Shrinkage in Call Centre: Formula and a 5-Person Roster
Shrinkage in a call centre is the share of paid hours when agents cannot take calls. Divide hours off the phones by paid hours, then roster by quiet and busy half hours.
By Tej Pandya, Founder · Updated 12 Nov 2026

Key takeaways
- Shrinkage is hours off the phones divided by paid scheduled hours, times 100. To have N agents free, roster N ÷ (1 − shrinkage), rounded up. Never add the percentage to N.
- For a small team, the half hour is the decision, not the percentage: each half hour's call volume sets how many agents must be on the phones.
- Put breaks, lunch and coaching in half hours with spare agents. Grant a day's leave only if that weekday's busiest slot needs at most one fewer agent than you have, and the leave day still has enough spare slots for lunch and breaks.
- Take the volumes from your call log and peak-hour report, build the grid in a spreadsheet, and recount shrinkage every week.
In this article
- 1What is shrinkage in a call centre and how do you calculate it?
- 2How many agents must you roster to cover shrinkage?
- 3Why does an average shrinkage figure mislead a five-agent team?
- 4Who can be on a break or leave in the busiest half hour?
- 5How do you build this roster from your own call log?
- 6Get your busiest half hours from your call log
- 7Frequently asked questions
A five-person phone team rarely loses calls on an average day. It loses them in one or two half hours. If your busiest half hour of the morning needs all five people on the phones, one person at the tea counter then costs you answered calls, however healthy your monthly shrinkage figure looks.
That is the gap in the usual advice, which stops at a percentage and a multiplier. The percentage says how much paid time disappears. It does not say which half hours can afford the loss, and for a small team that is the whole decision.
Below, you get the formula and the mistake that makes people under-roster. Then comes one worked day for five people: which half hours need everyone, which can spare one or two, and where lunch, coaching and leave can go. The volumes come from your own call log, so you can repeat the exercise this week with a spreadsheet and the lookup table below.
What is shrinkage in a call centre and how do you calculate it?
Shrinkage is the percentage of paid, scheduled hours in which an agent, meaning a person who answers calls, is not available to take them. Calculate it as hours off the phones divided by paid scheduled hours, times 100. Weekly offs are not scheduled hours, so leave them out of both numbers.
Hours off the phones come in two kinds. Planned shrinkage is time you schedule: breaks, lunch, coaching, team huddles, approved leave. Unplanned shrinkage is time that arrives uninvited: sickness, late logins, a dead headset.
Here is one week for five agents working 10:00 to 18:00, five days. These are illustrative inputs in our worked example, not customer data.
| Item | Hours | Share of 200 paid hours |
|---|---|---|
| Breaks and lunch (1 hour a day each) | 25 | 12.5% |
| Team huddle (1 hour each) | 5 | 2.5% |
| Coaching (2 hours each) | 10 | 5% |
| Planned leave (one day, one agent) | 8 | 4% |
Planned total | 48 | 24% |
| Sickness and late logins | 8 | 4% |
Total | 56 | 28% |
Paid scheduled hours are 5 agents × 5 days × 8 hours = 200, so shrinkage is 56 ÷ 200 × 100 = 28%. If lunch is unpaid on your roster, take it out of both sides of the sum.
For a month, add up the hours off the phones across all weeks and divide once by the month's paid scheduled hours. Do not average the weekly percentages, because a week with more paid hours should weigh more.
How many agents must you roster to cover shrinkage?
Roster the agents you need on the phones divided by (1 − shrinkage), then round up. Shrinkage is a share of rostered hours, not of the hours you need, so adding the percentage under-staffs. For 10 agents free at 30% shrinkage, 10 ÷ 0.70 = 14.3, so roster 15.
Adding 30% instead gives 13, and the check shows why that fails: 30% of 13 agents is 3.9 off, leaving 9.1 free, which is short of 10. Of 15 rostered, 4.5 are off, leaving 10.5 free.
