Erlang C Calculator

See how one more agent changes modeled waiting time. Compare your available agents with the staffing needed for an entered service target and optional occupancy limit.

Erlang C · One pooled queue

See what one more agent changes.

Compare your staffing with the count needed for an entered service target. Explore waiting times and occupancy with a stationary queue model, using your own arrival and handling assumptions.

Agents are simultaneously available to take work. This calculation does not add breaks, shrinkage, shifts, routing or abandonment.

Target staffing: 14 concurrent agents.

Work arriving

Average requests in the entered interval. Fractional forecasts are allowed; zero means no new demand.

1–1,440 minutes. The interval defines a steady arrival rate, not a simulation horizon.

Include all time an agent stays occupied by one request. Range: 0.001–86,400 seconds.

Offered work

10 Erlangs

Arrival rate × mean handling time. A positive-demand queue needs more agents than this load to be stable.

Your staffing

Whole concurrent agents, 0–10,000. The result does not represent scheduled headcount.

Service target

Percent of arrivals starting service within the waiting threshold. Enter 0–100%.

Waiting before service starts, 0–86,400 seconds. Zero counts arrivals served immediately.

When enabled, staffing must also keep offered work per agent at or below the entered ceiling.

Retained for later. This draft is editable and ignored while Limit occupancy is off.

Your available staffing

12 agents

Answered within threshold
≈ 60.6707 %
Occupancy
≈ 83.3333 %
Average wait, all arrivals
≈ 67.4082 s
Arrivals that wait
≈ 44.9388 %

Target staffing

14 agents

Smallest modeled count meeting the active targets.

2 additional concurrent agents.

Answered within threshold
≈ 86.6627 %
Occupancy
≈ 71.4286 %
Average wait, all arrivals
≈ 13.0599 s
Arrivals that wait
≈ 17.4132 %

This is the smallest stable integer staffing meeting the entered targets under the declared numerical resolution.

Explore the staffing tradeoff

Select a whole-agent count to inspect its performance and recenter the chart. This does not change your entered staffing.

Service level as staffing changes

Whole-agent counts are calculated individually. The line helps you read the curve; an overloaded queue has no stationary service level.

■ Your staffing○ Target staffing◆ Inspecting– – Entered service target
Percentage of arrivals starting service within the entered waiting threshold, plotted against concurrent agents. The staffing comparison table provides the calculated values and their statuses.025507510001530Concurrent agents

Vertical scale: arrivals answered within the threshold (%). Use the inspection control and table for values that coincide on the chart.

Inspecting 12 agents

Answered within threshold
≈ 60.6707 %
Occupancy
≈ 83.3333 %
Average wait, all arrivals
≈ 67.4082 s
Arrivals that wait
≈ 44.9388 %

Active targets: Does not meet targets.

Waiting-time detail
How likely is a longer wait?

For 12 concurrent agents. Each value is the fraction of all arrivals waiting longer than the stated time before service starts.

Waiting-time survival curve. Report values retain tiny positive probabilities even when their plotted height rounds to zero.02550751000375750Waiting time (seconds)

Vertical scale: arrivals waiting longer (%). The larger dot identifies the entered threshold. The drawing may not resolve a tiny positive probability; the table and report retain it.

P(wait > time) starts at the chance of a positive wait; immediately served arrivals are excluded. The display window is not a maximum possible wait. Positive probabilities below plot resolution retain their logarithmic values in the table/report.

