Shrinkage is the percentage of paid time your agents are not available to handle contacts. It is the bridge between the staffing requirement your Erlang C model produces and the number of people you actually have to employ, roster and pay for. Get shrinkage wrong and every downstream number, the roster, the hiring plan, the budget, is wrong with it.
The formula
Total shrinkage % = (planned shrinkage hours + unplanned shrinkage hours) / total scheduled hours x 100
Staff to schedule = base staff requirement / (1 - total shrinkage %)
Note the division in the second line. This is where most people go wrong, and we will come back to it.
Planned vs unplanned shrinkage
Planned shrinkage is time you deliberately schedule away from the queue: paid breaks, team meetings, coaching sessions, one-to-ones, training, project work, and system-mandated activities. You control it. It should be forecast, budgeted and placed in the roster where it does the least damage.
Unplanned shrinkage is time you lose without choosing to: sickness, unauthorised absence, late logins, early logouts, extended after-call work, and system outages. You influence it, but you cannot schedule it. Forecast it from historical patterns by day of week and by season.
Splitting the two matters because they are owned by different people. Planning owns planned shrinkage; operations owns unplanned. Reporting a single blended figure hides which of the two is deteriorating.
Worked example 1: hours method
A site has 4,000 scheduled hours in a month.
- Breaks: 400 hours
- Training and coaching: 250 hours
- Team meetings: 150 hours
- Sickness and absence: 320 hours
- Late logins and system downtime: 80 hours
Planned = 400 + 250 + 150 = 800 hours, which is 20% of 4,000. Unplanned = 320 + 80 = 400 hours, which is 10% of 4,000. Total shrinkage = 1,200 / 4,000 = 30%.
Productive time available = 4,000 - 1,200 = 2,800 hours, or 70% of what you pay for.
Worked example 2: applying it to headcount
Your interval-level Erlang C model says you need an average of 100 agents on the phone across the operating day. With 30% shrinkage:
Staff to schedule = 100 / (1 - 0.30) = 100 / 0.70 = 143 agents
Now compare the common shortcut of adding 30%: 100 x 1.30 = 130 agents. That is 13 agents short, a 9% understaff that will show up as missed service levels every single day, and which no amount of intraday management can fix.
The gap widens as shrinkage rises. At 40% shrinkage the correct answer is 167 and the shortcut gives 140: a 16% understaff.
Getting the denominator right
The most frequent source of argument in a shrinkage review is what counts as scheduled hours. Two decisions to make once and document:
- Are unpaid lunches inside or outside scheduled hours? If you exclude unpaid lunch from both the numerator and the denominator, your shrinkage percentage will be lower, and that is fine, as long as your headcount maths uses the same definition.
- Do you count holiday and leave? Annual leave is usually handled separately in the FTE-to-headcount conversion rather than as interval shrinkage. Mixing it into interval shrinkage double-counts it.
Consistency matters more than which convention you pick. What you must never do is change the definition mid-year, because a definition change looks exactly like a performance change.
Shrinkage by interval, not just by day
A single daily average hides a lot. If you schedule all your coaching between 14:00 and 16:00, the shrinkage in those intervals might be 45% while the daily average is 30%. Erlang C is interval-level maths, so shrinkage must be interval-level too. Build a shrinkage profile by 30-minute interval and by day of week, and place discretionary planned shrinkage in your lowest-volume intervals.
Benchmarks and what to do with them
Most contact centres land between 25% and 35% total shrinkage. Under 20% almost always means something is missing from the calculation. Over 40% is worth investigating: it is either a genuine absence problem or a month distorted by a one-off training programme.
Track planned and unplanned separately over time. A rising unplanned figure is one of the earliest reliable warning signs of burnout, usually visible months before it appears in your attrition numbers. Cross-check it against occupancy: sustained occupancy above 92% and rising unplanned shrinkage together are a reliable predictor of an attrition spike.
Use the shrinkage calculator to run your own numbers, then feed the result into the Erlang C calculator to see the headcount impact immediately.