I have been staffing contact centers for over 30 years. Every one of them ran on some version of this math. The calculator above is the one I wanted when I was building schedules: free, no signup, honest about what Erlang C gets wrong, and it gives you a workbook you can actually take into Excel or Google Sheets.
Use the interval planner for a whole day and the single interval calculator for a quick answer. Use the Excel download when you need a whole day, 48 half-hour intervals, with totals at the bottom.
Staffing math tells you how many agents you need. QA tells you how well they are doing. If your QA team can only score a handful of calls per agent, see what automated scoring costs on our OttoQA pricing page.
What you get in the free workbook
- Calculator tab. One interval. Calls, interval length, AHT, service level, shrinkage, occupancy cap and an optional ASA target. Returns agents on the phones, agents to schedule, service level, ASA, occupancy and queue length, plus a staffing table with your answer highlighted.
- Day Planner tab. 48 half-hour intervals. Enter calls and AHT per interval and it staffs every one, then totals calls, workload hours and peak agents for the day.
- Erlang A tab. Add caller patience and an abandon target. You get agents needed, service level, abandon rate, ASA and occupancy.
- Start here tab. How to use it, how to open it in Google Sheets and the formulas written out.
Plain formulas only. No macros, no add-ins, no Python in Excel, no login. It opens in Excel, Google Sheets and LibreOffice.
How many agents do I need? Quick table
Here is a quick answer for common volumes. Calls per half hour, at a 180 second or 300 second average handle time, for 80% of calls answered in 20 seconds, with 30% shrinkage. Erlang C, no abandons. For your own numbers, use the calculator above.
| Calls per 30 min | Erlangs (180s AHT) | Agents on phones (180s) | Agents to schedule (180s) | Erlangs (300s AHT) | Agents on phones (300s) | Agents to schedule (300s) |
|---|---|---|---|---|---|---|
| 25 | 2.5 | 5 | 8 | 4.17 | 7 | 10 |
| 50 | 5 | 8 | 12 | 8.33 | 12 | 18 |
| 100 | 10 | 14 | 20 | 16.67 | 21 | 30 |
| 200 | 20 | 24 | 35 | 33.33 | 39 | 56 |
| 400 | 40 | 45 | 65 | 66.67 | 74 | 106 |
| 1,000 | 100 | 107 | 153 | 166.67 | 176 | 252 |
Look at the bottom rows. 100 Erlangs needs 107 agents, only 7 more than the workload, at 93.5% occupancy. 2.5 Erlangs needs 5, double the workload. Small queues pay a lot more per call for the same service level.
How many agents for 100 calls in a half hour? Worked example
Here is the default example in the calculator. 100 calls in 30 minutes. 180 second average handle time. Target is 80% of calls answered in 20 seconds. Shrinkage is 30%.
Step 1: workload. 100 calls x 180 seconds = 18,000 seconds of talk. Divide by the 1,800 seconds in the interval and you get 10 Erlangs. That means 10 people talking nonstop for 30 minutes, with zero waiting room for the next call.
Step 2: find the agent count. 10 agents won't work. The queue never clears. Erlang C tells you what happens at each staffing level:
- 12 agents: 64.0% in 20 seconds, ASA 40 seconds.
- 13 agents: 79.6%. ASA 17 seconds. Close, but it misses.
- 14 agents: 88.8% in 20 seconds, ASA about 7.8 seconds, occupancy 71.4%.
- 15 agents: 94.1%, ASA 3.7 seconds.
14 is your number on the phones. Look at how fast service level moves. One agent is the difference between missing and beating the goal by nine points. That is why you look at the table, not just the answer.
Step 3: shrinkage. 14 / (1 - 0.30) = 20.0, rounded up to 20. You schedule 20 agents to have 14 on the phones.
Bottom line: 10 Erlangs of work, 14 agents on the phones, 20 on the schedule.
What service level will N agents give me?
Sometimes the schedule is already set and you just want to know what it buys you. Use the "Check a staffing level" box in the single interval calculator. Type the agents you will have on the phones and it shows service level, ASA, occupancy and, in Erlang A mode, abandons. The chart under it plots service level at each agent count with your target as a line.