Mean wait among arrivals that wait
≈ 150 s
Expected requests waiting
≈ 2.24694112149 requests
Longer than the entered threshold
≈ 39.3292583798 %
Waiting-time probability table
Waiting longer than (s)Arrivals (%)Entered threshold
0≈ 44.9388—
≈ 7.5≈ 42.7471—
≈ 15≈ 40.6623—
≈ 20≈ 39.3293Yes
≈ 22.5≈ 38.6792—
≈ 30≈ 36.7928—
≈ 37.5≈ 34.9984—
≈ 45≈ 33.2915—
≈ 52.5≈ 31.6679—
≈ 60≈ 30.1234—
≈ 67.5≈ 28.6543—
≈ 75≈ 27.2568—
≈ 82.5≈ 25.9274—
≈ 90≈ 24.6629—
≈ 97.5≈ 23.4601—
≈ 105≈ 22.316—
≈ 112.5≈ 21.2276—
≈ 120≈ 20.1923—
≈ 127.5≈ 19.2075—
≈ 135≈ 18.2708—
≈ 142.5≈ 17.3797—
≈ 150≈ 16.5321—
≈ 157.5≈ 15.7258—
≈ 165≈ 14.9588—
≈ 172.5≈ 14.2293—
≈ 180≈ 13.5353—
≈ 187.5≈ 12.8752—
≈ 195≈ 12.2473—
≈ 202.5≈ 11.65—
≈ 210≈ 11.0818—
≈ 217.5≈ 10.5413—
≈ 225≈ 10.0272—
≈ 232.5≈ 9.53817—
≈ 240≈ 9.07299—
≈ 247.5≈ 8.6305—
≈ 255≈ 8.20958—
≈ 262.5≈ 7.8092—
≈ 270≈ 7.42834—
≈ 277.5≈ 7.06605—
≈ 285≈ 6.72144—
≈ 292.5≈ 6.39363—
≈ 300≈ 6.08181—
≈ 307.5≈ 5.78519—
≈ 315≈ 5.50305—
≈ 322.5≈ 5.23466—
≈ 330≈ 4.97936—
≈ 337.5≈ 4.73652—
≈ 345≈ 4.50551—
≈ 352.5≈ 4.28578—
≈ 360≈ 4.07676—
≈ 367.5≈ 3.87793—
≈ 375≈ 3.6888—
≈ 382.5≈ 3.5089—
≈ 390≈ 3.33777—
≈ 397.5≈ 3.17498—
≈ 405≈ 3.02014—
≈ 412.5≈ 2.87284—
≈ 420≈ 2.73273—
≈ 427.5≈ 2.59946—
≈ 435≈ 2.47268—
≈ 442.5≈ 2.35208—
≈ 450≈ 2.23737—
≈ 457.5≈ 2.12825—
≈ 465≈ 2.02446—
≈ 472.5≈ 1.92572—
≈ 480≈ 1.83181—
≈ 487.5≈ 1.74247—
≈ 495≈ 1.65749—
≈ 502.5≈ 1.57665—
≈ 510≈ 1.49976—
≈ 517.5≈ 1.42661—
≈ 525≈ 1.35703—
≈ 532.5≈ 1.29085—
≈ 540≈ 1.2279—
≈ 547.5≈ 1.16801—
≈ 555≈ 1.11105—
≈ 562.5≈ 1.05686—
≈ 570≈ 1.00532—
≈ 577.5≈ 0.956286—
≈ 585≈ 0.909648—
≈ 592.5≈ 0.865284—
≈ 600≈ 0.823083—
≈ 607.5≈ 0.782941—
≈ 615≈ 0.744757—
≈ 622.5≈ 0.708434—
≈ 630≈ 0.673884—
≈ 637.5≈ 0.641018—
≈ 645≈ 0.609755—
≈ 652.5≈ 0.580017—
≈ 660≈ 0.551729—
≈ 667.5≈ 0.524821—
≈ 675≈ 0.499225—
≈ 682.5≈ 0.474878—
≈ 690≈ 0.451718—
≈ 697.5≈ 0.429687—
≈ 705≈ 0.408731—
≈ 712.5≈ 0.388797—
≈ 720≈ 0.369835—
≈ 727.5≈ 0.351798—
≈ 735≈ 0.334641—
≈ 742.5≈ 0.31832—
≈ 750≈ 0.302795—
Staffing comparison table

These are the consecutive counts in the visible chart window. The target search checks all necessary earlier eligible counts separately. Below-target and near-target statuses remain distinct.

