Skip to content
Tutorials Center
My learning

Topics

Lesson 20 of 38 Ā· Multi-Server Queues

Staffing a Call Center with Erlang C

Choose the number of agents with Erlang C to meet a service level such as 80/20, and learn the square-root staffing rule for large systems.

The most famous use of the M/M/c model is staffing: deciding how many agents to schedule so that customers wait an acceptable time. This lesson turns the formulas of the chapter into a method, and ends with an elegant rule of thumb for large systems.

Service level

Call centers state targets as a service level: "X% of calls answered within T seconds". The classic target is "80/20": 80% within 20 seconds. In an M/M/c queue,

P(Wq ≤ T) = 1 - C(c, a) e^(-(cμ - Ī») T)

(by the tail formula in the last lesson), where each call that finds a free agent counts as answered in 0 seconds. We want the smallest c for which this probability is at least the target.

Because c must exceed a for stability, we start at c = ⌊aāŒ‹ + 1 and increase c until the target is met.

Worked example

A center receives 120 calls per hour; each call takes 5 minutes on average to handle (including after-call work). Then λ = 120 per hour, μ = 12 per hour, a = 10 erlangs. At least 11 agents are needed for stability, but 11 gives very long waits. We search for the smallest c achieving 80/20.

Python Find the staff for 80/20

Press Run (or Ctrl + Enter) to see the result here.

Program input Ā· what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

You can see the trade-off: more agents raise the service level but lower occupancy (the fraction of time agents are busy). Management chooses the cheapest staffing level that meets the promise.

The square-root staffing rule

For large centers, a neat pattern emerges. To keep the probability of waiting C(c, a) at a fixed level, you need

c ā‰ˆ a + β √a

where β depends only on the target (a higher β means a smaller waiting probability). Offered load a gives the base staffing, and the spare capacity grows only like √a. So big centers run at higher occupancy than small ones at the same service quality. This is the same effect as the economy-of-scale gain from pooling.

Let us find, for several offered loads, the smallest c with C(c, a) ≤ 0.2 and compute β = (c - a)/√a.

Python Square-root staffing

Press Run (or Ctrl + Enter) to see the result here.

Program input Ā· what you'd type while the program runs

Type the answers before you press Run, one per line. Each time the program reads a line (with input()), it gets the next line from this box.

The coefficient β stays in a narrow band for large a, while occupancy rises toward 100%. That is the practical content of the rule: a center ten times larger needs only about three times the spare agents.

Erlang C assumes callers never hang up, agents are identical and the rate is constant. Real centers have abandonment, skills and daily patterns, so use these results for a first estimate, then refine with simulation or the Erlang A model (which adds abandonment).

Staffing step by step

  1. Forecast arrivals per interval (for example 30 minutes) and the mean handling time, giving a = λ/μ.
  2. Choose the service-level target (for example 80/20).
  3. Find the smallest c meeting it using P(Wq ≤ T).
  4. Convert agents on the phone into scheduled staff by allowing for breaks, absences and training (a "shrinkage" factor).
  5. Repeat for each interval of the day.

Check your understanding

Quick check Ā· 4 questions

of 4 answered You scored / 4 Ā· perfect!

  1. 1. What does an 80/20 service level mean?

    The first number is the percentage answered, the second the time limit in seconds.
  2. 2. A center gets 90 calls per hour of 6 minutes each. What is the offered load?

    a = lambda * mean handling time = 90 * (6/60) = 9 erlangs.
  3. 3. According to square-root staffing, how does spare capacity c - a scale with a?

    c is about a + beta * sqrt(a), so the margin grows like sqrt(a).
  4. 4. Why does occupancy rise as a center grows, at the same service quality?

    Pooling economies of scale mean the margin sqrt(a) is a shrinking fraction of a.

Summary

  • Service level P(Wq ≤ T) = 1 - C e^(-(cμ-Ī»)T) drives staffing: find the smallest c that meets the target.
  • Offered load a = Ī»/μ is the base; extra agents are needed for randomness.
  • Square-root staffing: c ā‰ˆ a + β√a.
  • Erlang C ignores abandonment, skills and time-varying rates, so treat results as estimates.

This is the end of the first part of the track. Continue with the next chapters on finite capacity and priorities on the Queueing Theory track page.