Same example, 100 calls in 30 minutes at 180 seconds: 12 agents give you 64.0% in 20 seconds. 13 give you 79.6%. 14 give you 88.8%. If you only have 13 on the phones, now you know you will land just under 80/20.
80/20 service level explained
80/20 means 80% of calls answered within 20 seconds. It is the most common voice target, but there is nothing magic about it. It is a choice about how long you are willing to let people wait, and it drives your headcount.
In this calculator, service level counts calls answered within the time limit divided by all calls offered. In Erlang A mode, a caller who hangs up counts as missed. Check how your phone system counts abandons before you compare.
Here is what different targets cost for the same 10 Erlangs (100 calls in 30 minutes at 180 seconds):
| Target | Agents on phones | Service level you get | ASA |
|---|---|---|---|
| 70/30 | 13 | 82.7% | 17 seconds |
| 80/30 | 13 | 82.7% | 17 seconds |
| 80/20 | 14 | 88.8% | 7.8 seconds |
| 90/20 | 15 | 94.1% | 3.7 seconds |
| 95/20 | 16 | 97.1% | 1.7 seconds |
Going from 80/20 to 95/20 costs 2 more agents on every interval in this example. Pick the target your customers need, not the one somebody used 20 years ago. ASA (average speed of answer) is the other common target. Set it in the ASA field and the calculator meets both.
Erlang C overstaffs. Use Erlang A when you know your abandons.
Erlang C assumes every caller waits forever. Nobody does that. Real callers hang up, and every caller who hangs up stops adding to the queue. So Erlang C overstates the wait and tells you to hire more people than you need. On a small queue it might be one agent. On a big queue that runs hot, it can be a lot.
Erlang A fixes this by adding one input: average caller patience. It is the M/M/N+M model, first worked out by Palm in 1946 and made practical for call centers by Garnett, Mandelbaum and Reiman in 2002. Same example, with patience set at 120 seconds:
- Erlang A says 12 agents hit 80/20. Service level 82.6%. But 6.2% of callers abandon.
- If you also set a 5% abandon target, it goes to 13 agents with 3.8% abandons.
Read that carefully. Erlang A is not telling you to cut staff and call it a day. It is telling you that part of your service level is coming from people giving up. Put an abandon target in. A queue that hits SL by losing 10% of callers is not a good queue.
Getting patience is the hard part. A rough place to start is average time to abandon divided by abandon rate, from your phone system reports for the same intervals. Then check the model against what really happened last week. If predicted abandons are way off, adjust patience.
Erlang B vs Erlang C vs Erlang A
Three formulas, three different assumptions about what happens when every agent is busy.
| Model | When all agents are busy | Use it for | Same 10 Erlangs |
|---|---|---|---|
| Erlang B | The caller is blocked (busy signal). Nobody waits. | Phone lines and trunks | 18 lines keep blocking under 1% (0.71%) |
| Erlang C | The caller waits forever. Nobody hangs up. | Agent staffing, the standard | 14 agents: 88.8% in 20 seconds, 17.4% of callers wait |
| Erlang A | The caller waits, then hangs up if the wait runs past their patience. | Agent staffing when you know your abandons | 14 agents, 120 second patience: 93.3% in 20 seconds, 2.28% abandon |
Erlang B is still the right tool for sizing lines. For agents, Erlang C is the standard and it is a little pessimistic. Erlang A is the most realistic if you have a decent patience number. This calculator does C and A, and the formula section below shows the Erlang B recursion it uses under the hood.
Shrinkage goes last. Always.
I see this mistake all the time. Someone bumps AHT or the service level target to "cover" breaks and training. Don't. Erlang only answers one question: how many people need to be on the phones right now. Then you divide by (1 - shrinkage) to get how many to schedule.
Shrinkage is everything that keeps a paid agent off the phones: breaks, lunch, coaching, training, meetings, sick time, PTO. 30% to 35% is common. If you have never measured yours, measure it before you trust any staffing plan. Coaching time counts here too. If you are adding QA sessions, see how many QA evaluations per agent actually makes sense, because every session comes out of your shrinkage.
