An operation needs thirty seats covered at peak. The provider hires thirty people. By week three the service level is missed and the conversation turns tense, because the team is complete and contacts are still waiting.
It's almost always the same problem and it has a name: headcount is not capacity. The gap between the two is called shrinkage, and it's the most frequent reason operations launch short without either side seeing it coming.
What shrinkage is
Shrinkage is the share of paid time during which a hired person isn't available to handle work. It isn't wasted or unproductive time. Much of it is necessary: training, coaching, rest. But it isn't time on the line, and that difference is what decides whether the SLA holds.
The usual components:
- Vacation and compensatory time off.
- Initial training and later refreshers. A team doesn't stop learning the day it leaves the classroom.
- Meetings, quality calibrations and one-on-one coaching.
- Breaks within the shift, including those required by labour rules.
- Absenteeism, sick leave and personal leave.
- Late starts, early finishes and dead time between shifts.
- System outages and access failures that leave someone at their desk unable to work.
It's worth splitting these into two groups, because they're managed differently. Planned shrinkage — vacation, training, meetings — is scheduled: somebody decides when it happens. Unplanned shrinkage — absence, sick leave, technical failures — is estimated and watched, but not scheduled. Reporting both as a single number hides exactly the part you can act on.
It's calculated on hours, not on people
The most common arithmetic mistake is reasoning in heads. "I need thirty seats covered, so I hire thirty." That would work if every person were available one hundred per cent of their shift, every week of the year, and nobody ever got sick or went to training. Nobody operates that way.
The correct calculation runs the other way. Start from workload by time slot, convert it into seats occupied per slot, and only then adjust for shrinkage. And the adjustment is a division, not a subtraction: required headcount equals seats on the line divided by one minus shrinkage. Subtract the percentage instead of dividing and the operation is short every time — and shorter the higher the assumption. How the first half of that calculation gets built — the demand curve by time slot — is covered in how to size a support operation.
One more detail that gets skipped: shrinkage isn't uniform across the day or the year. December doesn't look like March. The first week after a training wave doesn't look like month six. A single assumption applied to the whole year will be generous in some months and short in exactly the ones that hurt.
Adherence and occupancy are not the same thing
Three metrics get confused constantly, and confusing them leads to decisions opposite to the ones you need.
- Shrinkage: of paid time, how much wasn't available for work.
- Adherence: how closely what the person actually did matched what their schedule said. Arriving late lowers adherence; taking lunch at a different time than scheduled lowers it too, even if total hours add up.
- Occupancy: of the time they were available, what share was actually spent handling work.
They serve different purposes. Adherence is fixed with supervision and, more often than people admit, with realistic schedules: a schedule nobody can meet produces poor adherence without anyone failing. Occupancy is a load signal, not an efficiency one. Held very high for weeks, it means there's no air between contacts, and that ends in quality errors and early exits. It's one of the least-named causes of attrition in review meetings, and we look at it in BPO attrition.
Very high occupancy doesn't describe an efficient operation: it describes one with no slack. The difference shows up the day a peak arrives.
What to do with the planned part
This is where the quick win is, and it isn't about reducing planned shrinkage but about moving it. Training has to happen; the question is when.
- Refreshers, calibrations and team meetings in the valley slots, never at peak. It sounds obvious and it's the first thing that breaks when the calendar is built by someone who isn't looking at the demand curve.
- Vacation spread across the year, with a cap on people per week and a block on high-season months.
- Meetings with a fixed start and a fixed length. A meeting that runs thirty minutes over in a large team is capacity that vanishes without being recorded anywhere.
- Staggered hiring cohorts where volume allows, so training doesn't pull half the operation off the floor in the same week.
What to do with the unplanned part
Absenteeism is almost never random. Broken down, a pattern appears, and the pattern tells you what to fix.
If it concentrates in one shift, it's a schedule design or a transport problem at that hour. If it concentrates in new hires' first month, the cause sits upstream: selection, or expectations set badly at the offer. If it shows up in one team and not in another with the same schedule and the same profile, it's supervision. Three different diagnoses, three different fixes, and none of them visible if the report only carries the monthly average.
When shrinkage becomes an excuse
It gets misused in two directions. The first is on the provider side: explaining every miss with absenteeism, month after month, with no breakdown and no plan. Shrinkage isn't weather. It's measured, explained and managed; if the report doesn't separate it, there's no management, there's narrative.
The second is on the client side: refusing to accept it exists and demanding that contracted hours equal handled hours. That demand doesn't remove shrinkage, it just pushes it somewhere it can't be seen: no training, no coaching, people working without breaks. Capacity looks right on paper and quality falls over in practice.
What to put in writing
Four things prevent most of the later argument:
- The shrinkage assumption the quote was built on. Without it, the proposal isn't comparable to any other, because the number of people quoted depends directly on that assumption.
- How it gets reported. Planned and unplanned separated, broken down by component.
- What happens when the assumption breaks. If real absence drifts away from the estimate for a sustained period, there has to be a rule agreed in advance: adjust headcount, adjust scope, or adjust the service target. Deciding it in the middle of the crisis always goes worse.
- The conditions the service level is measured against. An SLA built on assumptions the operation doesn't control isn't a commitment, it's a scheduled argument. We work through it in how to write an SLA you can meet.
How we handle it at smartBPO
We put the shrinkage assumption on the table before signing, not after the first missed month, and we explain which components it's made of. In reporting we show it split into planned and unplanned, because those are two different conversations: one is about the calendar, the other about root cause. We schedule training and calibrations against the client's demand curve rather than our internal convenience, and we review absence broken down by shift, by team and by tenure so we act on the pattern instead of the average. When the assumption doesn't hold, we say so with the data next to it and propose the adjustment. We don't promise shrinkage will disappear: we promise it will be measured, explained and built into the sizing from day one.