Call Centre Helper, a contact-centre trade site, reports that most professionals put shrinkage at 30 to 35%, and cites Dimension Data's 35% average (global contact-centre figures, not Indian data). One agent's leave moves a five-agent team's figure by several points, so use your own number from the weekly count above, and treat benchmarks only as a sanity check.
The multiplier also assumes shrinkage lands evenly across the day. For five agents it does not, and that is where the real decision sits.

Why does an average shrinkage figure mislead a five-agent team?
Calls arrive unevenly, so service is lost in the busiest half hours, not in the daily average. A peer-reviewed review of call-centre operations (Gans, Koole and Mandelbaum, 2003) reports one centre whose offered calls peaked from 11:00 to 11:30 a.m., dipped over lunch, and peaked again from 2:30 to 3:00 p.m.
The same review notes that service capacity cannot be inventoried. A spare agent in a quiet half hour cannot be saved for a busy one.
Standard practice sets agents per half hour with the Erlang C formula. It is a queueing calculation that turns the number of calls and the average call length into agents needed. It assumes calls arrive at random around a steady average within the half hour, call lengths vary randomly around their average, callers are served first come, first served, and nobody hangs up while waiting. Erlang C ignores callers who hang up, so it tends to recommend more agents than a real queue needs. Treat the table as a cautious starting point.
We ran it with these inputs, as our model and not an official figure:
- Workload (Erlangs): A = calls in the half hour × AHT (average handle time: talk time plus hold time plus after-call work, in seconds) ÷ 1,800.
- Service level (the share of calls answered within a set time): we use 80% within 20 seconds, so the share answered within 20 seconds = 1 − C × e^(−(N − A) × 20 ÷ AHT). Here N is the agents on the phones and C is the Erlang C chance that a caller waits.
- AHT: 240 seconds (4 minutes); the 240 seconds in this table covers all three.
- Worked case: 20 calls give A = 20 × 240 ÷ 1,800 = 2.67.

| Calls in the half hour | Agents needed on the phones | Answered within 20 s with that many | With one fewer |
|---|---|---|---|
| 4 | 2 | 90.1% | 48.7% |
| 6 | 3 | 95.7% | 79.3% |
| 8 | 3 | 90.9% | 65.7% |
| 10 | 3 | 84.3% | 49.5% |
| 12 | 4 | 92.6% | 75.6% |
| 15 | 4 | 85.3% | 59.1% |
| 20 | 5 | 86.7% | 66.1% |
| 22 | 5 | 81.5% | 55.9% |
| 24 | 6 | 89.9% | 75.2% |
Round your call count up to the next row. At 22 calls, five agents only just hold 80%, and at 24 calls you need six, which a team of five does not have. A longer average call raises every number, so recompute with your own handle time. Average speed of answer is a different measure, the mean wait, and our guide shows how to count it.
Who can be on a break or leave in the busiest half hour?
Nobody can be off in a half hour that needs all five agents, and one or two can be off in half hours that need four or three. This grid applies the table above to a made-up day for five agents, 10:00 to 18:00, shaped like a two-peak day with a lunch dip. The volumes are ours and illustrative.
| Half hour | Calls | Agents needed | Can be off the phones |
|---|---|---|---|
| 10:00 | 10 | 3 | 2 |
| 10:30 | 15 | 4 | 1 |
11:00 | 20 | 5 | 0 |
11:30 | 20 | 5 | 0 |
| 12:00 | 15 | 4 | 1 |
| 12:30 | 10 | 3 | 2 |
| 13:00 | 6 | 3 | 2 |
| 13:30 | 6 | 3 | 2 |
| 14:00 | 10 | 3 | 2 |
| 14:30 | 12 | 4 | 1 |
| 15:00 | 15 | 4 | 1 |
| 15:30 | 15 | 4 | 1 |
| 16:00 | 12 | 4 | 1 |
| 16:30 | 10 | 3 | 2 |
| 17:00 | 8 | 3 | 2 |
| 17:30 | 6 | 3 | 2 |
The day carries 190 calls and needs 58 agent-half-hours of the 80 rostered (5 agents × 16 half hours). That leaves 22 spare agent-half-hours, or 11 hours, which is 27.5% of the team's 40 paid hours. Four things follow:
- Lunch is 30 minutes and fits between 12:30 and 14:00, where two agents can be off at a time: two at 12:30, two at 13:00, one at 13:30.
