Should vacancies and attrition be included in shrinkage?

No. Shrinkage describes the time your existing agents are unavailable. Vacant positions are a gap between the headcount you have and the headcount you planned, and are better tracked as a separate vacancy factor on the establishment. Mixing the two hides which problem you actually have.

How do I handle part-time agents?

Shrinkage is a percentage, so it scales. Breaks and meetings that are fixed per day, though, take a larger share of a shorter shift, so part-time groups usually run a few points higher. Calculate shrinkage per contract type if the groups are large enough to matter, or use the contracted hours of the typical agent in the group.

How to calculate contact centre shrinkage

Shrinkage is the share of paid time your agents are not available to handle contacts. It is the difference between the number of agents an Erlang C calculation says you need on the phones and the number you must actually schedule, and it is the input planners most often get wrong, because the temptation is to borrow a benchmark instead of measuring it.

The shrinkage formula

shrinkage = (paid time − time available for contacts) ÷ paid time

There are two ways to fill that in. For planning, add up the categories of lost time from contracts and policy. For measuring, take the states from your ACD and the absence data from HR. Both should land in the same place; when they do not, the gap is usually an unrecorded category.

A worked example

Take an agent contracted for 37.5 hours over five days, which is 7.5 hours a day and 1,950 paid hours a year.

Category	Input	Hours a year
Annual leave	25 days × 7.5 h	187.5
Public holidays	8 days × 7.5 h	60
Sickness	8 days × 7.5 h	60
External subtotal		307.5 (15.8%)
Training and coaching	2 h a week × 52	104
Meetings and one-to-ones	1 h a week × 52	52
Paid breaks	30 min × 5 days × 52	130
Other off-phone time	1 h a week × 52	52
Internal subtotal		338 (17.3%)
Total shrinkage		645.5 h (33.1%)

Each agent is available for contacts 66.9 per cent of the time they are paid, which is 25.1 productive hours from a 37.5-hour week.

External versus internal shrinkage

The split is not cosmetic.

External shrinkage is time out of the building: annual leave, public holidays, sickness, and other absence. It is mostly set by contracts and HR policy, it is forecastable from history, and it has a shape across the year: leave peaks in summer and at Christmas, sickness in winter.

Internal shrinkage is time in the building but off the phones: training, coaching, meetings, breaks, system downtime, admin. It is under the planner’s control. Training and coaching can be moved into the troughs of the demand curve; breaks can be staggered. Reducing internal shrinkage in the peak hour is often cheaper than hiring.

Reporting them separately also makes the number credible with finance. A single 33 per cent figure invites argument; 16 per cent leave and sickness plus 17 per cent of scheduled development and breaks is harder to dispute.

The shrinkage multiplier

To turn a productive requirement into a scheduled one, divide by what is left:

multiplier = 1 ÷ (1 − shrinkage) = 1 ÷ (1 − 0.331) = 1.495

If an Erlang C calculation says you need 20 agents on the phones, you need 20 × 1.495 = 29.9 scheduled, which rounds up to 30. Always round up: you cannot schedule nine tenths of a person, and rounding down puts you short before the interval starts.

The multiplier is why shrinkage matters more than most forecast errors. Moving shrinkage from 30 to 35 per cent raises the multiplier from 1.43 to 1.54, an 8 per cent increase in the agents you must schedule for the same demand. A forecast that is 8 per cent high is a bad week; a shrinkage figure that is five points low is a bad year.

Measuring shrinkage from your own data

ACD states. Every minute an agent is logged in sits in a state. Available and handling are productive; break, meeting, training, coaching and admin states are internal shrinkage. Sum them by week.
HR data. Leave taken, public holidays, and sickness by week. Add anything paid but not logged in, such as a full-day training course.
Measure by week and by interval. Annual shrinkage is one number for budgeting. For rostering you need it by week (seasonality) and ideally by time of day, because breaks and coaching cluster mid-morning and mid-afternoon, and a flat 33 per cent understaffs those intervals.
Reconcile. Paid hours minus productive hours from the ACD should match the category totals. Where it does not, find the missing category rather than adjusting the number.

Common mistakes

Borrowing a benchmark. Industry figures of 30 to 35 per cent are a sanity check, not an input. Your leave policy and your absence rate are what count.
Mixing denominators. Some tools measure shrinkage against scheduled time rather than paid time. Either is fine, but the multiplier only works if the requirement you apply it to uses the same base.
Double counting. If unpaid lunch is outside paid hours, it must not appear in the numerator. If leave is already removed from the hours you schedule against, do not remove it again through shrinkage.
Applying one number all year. July with a 20 per cent leave rate and February with 8 per cent need different rosters.
Ignoring within-interval loss. Late log-ins, early log-offs, and overrunning breaks are shrinkage inside the interval. If you plan at 30-minute resolution, a five-minute late start is a 17 per cent loss for that agent in that interval.

Try it with your own numbers

Inputs

Results

Total shrinkage

33.1%

Share of paid time not available for contacts.

External shrinkage: 15.8%
Internal shrinkage: 17.3%
Shrinkage multiplier: 1.49×
Productive hours per agent per week: 25.1 h
Agents to schedule: 30

Show the working

Paid hours a year = 37.5 × 52 = 1950 h; a working day is 7.50 h.
External = (25 + 8 + 8) days × 7.50 h = 307.5 h, which is 15.8% of paid time.
Internal = (2 + 1 + 1) h × 52 + 30 min × 5 days × 52 ÷ 60 = 338.0 h, which is 17.3%.
Total shrinkage = 15.8% + 17.3% = 33.1%.
Multiplier = 1 ÷ (1 − 0.331) = 1.495; 20 on the phones × 1.495 = 29.9, rounded up to 30.

Doing this for every interval of the week? Pebble WFM computes the requirement from your forecast and builds the roster. Free month, no card needed.

Where next

How to calculate Erlang C staffing: the productive requirement this multiplier is applied to.
How to calculate a staffing requirement: the whole chain, from forecast to agents scheduled.