Every contact center is pulled in two directions. Customers want their call answered quickly. Finance wants agents to be busy, not sitting idle. These two goals map to two metrics — service level and occupancy — and pushing one up always pushes the other down. Understanding that tradeoff is one of the most useful things a WFM analyst can do.
The Two Metrics, Defined
Service level is the percentage of contacts answered within a target time. "80/20" means 80% of calls answered within 20 seconds.
Occupancy is the percentage of logged-in time an agent spends handling contacts (talk time plus after-call work), as opposed to waiting for the next one.
High service level means customers rarely wait, which requires agents to often be sitting idle and ready. That's low occupancy. High occupancy means agents are almost always busy, which means calls are often queueing. That's low service level.
A Real Example
Take a half-hour interval with 100 calls and an average handle time of 5 minutes (300 seconds). That's about 16.7 agents' worth of work. Running Erlang C for that interval:
| Agents | Service level (20 sec) | Occupancy |
|---|---|---|
| 17 | 11% | 98% |
| 18 | 39% | 93% |
| 19 | 59% | 88% |
| 20 | 73% | 83% |
| 21 | 83% | 79% |
| 22 | 89% | 76% |
| 23 | 93% | 72% |
| 24 | 96% | 69% |
Two things stand out. First, the very first agent above the bare minimum matters enormously: going from 17 to 18 agents lifts service level by nearly 30 points. (We covered this in The Power of One.) Second, the gains flatten out fast — the jump from 23 to 24 agents buys three points of service level but costs an entire salary.
What Does a Higher Target Cost?
Using the same interval:
- 80/20 needs 21 agents
- 90/20 needs 23 agents
- 95/20 needs 24 agents
Raising the target from 80% to 90% takes you from 21 agents to 23 — about 10% more headcount for ten more points of service. Whether that's worth it depends on your business. For a hospital or a bank's fraud line, absolutely. For a general inquiries queue, probably not.
Want to see this with real numbers? Open the calculator with a sample call center already loaded, then swap in your own.
Try it with sample data →Why Bigger Teams Have It Easier
The same math is kinder to large queues. Here's the staffing needed to hit 80/20 at a 5-minute AHT:
- 20 calls per half hour: 6 agents — occupancy only 56%
- 100 calls per half hour: 21 agents — occupancy 79%
- 400 calls per half hour: 74 agents — occupancy 90%
This is the "economies of scale" effect in queueing. A small team has to carry a lot of idle capacity to absorb random bursts. A large pooled team can run at much higher occupancy and still hit the same service level. It's also why splitting one team into several small skills can quietly raise your cost.
How High Is Too High for Occupancy?
A common benchmark is that sustained occupancy above roughly 85–90% starts to cause problems. Agents get no recovery time between contacts, quality drops, handle times creep up, and attrition rises. And since longer handle times demand even more agents, a stressed team can spiral.
That's why a good staffing plan checks occupancy as a guardrail, not just a result. If your target service level requires 94% occupancy in some interval, that's a signal: either relax the target for that period, add a few agents, or find ways to shorten handle time.
Practical tip: Set a maximum occupancy cap (for example 85–90%) when you run your staffing model. If the cap is binding in an interval, that's where you're asking too much of your team.
How to Choose a Target
- Start from the customer. How long will your customers actually tolerate waiting? Check your abandonment rate against average speed of answer.
- Price the options. Run 80/20, 85/20, and 90/20 and look at the headcount difference.
- Check the occupancy at each. If the agent experience at 90/20 is 92% occupancy all day, you've picked a target your team can't sustain.
- Revisit it. Targets set years ago are often just inherited. Re-evaluate them against current business needs.
The Bottom Line
Service level and occupancy are two sides of one trade. Treating them separately is how contact centers end up with a service level goal that's mathematically impossible at the occupancy they're willing to run. Model both, choose deliberately, and you'll spend a lot less time explaining misses.