Staffing comparison table
AgentsQueue stateWithin threshold (%)Occupancy (%)Waits (%)Average wait (s)Expected queueTargets
0No stationary queueNot availableNot availableNot availableNot availableNot availableDoes not meet targets
1No stationary queueNot available1000Not availableNot availableNot availableDoes not meet targets
2No stationary queueNot available500Not availableNot availableNot availableDoes not meet targets
3No stationary queueNot available≈ 333.333Not availableNot availableNot availableDoes not meet targets
4No stationary queueNot available250Not availableNot availableNot availableDoes not meet targets
5No stationary queueNot available200Not availableNot availableNot availableDoes not meet targets
6No stationary queueNot available≈ 166.667Not availableNot availableNot availableDoes not meet targets
7No stationary queueNot available≈ 142.857Not availableNot availableNot availableDoes not meet targets
8No stationary queueNot available125Not availableNot availableNot availableDoes not meet targets
9No stationary queueNot available≈ 111.111Not availableNot availableNot availableDoes not meet targets
10No stationary queueNot available100Not availableNot availableNot availableDoes not meet targets
11Stable≈ 36.1874≈ 90.9091≈ 68.2118≈ 204.635≈ 6.82118Does not meet targets
12Stable≈ 60.6707≈ 83.3333≈ 44.9388≈ 67.4082≈ 2.24694Does not meet targets
13Stable≈ 76.644≈ 76.9231≈ 28.527≈ 28.527≈ 0.950902Does not meet targets
14Stable≈ 86.6627≈ 71.4286≈ 17.4132≈ 13.0599≈ 0.43533Meets targets
15Stable≈ 92.6883≈ 66.6667≈ 10.2042≈ 6.12254≈ 0.204085Meets targets
16Stable≈ 96.156462.5≈ 5.73403≈ 2.86702≈ 0.0955672Meets targets
17Stable≈ 98.0638≈ 58.8235≈ 3.08761≈ 1.32326≈ 0.0441087Meets targets
18Stable≈ 99.0656≈ 55.5556≈ 1.59283≈ 0.59731≈ 0.0199103Meets targets
19Stable≈ 99.5679≈ 52.6316≈ 0.787356≈ 0.262452≈ 0.0087484Meets targets
20Stable≈ 99.808450≈ 0.373113≈ 0.111934≈ 0.00373113Meets targets
21Stable≈ 99.9185≈ 47.619≈ 0.169625≈ 0.0462615≈ 0.00154205Meets targets
22Stable≈ 99.9667≈ 45.4545≈ 0.0740478≈ 0.018512≈ 0.000617065Meets targets
23Stable≈ 99.9869≈ 43.4783≈ 0.0310698≈ 0.00716996≈ 0.000238999Meets targets
24Stable≈ 99.9951≈ 41.6667≈ 0.0125438≈ 0.00268797≈ 8.95988e-5Meets targets
25Stable≈ 99.998240≈ 0.00487818≈ 0.000975635≈ 3.25212e-5Meets targets
26Stable≈ 99.9994≈ 38.4615≈ 0.00182932≈ 0.000342997≈ 1.14332e-5Meets targets
27Stable≈ 99.9998≈ 37.037≈ 0.000662197≈ 0.000116858≈ 3.89528e-6Meets targets
28Stable≈ 99.9999≈ 35.7143≈ 0.000231633≈ 3.86054e-5≈ 1.28685e-6Meets targets
29Stable≈ 100≈ 34.4828≈ 7.8372e-5≈ 1.23745e-5≈ 4.12484e-7Meets targets
30Stable≈ 100≈ 33.3333≈ 2.56736e-5≈ 3.85104e-6≈ 1.28368e-7Meets targets

Target search boundary

The neighboring rows explain the selected minimum, numerical-resolution range or supported search limit. A search limit is not a solved staffing target.

