What does it take to staff 1000 calls an hour?

1000 calls an hour at a 240-second handle time brings 66.7 erlangs of work, and the smallest team that clears 80/20 is 74 agents on the phones, at 84.6 percent answered within 20 seconds and a 9.3-second average wait. Scheduling 106 at 30 percent shrinkage keeps 74 in seats, and at this scale that hourly figure is the floor for an interval-level plan rather than the plan itself.

How many agents should be scheduled for 1000 calls an hour after shrinkage?

106. The distance between 74 on the phones and 106 on the schedule is 32 salaries of shrinkage, which at this scale is the largest single line item the Erlang number hides. Managing breaks, meetings, and training placement at 1000 calls an hour frees more capacity than most routing projects.

Why is Erlang C unreliable at 1000 calls per hour?

Two of its assumptions fail at this scale. It assumes no caller ever abandons, but at 90.1 percent occupancy real callers hang up and relieve the queue, so the formula runs pessimistic. It also assumes calls arrive evenly across the hour, which never happens, so the hourly answer of 74 agents overstaffs some 15-minute intervals and understaffs others. Treat 74 as the floor and plan at the interval level.

How close is 73 agents to holding 1000 calls an hour?

Exactly at the line. Seventy-three agents on the phones hit the 80/20 target with zero buffer for forecast error or a heavy handle time day, and 72 drop the service level to 73 percent. At 66.7 erlangs the curve is steep near the target, so the last two agents are worth several points each.

Should you staff 1000 calls an hour by the hour or by 15-minute intervals?

By 15-minute intervals. The 74-agent answer assumes the 1000 calls arrive evenly, and real arrival curves do not, so one hourly number overstaffs the quiet intervals and understaffs the surges at the same time. Let the hourly Erlang figure size the budget, then let interval data place the people.

The call center staffing ladder

Call center staffing

How many agents do you need for 1000 calls an hour?

Erlang C says 74 agents on the phones and 106 scheduled, and at this volume that answer is where the planning starts, not where it ends.

The short answer

Handling 1000 calls an hour at a 4-minute average handle time with an 80/20 service level takes 74 agents on the phones; at 30 percent shrinkage you schedule 106.

Calls per hour: 1000
Offered load: 66.7 erlangs
Agents on phones: 74
Scheduled agents: 106
Service level hit: 84.6%
Occupancy: 90.1%

Why the math lands there

1000 calls an hour at 240 seconds of average handle time is 240,000 seconds of talk, hold, and wrap-up work, or 66.7 erlangs of offered load. Erlang C puts 74 agents on the phones against that load: 84.6 percent of calls answered within 20 seconds, a 9.3 second average speed of answer, and 90.1 percent occupancy. The floor below is close. 73 agents hit the 80 percent target exactly, and 72 drops the service level to 73 percent.

Those 74 are agents on the phones, logged in and available. With 30 percent shrinkage for breaks, meetings, training, and absence, you schedule 106 to reliably field 74. The reward for this scale is the pooling economy: 7.4 agents on the phones per 100 calls, the strongest ratio on this ladder. A queue this large absorbs the randomness of individual calls so efficiently that almost every scheduled body is doing measurable work.

The catch is that at this volume Erlang C stops being a staffing plan and becomes a floor. The formula assumes a flat hour and infinitely patient callers, and both assumptions fail harder at 1000 calls an hour than anywhere below it. Real planning at 74 agents means 15-minute staffing curves, intraday reforecasting against actuals, and planned occupancy relief.

11000 calls an hour times 240 seconds of handle time is 66.7 erlangs of offered load: 66.7 hours of phone work arriving every hour.
2Counting up from 67 agents, 74 is the first count where Erlang C clears the 80/20 target: 84.6 percent of calls answered within 20 seconds, with an average wait of 9.3 seconds.
3At 74 agents, occupancy is 90.1 percent: the share of each logged-in hour spent handling calls rather than waiting for the next one.
4Breaks, meetings, training, and absence take agents off the phones for 30 percent of paid time, so keeping 74 on the phones means scheduling 106.

One agent either way

Erlang math is a cliff, not a slope. This is the same hour with one or two agents more or fewer on the phones.

Agents on phones	Answered in 20s	Average wait	Occupancy
72	73.4%	18.7s	92.6%
73	79.7%	13s	91.3%
74the answer	84.6%	9.3s	90.1%
75	88.4%	6.7s	88.9%
76	91.4%	4.8s	87.7%

Occupancy check: at 74agents, 90.1 percent of every logged-in hour is call handling. Above 85 percent, a queue that looks calm on paper is brutal in the chair; plan headcount above the Erlang minimum or the roster will shed people faster than you can hire them.

