Agent Occupancy: A Small-Team Calculator and Ceiling Table

• A small inbound team has a lower occupancy ceiling than 85%. In our Erlang C model (80% of calls answered within 20 seconds, 4-minute calls) the ceiling is 60% for 5 agents, 72% for 10 and 81% for 20.
• Occupancy is (talk + hold + after-call work minutes) ÷ logged-in minutes × 100. Measure it in your busiest hour, not across the whole day.
• Push a 5-agent line to 85% and, in our model, 65% of callers wait, and the average caller waits about 3.5 minutes.
• Outbound telecalling has no queue, so the ceiling table does not apply. Treat 85% as the upper limit and watch whether handling time climbs.

Agent occupancy is the share of an agent's logged-in time spent on calls and call-related work: talk time, hold time and after-call work (ACW), which is the note-writing and updating an agent does after a call ends (Call Centre Helper). Enter the values below, replacing our starting values with your own.

Part 1: the occupancy you have

Occupancy % = (talk + hold + ACW) ÷ logged-in × 100. Starting values: (120 + 10 + 30) ÷ 300 × 100 = 160 ÷ 300 × 100 = 53%.

Occupancy is not utilisation. Utilisation, as this page uses the term, divides productive time by all paid time, so it includes meetings, training and breaks. Occupancy counts only logged-in time, including time on calls and after-call work.

Part 2: the occupancy your call volume forces on you

1. Offered load in erlangs = calls per hour × AHT in minutes ÷ 60. One erlang is one agent busy for a whole hour (Wikipedia: Erlang). Starting values: 40 × 4 ÷ 60 = 2.67 erlangs.
2. Forced occupancy % = offered load ÷ agents × 100 (Call Centre Helper). Starting values: 2.67 ÷ 5 = 53%.
3. Look up your team size in the ceiling table below. Five agents have a ceiling of 60%, so the line has about 7 points of headroom.

The ceiling also converts to calls. Five agents at 60% carry 3.0 erlangs, which is 3.0 × 60 ÷ 4 = 45 calls an hour. The line takes 40, so it passes. Take one agent away and the ceiling falls to 55%, about 33 calls an hour. Forced occupancy becomes 2.67 ÷ 4 = 67%, so callers would wait longer than the target allows.

If you want to know where your handling time comes from, see average call duration and after-call work.

Ceilings run from 39% for 2 agents to 89% for 50, and they climb with team size. Our model (below) finds the highest occupancy at which 80% of calls are still answered within 20 seconds, for calls averaging 4 minutes.

Our model, so you can redo it: we used the Erlang C waiting probability (Wikipedia: Erlang). The service-level and average-wait expressions below are our derivation from the Erlang C waiting probability, assuming exponentially distributed call lengths: share answered within T seconds = 1 − P(wait) × e^(−(agents − load) × T ÷ AHT). For each team size we raised the load in steps of 0.001 erlang and kept the highest load that still met the target, then divided by agents. We ran it in October 2026. Assumptions: calls arrive at random, callers never hang up while waiting, each agent takes one call at a time, nobody is on leave or break, call lengths are exponentially distributed around the average, the queue has no limit, and all agents handle every call.

Changing the service target moves the ceiling. With the same 4-minute calls:

Call length barely moves the ceiling. At 80/20 with 5 agents, 2-minute calls give 63%, 4-minute calls 60% and 6-minute calls 59%.

One Call Centre Helper reader reports about 45% occupancy at a 99%-in-5-seconds target with calls of about 60 seconds (Call Centre Helper reader comments, updated 22 July 2026). Another says centres under 100 agents can run at 70 to 75%. These are practitioner comments, not studies, so use the table as a planning guide, not a guarantee.

Below about five agents, a single lunch break or absence swings the figure sharply. Our recommendation: average your busiest hour over at least five working days before you compare it with the table.

For a small inbound team, yes. At 85% a five-agent line makes about two-thirds of callers wait, and the average wait across all callers runs past three minutes. The usual 85% limit comes from fatigue: 51 minutes of handling in each logged-in hour is 85%, and above it absence and attrition rise (Call Centre Helper, updated June 2025). That is a ceiling on how hard agents are worked. It says nothing about how long callers wait.

Our model for 5 agents and 4-minute calls (same assumptions as above):

Average wait = P(wait) × AHT ÷ (agents − load), with P(wait) from the Erlang C formula (our derivation, same assumptions).

Between 60% and 70% the average wait doubles. Past 80% it grows steeply. High occupancy on a small line therefore costs you callers, and callers who hang up show as abandoned calls, which this model assumes do not happen.

Subtract your forced occupancy from the ceiling for your team size: more than 15 points of gap suggests slack, 0 to 15 points is healthy, and at or above the ceiling callers wait longer than your target. These thresholds are our recommendation.

Measured occupancy (Part 1) and forced occupancy (Part 2) should land within about 5 points of each other for the same hour. Within 5 points is a pass. A bigger gap is a fail: one of your inputs is wrong, so recheck the logged-in minutes and the handling time.

Partly. The formula is the same, but outbound agents are not answering a queue, so there are no callers to keep waiting and the ceiling table does not apply. The limit is how hard you work people: our recommendation is to keep sustained occupancy under 85%, the figure above which absence and attrition rise (Call Centre Helper).

A better outbound check is to compare average handling time in the first and last hours of a shift. Our recommendation: if average handling time in the last hour is more than 10% above the first hour, treat the workload as too heavy. One Call Centre Helper reader comment says places running near 90% occupancy typically see handling time increase (Call Centre Helper reader comments, updated 22 July 2026). For output targets rather than workload, see sales calls per day, and for a full week of call data see sales team productivity.

Your call log gives most of them. Our Sales Call Analytics reports total calls, answer rate, missed calls, average talk time, team performance and peak calling hours, in daily and monthly reports you can download. The peak calling hours show your busiest hour, and total calls and average talk time give you most of the Part 2 inputs. Add hold and after-call work from your own notes to reach handling time, and take logged-in time from your shift sheet.

Missed calls show where the line failed to answer, which is where the ceiling is most likely being breached. If routine questions fill your busiest hour, our AI Receptionist answers calls, handles common questions and captures leads. Calls it resolves do not reach your agents, so subtract them from the peak-hour call count in Part 2; calls it transfers still count.

As of October 2026, TalkEasy Pro costs ₹999 + GST a month and includes unlimited users, call recording and history, and analytics. Extra talktime beyond the 3 hours a day included is ₹49 + GST an hour.