Shrinkage calculator

How this is calculated

Shrinkage is unavailable hours divided by scheduled hours. The number that matters downstream is the gross-up: agents needed on the phones divided by one minus shrinkage gives the headcount you actually roster. At 30% shrinkage, 14 agents on the phones means 20 people scheduled.

shrinkage = unavailableHours / scheduledHours
roster    = agentsOnPhones / (1 − shrinkage)

Worked example

Across 4,000 scheduled hours: 400 on breaks, 200 in training, 120 in meetings, 300 on holiday, 160 lost to sickness and 60 to late starts and downtime. That is 1,240 unavailable hours — 31% shrinkage, split 25.5% planned and 5.5% unplanned. To keep 14 agents on the phones you must roster 20.3 people.

The gross-up is the whole point

An Erlang calculation returns agents needed on the phones. It says nothing about breaks, training or absence. Rostering that raw number puts fewer people on the phones than the model assumed, and the queue misses target even though the maths was right — which is the single most common reason a correct staffing calculation still fails in practice.

The planned-versus-unplanned split matters more than the headline figure. Two centres both running 31% behave completely differently if one is mostly scheduled training and the other is mostly unscheduled absence: the first can be planned around, the second has to be absorbed in real time. Compute your own from payroll and adherence data rather than adopting the 30% figure usually quoted as standard.

Frequently asked questions

How do you calculate call centre shrinkage?

Add up every hour agents are paid but unavailable to take contacts — breaks, training, meetings, coaching, holiday, sickness, system downtime — and divide by total scheduled hours. Splitting the categories matters more than the headline: planned and unplanned shrinkage are managed in completely different ways.

What is a typical shrinkage rate?

30% is the figure most often quoted as standard, with measured averages nearer 26.6%. Anything above about 35% is worth investigating. Your own number is better than any benchmark — compute it from payroll and adherence data rather than adopting a published figure.

What is the difference between planned and unplanned shrinkage?

Planned shrinkage — holiday, training, scheduled meetings — is known in advance and can be scheduled around. Unplanned shrinkage — sickness, late starts, system outages — cannot, and is what real-time management exists to absorb. Two centres with identical 30% shrinkage can behave completely differently depending on the split.

Why does my Erlang calculation still miss target?

Usually because the shrinkage gross-up was skipped. Erlang C returns agents needed on the phones; it says nothing about breaks, training or absence. Rostering that raw number puts fewer people on the phones than the model assumed, and the queue misses target even though the maths was right.

Related reading