Target search boundary table
AgentsQueue stateWithin threshold (%)Occupancy (%)Waits (%)Average wait (s)Expected queueTargets
13Stable≈ 76.644≈ 76.9231≈ 28.527≈ 28.527≈ 0.950902Does not meet targets
14Stable≈ 86.6627≈ 71.4286≈ 17.4132≈ 13.0599≈ 0.43533Meets targets

Includes current inputs, target search, visible staffing window, selected waiting-time curve and assumptions.

Complete report CSV

The selectable text is identical to the download for this view. Text cells use a declared apostrophe prefix; numerical cells retain their report precision.

Assumptions, method and limits

Erlang C assumes one stationary, first-come-first-served queue with independent Poisson arrivals, exponentially distributed handling times, identical agents, unlimited waiting room and no abandonment. Real bursts, changing demand, skill routing and breaks can change actual results.

Offered load is arrivals × handling seconds ÷ (60 × interval minutes). For positive demand, agents must strictly exceed offered load. A target of 100% within any finite wait cannot be guaranteed by finite staffing under this model.

The search stops at 10,000 agents and distinguishes a clear pass, a value within declared numerical resolution, and an unresolved search limit. The resolution policy is not a certified mathematical error bound. Very small positive probabilities remain available in scientific and logarithmic form.

Numeric inputs allow up to 12 significant digits and six effective decimal places, within each field’s stated range. Available agents requires a whole decimal count. Blank drafts are not zero; an inactive occupancy draft is retained without being used.

Method sources: Tufts M/M/c queueing notes and Mandelbaum’s service-engineering notes.

How to use this calculator

  1. Enter Expected arrivals, Interval (minutes) and Average handling time (seconds). The interval converts the expected workload into a rate. Include all time an agent remains occupied by a modeled request in handling time; this does not simulate the interval's start-up or end backlog.
  2. Enter Available agents as the number simultaneously available to the pooled queue. Read current occupancy, delay probability and waiting time. If agents do not exceed offered load, stationary waiting results are unavailable; the target search can still identify a supported stable staffing level.
  3. Enter Target service level (%) and Answer within (seconds). The target measures waiting before handling begins, not completed handling. A zero-second threshold includes immediate service. The smallest supported staffing is shown only when earlier candidates clearly fail under the declared numerical policy.
  4. Optionally enable Limit occupancy and enter Maximum occupancy (%). Both the service and occupancy constraints must pass. Disabling the limit preserves its draft while excluding it from the calculation. Correct associated active errors before downloading a complete report.
  5. Use Inspect staffing to compare integer agent counts, the service-level curve and the waiting-time tail. The staffing window contains consecutive counts; off-window current or target results remain separate comparison rows. The tail starts at the probability of delay, and its drawing window is not a maximum wait.
  6. Open the Staffing comparison table and target search boundary rows to inspect the same calculated values. Download complete report or select Complete report CSV for the inputs, current and target results, displayed staffing window and tail samples. Recognized unstable, no-demand and bounded-search outcomes remain explicitly labeled in reports.

The Formula

The mathematical logic powering this calculation is:

For V expected arrivals in I minutes and average handling time H seconds, arrival rate λ = V/(60I), service rate μ = 1/H, and offered load a = λH Erlangs. With n simultaneously available agents, positive demand has a stationary M/M/n queue only when n > a. Occupancy is a/n; a 100% occupancy limit does not admit an unstable n = a queue.
Let B = aⁿ/[n!(1 − a/n)]. Erlang C delay probability is C = B/[Σ(k=0 to n−1) aᵏ/k! + B]. For a stable queue, P(wait > t) = C exp[−(n − a)t/H], service level P(wait ≤ t) = 1 − P(wait > t), mean wait = CH/(n − a), mean wait if delayed = H/(n − a), and mean queue length = Ca/(n − a). At t = 0, immediate service counts: service level is 1 − C.
The implementation evaluates the equivalent finite-n formula through a bounded logarithmic recurrence. Authored decimals give exact rational offered load, stability and occupancy comparisons. Other probabilities are numerical estimates. A declared resolution policy distinguishes a clear target pass from a near-target candidate; an uncertain first candidate is not labeled a proven minimum. A positive tiny probability remains positive in scientific/logarithmic form.
This is one stationary, pooled queue with Poisson arrivals, independent exponential handling times, identical available agents, first-come-first-served service, unlimited waiting room and no abandonment. The arrival interval defines a rate, not a simulated shift. Reports and charts describe these assumptions rather than guaranteed observed performance.