- A whole-team huddle does not fit, because no half hour has five spare. Run it in two or three small rounds.
- Leave cannot be granted on this day. The 11:00 and 11:30 half hours are locked, meaning nobody may be off.
- Sickness at 11:00 takes the half hour from 87% to 66% answered within 20 seconds, so cover for locked half hours matters more than the weekly percentage.
The rest of each agent's hour a day is 30 minutes of breaks, five a day across the team. They go at 10:00 (two agents), then 10:30, 12:00 and 14:00 (one agent each). Coaching, 2 hours a day across the team or four agent-half-hours, goes at 17:00 and 17:30, two agents each. Lunch takes 5 of the 22 spare agent-half-hours, breaks 5 and coaching 4, so 14 are used and 8 remain for the weekly huddle rounds.
Five days like this give 55 spare hours, against 56 hours of shrinkage in the weekly example, including a leave day this profile cannot absorb. We would fix that by moving coaching outside phone hours, granting leave only on lighter days, or adding a sixth agent.
How do you build this roster from your own call log?
Pull four weeks of call volume by time of day, convert each half hour to agents needed, and spend the spare agents on breaks, coaching and leave. In our Sales Call Analytics, the report covers call volume, answer rates, missed calls, talk time, team performance and peak hours, and reports download so you can paste them into a sheet.
- Export four weeks of volume.Pass: each weekday appears four times. Fail: any weekday has fewer than three weeks, so do not plan its leave yet.
- Average the calls per slot by weekday.
Use the smallest slot your report shows. If it shows hours, halve each hour's calls as a starting figure, then round up; this smooths real swings, so treat it as a guide. Pass: every slot has a number. Our guide to call volume covers counting.
- Look up agents neededin the table, rounding calls up; 1 to 3 calls: 2 agents; more than 24 calls: recompute with the formula. Pass: no slot needs more agents than you have. A fail marks a slot where callers will queue or give up.
- Mark spare agentsas team on shift minus agents needed. Slots at zero are locked.
- Place lunch, breaks and coaching only in slots with spare agents,never more people off than the spare count.
- Apply the leave rule.Our recommendation: grant a day's leave only if that weekday's busiest slot needs at most one fewer agent than you have, and the leave day still has enough spare slots for lunch and breaks. Pass: the rule holds for every approved leave.
- Recount shrinkage weekly,as hours off the phones ÷ paid scheduled hours. Also count how often someone was off in a locked slot. Our target is zero.
Our Pro plan, ₹999 + GST a month as of November 2026, includes the analytics dashboard, team management and unlimited users. Routing rules in Team Centre let you set which person incoming calls reach, and our guide to business call forwarding for a team shows who should answer first.
Get your busiest half hours from your call log
Bring four weeks of call volume and see how our Sales Call Analytics reports it by time of day for your team. Talk to Our Expert or write to support@groweasy.ai.
Get your busiest half hours from your call log
Frequently asked questions
A five-agent team should aim for no one off the phones in a half hour its grid marks as locked, not for a target percentage. The benchmark from Call Centre Helper covers global contact centres, so use it only as a sanity check and track your own figure as a trend.
Shrinkage is the share of paid hours agents are not available to take calls. Occupancy is the share of available time they spend busy on calls. A 28% shrinkage figure does not set occupancy. See our agent occupancy calculator and ceiling table.
Sources
- 1.Telephone Call Centers: Tutorial, Review, and Research Prospects, Gans, N., Koole, G. and Mandelbaum, A. (2003), Manufacturing & Service Operations Management 5(2), INFORMS, 2003
- 2.What is Call Centre Shrinkage and How to Calculate It?, Call Centre Helper
- 3.Call Analytics, TalkEasy
- 4.Our own calculations: the weekly shrinkage example, the gross-up checks, the Erlang C table and the five-agent day grid


