 Erlangs.

Agents (N)P(wait)SL ≤ 20 sASAOccupancy
110.682138.96%122.8 s90.9%
120.449464.02%40.4 s83.3%
130.285379.56%17.1 s76.9%
140.174188.84%7.8 s71.4%

Thirteen agents miss the target by 0.44 points, so the answer is 14 agents on phones: service level 88.8%, ASA 7.8 seconds, 82.6% answered immediately and occupancy 71.4%, under the 85% cap. At 30% shrinkage you schedule 14 / 0.70 = 20 agents. These are the calculator's default inputs, and the P(wait), service-level and 14-agent results match Call Centre Helper's worked example.

The occupancy cap matters most in large queues. At 5,000 calls per 30 minutes with the same handle time (A = 500), 80/20 alone needs 510 agents at 98% occupancy. An 85% cap raises that to 500 / 0.85 = 588.2, rounded up to 589.

What 30% AI containment does to staffing

Keep the same inputs and let an AI agent fully handle 30% of contacts before they reach the queue. Offered calls fall to 70 and traffic to 7 Erlangs. Nine agents reach only 69.2%, so the answer is 10 agents on phones (SL 84.1%, ASA 13.3 seconds, occupancy 70.0%) and 10 / 0.70 = 14.3, rounded up to 15 scheduled.

30% containment cuts scheduled agents from 20 to 15, a 25% saving rather than 30%. A smaller queue needs more spare capacity per Erlang to answer in the same time, so staffing does not fall in a straight line with volume. Count only contacts that never reach a person: a deflected contact is not a resolved one, and escalated contacts still arrive, often with a longer handle time after AI.

Service level, shrinkage and occupancy conventions

These are conventions, not standards. Set yours by what callers tolerate and what an agent-hour costs. Figures as of November 2026.

InputWhat the source saysSource
Service level 80% in 20 secondsThe most popular target in call centres (Call Centre Helper). ICMI notes that 80/20 once appeared in ACD manuals as an "industry standard" but never was one.ICMI, "Set Service Level with Call Center Cost and the Customer in Mind" (updated Mar 24, 2020)Call Centre Helper, "Is 80/20 Still a Reasonable Service Level?" (modified Aug 13, 2025)
ShrinkageCall Centre Helper puts the industry average at around 30–35%.Call Centre Helper, "Erlang C Formula: Made Simple" (modified Jul 29, 2026)
Maximum occupancyCall Centre Helper warns that agents above 85% occupancy are likely to burn out; Call Centre Tools cautions against running above 85–90% for long periods.Call Centre Helper, "Erlang C Formula: Made Simple"Call Centre Tools, Erlang calculator

Related RTCI tools and guides

Frequently asked

What is Erlang C used for?

Sizing a live call center queue: how many agents you need on phones in an interval to answer a target share of calls within a target time, given calls offered and average handle time. It assumes callers wait in line rather than getting a busy signal.

What is the difference between Erlang B, Erlang C and Erlang A?

Erlang B assumes blocked calls are lost, with no queue, and is used for trunk or line sizing. Erlang C assumes every caller waits and nobody hangs up. Erlang A adds caller patience, so it models abandons and usually needs slightly fewer agents.

Where does shrinkage go in the Erlang C calculation?

After it. Erlang C gives the agents who must be on phones. Divide by (1 − shrinkage) to get the number to schedule: 14 on phones at 30% shrinkage means 20 scheduled.

Can I use Erlang C for live chat?

Only roughly. Erlang C assumes each agent handles one contact at a time, so if agents run two or three chats at once it overstates headcount. Email does not queue in real time; use the monthly customer support staffing calculator for it.

How does an AI agent change Erlang C staffing?

Only contacts the AI fully handles leave the queue, and the saving is smaller than the containment rate. With 100 calls per 30 minutes at 180 seconds and an 80/20 target, 30% containment cuts scheduled agents from 20 to 15 (25%). Escalated contacts still arrive and often take longer.