Worked Example

If you input:

  • Expected arrivals: 120
  • Interval (minutes): 60
  • Average handling time (seconds): 300
  • Available agents: 12
  • Target service level (%): 80
  • Answer within (seconds): 20
  • Limit occupancy: Disabled
  • Maximum occupancy (%): 85 (retained, inactive)

Result:

The offered load is exactly 10 Erlangs. With 12 agents, modeled occupancy is approximately 83.333333%, service within 20 seconds is 60.670742%, and mean wait is 67.408234 seconds. Thirteen agents reach approximately 76.644031%; fourteen reach 86.662742%, so the smallest supported staffing meeting 80% within 20 seconds is 14. At 14 agents, occupancy is approximately 71.428571%. The retained 85% occupancy limit is disabled. This is a steady-state model comparison, not a shift plan or a service guarantee.

Frequently Asked Questions

What does Erlang C assume about the queue?

It assumes independent Poisson arrivals, exponential handling times with the entered mean, identical continuously available agents and one first-come-first-served queue with unlimited waiting space. Nobody abandons or is blocked. These are model assumptions, not facts established about your operation. Multiple skills, priorities, changing arrival rates and finite-interval backlog require a different model.

Does the suggested agent count include breaks or shrinkage?

No. Agents are simultaneously available to handle the modeled work. The result is not a roster, shift allocation or paid headcount estimate, and it does not add breaks, training, absence or shrinkage. Time Tracking totals entered durations, Meeting Time Cost supplies its separate cost estimate, and Critical Path handles dependency planning; none silently supplies workforce assumptions here.

What does answered within the threshold mean?

Service level is the modeled fraction of arrivals whose waiting time before handling starts is at most the entered number of seconds. Handling completion is not included. At zero seconds, arrivals served immediately still count. The waiting-time tail instead reports the fraction waiting longer than each displayed time; its value at zero is the delay probability.

Why are zero demand and an overloaded queue different?

With no modeled arrivals, required agents are zero and expected queue length is zero, assuming no pre-existing work. Arrival-conditioned service and waiting probabilities are unavailable, not an observed 100% success rate. With positive demand at or above capacity, no stationary waiting distribution exists. Occupancy and exact spare capacity can still be shown, but a finite wait is not invented.

Why might a target have no reported minimum?

For positive demand, 100% service within any finite time is mathematically unattainable at finite staffing; an enabled 0% occupancy ceiling is also impossible. A search-limit outcome only says no qualifying candidate was found through 10,000 agents. A near-target outcome gives the earliest possible and first clear candidates without pretending the declared numerical resolution is a certified mathematical error bound.

How are invalid inputs and very small probabilities handled?

Blank required traffic is not zero. Invalid available agents retain a valid target search; invalid target fields retain independent current queue metrics. An invalid threshold removes current threshold-dependent results. Disabled occupancy text remains preserved and ignored numerically. Positive probabilities that underflow ordinary floating point retain scientific and logarithmic values, while a rounded 100% service figure never proves a 100% target.

What input limits and export scope apply?

Expected arrivals allow zero through one million, with a positive minimum of 0.000001; interval is 1–1440 minutes and handling time 0.001–86400 seconds. Agents are integers from 0 through 10,000; percentage targets allow 0–100 and the answer threshold 0–86400 seconds. Numeric atoms admit 12 significant digits and six effective decimal places. The complete CSV records assumptions, retained inputs and the displayed comparison data; it does not export a simulated shift or guarantee real service levels.

Similar Calculators