Call center staffing calculator: shrinkage to FTE
Erlang gives you agents on the phones for one interval. To get to a schedule and a headcount, do it in two steps.
Step 1: agents to schedule. Divide by (1 - shrinkage) and round up. For 14 agents on the phones:
| Shrinkage | Agents to schedule |
|---|---|
| 20% | 18 |
| 25% | 19 |
| 30% | 20 |
| 35% | 22 |
| 40% | 24 |
Step 2: FTE for the day. Add up scheduled agents x interval length across the day to get scheduled agent-hours, then divide by paid hours per FTE. Say you need 14 on the phones for 20 half-hour intervals (10 hours) at 30% shrinkage. That is 20 scheduled x 10 hours = 200 scheduled agent-hours. At 8 paid hours per FTE, that is 25 FTE for the day.
The interval planner at the top of this page does this for a whole day. It totals agent-hours and FTE, and the 8 paid hours per FTE is an input you can change.
Occupancy and burnout
Here is the other thing Erlang C won't tell you. Big queues are efficient. At 500 Erlangs, Erlang C says 510 agents hit 80/20. That is 98% occupancy. Every agent finishes a call and gets the next one instantly, all day.
That looks great in a spreadsheet. On the floor it is how you burn people out and lose them. I would cap voice occupancy somewhere around 85% to 90%. Set the max occupancy field and the calculator adds agents until you are under it. In the 500 Erlang example, an 85% cap means 589 agents on the phones, not 510. The calculator tells you when the cap, not service level, is setting the number.
And if you are staffing that tight, your quality numbers will show it. Rushed agents skip steps. That is one of the call center QA mistakes that is really a staffing problem.
Erlang C formula, with a worked example
No black box. Here is exactly what runs, in the browser and in the workbook. a is traffic in Erlangs, N is agents, T is the answer time target.
The textbook form, and a worked example. You will usually see Erlang C written with factorials:
It is the same number as the recursion. Run it for 100 calls in 30 minutes at 180 seconds (a = 10) with N = 14 agents and T = 20 seconds:
- Erlang B at 14 agents: B(14) = 0.0568.
- Chance a caller waits: C = 14 x 0.0568 / (14 - 10 x (1 - 0.0568)) = 0.1741, so 17.4% of callers wait.
- Service level: 1 - 0.1741 x e^(-(14 - 10) x 20 / 180) = 88.8%.
- ASA: 0.1741 x 180 / (14 - 10) = 7.8 seconds. Occupancy: 10 / 14 = 71.4%.
At 13 agents the same steps give B(13) = 0.0843, C = 0.2853 and a service level of 79.6%, which misses 80/20. So 14 it is.
Why the recursion? The textbook version uses a^N / N!. That number overflows a spreadsheet around 170 agents. The recursion never does, so the web calculator handles a 20,000 Erlang interval without blinking.
Erlang A uses the full birth and death model with patience. Abandon rate = average queue / (patience x call rate). Service level in Erlang A is the share of all offered calls answered within T. The web calculator computes it with an exact numerical method (uniformization). The Excel Erlang A tab uses an exact closed form built on the beta distribution (BETA.DIST in Excel and Google Sheets). The two match each other to better than one part in a million.
We tested all three versions against a separate high-precision reference before publishing: the browser code, the recalculated workbook and an independent Python model. Agent counts matched in every case we ran, from half an Erlang up to 20,000.
Erlang C Excel template: how to use it
- Download the .xlsx. No email, no macros.
- Open it in Excel, or in Google Sheets: upload it to Google Drive, then Open with > Google Sheets.
- Calculator tab: type over the blue cells for one interval: calls, interval minutes, AHT, service level target and seconds, shrinkage, max occupancy and an optional ASA target. You get agents on the phones, agents to schedule and the full staffing table.
- Day Planner tab: type calls (and AHT if it changes) for each of 48 half hours. Every row staffs itself, with totals at the bottom. Its max occupancy defaults to 85%, while the planner on this page defaults to 100% (no cap) and also offers 15- and 60-minute intervals, so set the tab to 100% before comparing the two.
