Erlang

An Erlang is the standard unit of telecom traffic. One Erlang equals one hour of continuous voice activity on a single line, or 60 minutes of call time spread across many lines. Call center planners use Erlang formulas to size trunks, seats, and schedules.

The unit is named after Agner Krarup Erlang, the Danish mathematician who published the first traffic engineering paper in 1909. His work still frames how contact centers forecast staffing today, even inside cloud-native platforms.

The word carries a second, unrelated meaning. Erlang is also a functional programming language built at Ericsson in 1986 for telecom switching. This entry covers the traffic-unit sense, the one call center operators use daily.

Erlang math matters commercially. Under-staff a queue by two agents and service levels collapse within minutes — over-staff it and margin evaporates. That’s why business process outsourcing contracts price voice work per hour, not per call.

Key takeaways

• One Erlang equals one hour of continuous voice traffic on a single circuit or agent.
• Erlang B, Extended Erlang B, and Erlang C convert forecast traffic into required headcount.
• Erlang C assumes callers queue, making it the default staffing model for inbound service.
• Agner Krarup Erlang’s 1909 traffic paper still supplies the underlying mathematics.
• NICE, Genesys, and Verint embed Erlang C inside their forecasting engines.

How it works

An Erlang measures traffic intensity: the average number of concurrent calls in progress over a defined period. If a queue handles 30 hours of talk time within one hour, that’s 30 Erlangs. Planners feed that figure into Erlang B or Erlang C.

Erlang C needs only four inputs: forecast contacts per interval, average handle time, the target answer speed, and the interval length. Get any one of them wrong and the headcount it returns will be confidently wrong too.

Three formulas dominate the field, and all three are still in daily use. Each treats blocked calls differently, which is why the choice matters more than it looks.

Erlang B assumes blocked callers vanish, so it’s the tighter formula. Telcos and outbound teams use it to size trunks where a busy signal simply ends the call.

Extended Erlang B layers in a retry rate. It’s the more realistic choice for outbound campaigns, where dialers redial and callers try again a minute later.

Erlang C assumes blocked callers wait in queue. That makes it the default for inbound call center staffing, because most customer-service callers hold on hook rather than abandon straight away.

Interval choice changes the answer. A 60-minute interval smooths a lunchtime spike into the average and understaffs the peak, which is why most planners model at 15- or 30-minute grain.

Erlang also assumes calls arrive randomly, which mostly holds. It breaks when a marketing email or an outage drops thousands of calls into one interval, so planners cover those spikes with manual overrides instead.

A working planner runs the numbers roughly like this:

1. Forecast call volume for a 30-minute interval. 300 calls at a six-minute average handle time equals 30 Erlangs.
2. Set a service level target. Answering 80% of calls within 20 seconds is the long-standing industry norm.
3. Run Erlang C to find the minimum agent count that hits the target.
4. Add shrinkage for breaks, adherence, and training. Most operations allow 25% to 35%.
5. Publish the headcount to the workforce management scheduler, interval by interval.

Shrinkage is where Erlang plans most often fail. Erlang C might return 42 agents on the phone, but at 30% shrinkage you need roughly 60 names on the roster to keep 42 of them available.

Erlang C returns two numbers, not one. Alongside required agents it reports occupancy, the share of paid time agents spend on calls — and sustained occupancy above 85% reliably predicts burnout and attrition.

Examples

Erlang math shapes staffing across every major BPO destination. Forecasting teams in Manila, Bogotá, Kraków, and Cape Town convert half-hour call volumes into Erlangs, then compute the seat count that hits contracted service levels.

Concentrix, a Fremont-headquartered customer experience provider, reported roughly 440,000 employees in its FY2024 annual filing. Its enterprise contracts run Erlang C forecasting inside a NICE workforce management stack, recalculated every 30 minutes.

Teleperformance, a Paris-based customer experience giant, runs voice campaigns in the Philippines, Colombia, India, and Portugal. Erlang-based scheduling is standard across its front-office queues — usually on Verint or Genesys.

Genesys Cloud CX, one of the largest cloud contact center platforms, documents Erlang C as the default forecasting method inside its workforce management module, with simulation offered as an alternative.

In the Philippines, the largest offshore voice market, Erlang output drives the night-shift rosters that cover North American business hours. A single interval miss there surfaces as a service-level penalty the following month.

By 2024, most enterprise WFM suites had moved from nightly batch forecasts to intraday re-forecasting, rerunning Erlang C every 15 minutes against the arrival pattern actually observed that day. The formula didn’t change; the cadence around it did.

Smaller operators without a full WFM licence run the same maths in free spreadsheet calculators published by erlang.com and other independent tools. The formula doesn’t care about licence tiers, and the arithmetic is identical.

Related terms

Erlang sits inside a small cluster of workforce planning terms, and the numbers only make sense together. These six entries cover the inputs Erlang consumes, the targets it solves for, and the outputs schedulers act on.

• Workforce Management: the wider discipline that folds Erlang forecasting into scheduling, adherence, and shrinkage.
• Average Handle Time: the seconds-per-call input every Erlang calculation depends on.
• Service Level Agreement: the answer-within-X-seconds target Erlang C solves for.
• Call Center: the operational venue where Erlang math is applied every day.
• Occupancy: the utilization output Erlang C reports alongside required agents.
• Shrinkage: the buffer added on top of Erlang output to cover breaks and off-phone time.

FAQ

What does one Erlang mean in a call center?

One Erlang equals one hour of continuous voice traffic. That can be a single 60-minute call, six 10-minute calls, or any mix whose talk time totals an hour. The unit says nothing about headcount until you pick a service target.

What’s the difference between Erlang B and Erlang C?

Erlang B assumes blocked callers hang up and never return, while Erlang C assumes they wait in queue. Inbound customer service almost always uses Erlang C. Trunk sizing and outbound dialer capacity usually use Erlang B.

Is Erlang C still accurate for modern contact centers?

Erlang C still delivers reliable staffing estimates for stable inbound voice queues. It overstates required agents when abandonment is high, so many WFM platforms blend it with discrete-event simulation — a correction that matters most in high-abandon queues.

Who invented the Erlang formula?

Danish mathematician Agner Krarup Erlang published the first traffic-loss formula in 1909 while working at the Copenhagen Telephone Company. The International Consultative Committee on Telephony adopted his name as the official unit of traffic in 1946.

Do chat and email channels use Erlang?

Erlang C was built for one-call-at-a-time voice work, so most WFM tools switch to modified formulas or simulation for chat and email.

Getting the Erlang math right is table stakes for voice work, so explore verified BPO providers whose workforce management practice is already mature.