If your handle time or target differs

Agents on the phones for 1000 calls an hour, across handle times and the two most common service level targets. Adjust every input live in our Erlang C calculator.

Average handle time	80/20 target	90/30 target
3 minutes	56 agents	57 agents
4 minutesthis page	74 agents	75 agents
5 minutes	91 agents	93 agents
6 minutes	109 agents	111 agents

Scheduled headcount sits above every number here: this page assumes 30 percent shrinkage, and your own rate comes out of our shrinkage tool.

What Erlang C does not tell you

Erlang C assumes every caller waits as long as it takes. At 90.1 percent occupancy that assumption breaks, because real callers facing a queue hang up, and every abandonment relieves the queue in a way the formula does not model. The practical effect is that Erlang C runs slightly pessimistic at this volume: the measured service level at 74 agents will often land above the predicted 84.6 percent because abandons thinned the queue for everyone behind them. Do not budget on that relief. A service level propped up by hang-ups is being met by the callers you failed.

The formula also assumes the 1000 calls arrive evenly across the hour, and they never do. A morning surge can push one 15-minute interval well above the hourly rate while the interval after it runs under, and at 74 agents a swing of a few percent is several bodies in either direction. This is the scale where the planning unit changes from the hour to the interval: build 15-minute staffing curves from your own arrival history, reforecast during the day as actuals come in, and move breaks and off-phone work into the troughs. The 74 is the correct answer to an averaged question that no real hour ever asks.

Finally, 90.1 percent occupancy is what the model tolerates for one hour, not a set point you can hold across a shift. Sustained at that level, handle times drift up and the queue feedback turns vicious, because at 66.7 erlangs there is almost no slack between a stable queue and a growing one. The plan around 74 needs relief designed into it: off-phone blocks, flexible lunch placement, and a clear escalation path for the intervals where the reforecast says the floor has moved.

The agent count is the input; the schedule is the work. Turning 106 scheduled agents into shifts that follow your call curve, respect breaks, and survive swaps is what Soon builds automatically, with intraday coverage tracking when the forecast misses.

Frequently asked questions

What does it take to staff 1000 calls an hour?: 1000 calls an hour at a 240-second handle time brings 66.7 erlangs of work, and the smallest team that clears 80/20 is 74 agents on the phones, at 84.6 percent answered within 20 seconds and a 9.3-second average wait. Scheduling 106 at 30 percent shrinkage keeps 74 in seats, and at this scale that hourly figure is the floor for an interval-level plan rather than the plan itself.
How many agents should be scheduled for 1000 calls an hour after shrinkage?: 106. The distance between 74 on the phones and 106 on the schedule is 32 salaries of shrinkage, which at this scale is the largest single line item the Erlang number hides. Managing breaks, meetings, and training placement at 1000 calls an hour frees more capacity than most routing projects.
Why is Erlang C unreliable at 1000 calls per hour?: Two of its assumptions fail at this scale. It assumes no caller ever abandons, but at 90.1 percent occupancy real callers hang up and relieve the queue, so the formula runs pessimistic. It also assumes calls arrive evenly across the hour, which never happens, so the hourly answer of 74 agents overstaffs some 15-minute intervals and understaffs others. Treat 74 as the floor and plan at the interval level.
How close is 73 agents to holding 1000 calls an hour?: Exactly at the line. Seventy-three agents on the phones hit the 80/20 target with zero buffer for forecast error or a heavy handle time day, and 72 drop the service level to 73 percent. At 66.7 erlangs the curve is steep near the target, so the last two agents are worth several points each.
Should you staff 1000 calls an hour by the hour or by 15-minute intervals?: By 15-minute intervals. The 74-agent answer assumes the 1000 calls arrive evenly, and real arrival curves do not, so one hourly number overstaffs the quiet intervals and understaffs the surges at the same time. Let the hourly Erlang figure size the budget, then let interval data place the people.

How many agents do you need for 500 calls an hour?

The Erlang answer is 39 agents on the phones and 56 scheduled, but 85.5 percent occupancy makes 39 a floor to staff above, not a target.

39 on phones · 56 scheduled · 80/20

See the math →

How many agents do you need for 2000 calls a day?

About 25 agents on the phones at peak and 36 on the schedule, and the volume where per-day arithmetic officially hands off to interval planning.

25 on phones · 36 scheduled · 80/20

See the math →

How many agents do you need for 200 calls an hour?

17 agents on the phones, 25 on the schedule, and exactly one point of margin over the 80/20 target.

17 on phones · 25 scheduled · 80/20

See the math →