- Erlang A tab: add caller patience and an abandon target, plus an optional service level target. It shows how many fewer agents you need once abandons count.
- Copy the Day Planner into your schedule, or export the interval planner on this page to Excel with the same columns.
Every formula is plain Excel (no add-ins), and the Start here tab writes out the math.
How it compares to other free Erlang tools
There are good tools out there. Here is a straight comparison based on what each page said when we checked it on October 27, 2026. Where a page didn't say, we wrote "not stated".
| Tool | Cost | Signup | Format | Abandons (Erlang A) | Full-day intervals |
|---|---|---|---|---|---|
| OttoQA (this page) | Free | None | Web calculator plus .xlsx (Excel, Google Sheets) | Yes | Yes, 48 half hours in the workbook |
| erlang.com Erlang C calculator | Free online. Excel version is the paid "Erlang for Excel" add-in | Not stated | Web, one hour of traffic | Not stated on the Erlang C page | Not stated |
| Call Centre Helper Excel calculator | Free | Newsletter signup | .xlsx that needs Python in Excel and a Microsoft 365 sign-in | Not stated | Not stated |
| Call Centre Helper online calculator | Free | Listing says no log-in | Web | Yes, per its listing | Yes, Day Planner per its listing |
| DataHub Pro Erlang C template | Free | Not stated | .xlsx, no macros. Opens in Excel, Google Sheets, Numbers, LibreOffice | Not stated | Yes, day plan sheet |
Checked October 27, 2026. The Call Centre Helper online calculator page did not load for us directly, so that row is from its search listing. Features change. Check each page before you choose.
When to use the multi-skill version instead
This page is one queue. If you are staffing voice, chat and email across several skills, use the WFM scheduling calculator. To see what share of your calls your QA team actually reviews, try the QA coverage calculator. More free tools are on our free tools page.
FAQ
Is this Erlang C calculator really free?
Yes. The calculator on this page and the Excel download are free. No signup, no email gate, no macros.
Does the Excel file work in Google Sheets?
Yes. Upload the .xlsx to Google Drive and open it with Google Sheets, or use File > Import. Every formula in the workbook is one Google Sheets supports. The Start here tab has the steps.
How many agents do I need for 100 calls in 30 minutes at a 180 second AHT?
That is 10 Erlangs. For 80% of calls answered in 20 seconds you need 14 agents on the phones. At 30% shrinkage you schedule 20.
What service level will 13 agents give me for 100 calls in 30 minutes?
At a 180 second AHT that is 10 Erlangs. 13 agents give about 79.6% of calls answered in 20 seconds under Erlang C, just under an 80/20 target. Use the Check a staffing level box to try any number.
What is the difference between Erlang C and Erlang A?
Erlang C assumes nobody hangs up. Erlang A adds caller patience, so some callers abandon. Real callers do abandon, which is why Erlang C tends to overstaff. Use Erlang A when you know your average patience.
How do I estimate caller patience for Erlang A?
A common estimate is the total time all callers waited (answered and abandoned) divided by the number of abandoned calls, over the same set of intervals. Start with your phone system's abandon data, then check whether predicted abandons match what you actually saw.
Where does shrinkage go in the math?
Last. Erlang tells you how many agents need to be on the phones. Then you divide by (1 minus shrinkage) to get how many to schedule. Shrinkage never goes into the Erlang formula itself.
Why does the calculator stop me at an occupancy cap?
Because Erlang C will happily staff a big queue at 97% occupancy. Nobody can sustain that. The cap adds agents until occupancy is under your limit.
How big a queue can it handle?
The web calculator runs up to 50,000 Erlangs per interval in Erlang C mode and 10,000 in Erlang A mode. The workbook handles 5,900 Erlangs on the Calculator tab, 900 per half hour on the Day Planner and 300 on the Erlang A tab.
Bottom line: get the agents-on-phones number right with the calculator, add shrinkage last, keep occupancy under a cap your people can live with, and use Erlang A once you trust your patience data.