Every contact centre dialler or ACD (automatic call distributor) quietly records what each agent is doing, second by second. Those records are the raw material for two of the most quoted workforce numbers in a BPO: occupancy and utilisation. They sound similar, they are often used interchangeably in review meetings, and they measure quite different things. Getting them straight starts with understanding the states an agent moves through in a shift.
The building blocks: agent states
Most telephony and workforce management platforms break logged-in time into a small set of states. The names change from vendor to vendor, but the logic is consistent.
- Talk time: the agent is connected to the customer.
- Hold time: the customer is still on the line but parked while the agent checks something.
- After-call work (ACW or wrap): the call has ended and the agent is updating the CRM, tagging a disposition or sending a follow-up email.
- Available or idle: logged in, ready, waiting for the next contact.
- Aux (auxiliary) codes: logged in but deliberately unavailable, for breaks, team huddles, coaching, training, system issues or personal time. Each reason usually has its own code.
Talk, hold and ACW together make up handle time, the time the agent spends actually working contacts. Handle time plus available time is the time the agent was ready to take or working contacts, sometimes called productive or staffed time. Login time is everything from sign-in to sign-out, aux included.
One shift, worked through
Take an agent on an inbound telecom support process with an 8-hour login. Her state report shows:
| State | Hours |
|---|---|
| Talk + hold | 5.00 |
| After-call work | 0.75 |
| Available (idle) | 0.75 |
| Aux: breaks, huddle, coaching | 1.50 |
| Total login | 8.00 |
Handle time is 5.75 hours. Productive time (handle plus available) is 6.5 hours.
Occupancy is handle time divided by productive time: 5.75 ÷ 6.5 = 88.5%. It answers the question, “Of the time she was ready to work, how much was she actually busy?” Aux time is excluded because she was not available to take contacts then.
Utilisation is the ratio of productive time to login (or paid) time: 6.5 ÷ 8 = 81.25%. It answers a different question: “Of the time she was on the clock, how much was available to customers at all?” The 1.5 hours of aux is what drags it down.
Here is the first trap. Some organisations define utilisation as handle time divided by login time, which gives 5.75 ÷ 8 = 71.9% for the same shift. Neither definition is wrong; they are just different. Before comparing utilisation across two clients, two sites or two vendors, check which formula each one uses, and whether the denominator is login hours, scheduled hours or paid hours. A good MIS report states its formula in a footnote.
What each number is good for
Occupancy is mostly a consequence, not a lever. On an inbound queue it is driven by how many agents are staffed against how many calls arrive. If calls are plentiful and agents are few, occupancy rises; if the floor is overstaffed for the hour, it falls. Queueing maths explains a useful pattern here: for a given service level target, a large team can run at a higher occupancy than a small one, because random gaps between calls are smoothed out across more people. A 15-seat team might hit its service level only at 75–80% occupancy, while a 300-seat team on the same calls can safely run in the high 80s.
Utilisation is more of a management lever. It reflects decisions about breaks, training, meetings and the discipline of aux usage. When utilisation drops, the question is which aux codes grew. Was it planned training for a new product launch, or did “system issue” time quietly double because a CRM screen has started freezing? The second one is a technology problem pretending to be a people problem.
Why “higher” is not automatically “better”
It is tempting to treat both numbers as efficiency scores to be pushed up. That works for a while and then backfires.
Sustained occupancy above roughly 90% on voice work leaves almost no breathing space between calls. Agents go from one upset customer straight to the next. Fatigue shows up first in quality scores and sick leave, then in attrition. Many workforce teams cap planned occupancy somewhere in the 80s for voice and accept higher levels for chat or email, where the work is less emotionally continuous. There is no universal limit, but the direction of risk is well understood.
Very low occupancy has its own cost. If agents are idle 30% of their ready time, the client is paying for capacity that is not being used, or the BPO is absorbing that cost under a fixed price. Low occupancy also hurts skill retention on complex processes; agents who take too few contacts forget the less common scenarios.
Utilisation pushed too high usually means coaching, refresher training and team huddles have been squeezed out. The short-term numbers look excellent. Three months later, quality and first-contact resolution decline because nobody had time to learn the new policy update.
Reading the numbers together
The two metrics become informative when looked at side by side, along with service level.
- High occupancy, poor service level: genuinely understaffed for the volume, or handle time has risen. Check AHT trend and intraday staffing against forecast.
- Low occupancy, poor service level: a scheduling mismatch. Enough people across the day, but not in the right intervals. Typical when shift start times do not follow the arrival curve.
- Low utilisation, normal occupancy: too much aux. Break down aux by code and by team leader; one team often accounts for most of the gap.
- Occupancy well above plan for weeks: the floor is running hot. Expect a lagging hit on quality and attrition even if service level looks fine today.
Where the data itself goes wrong
Both metrics are only as honest as the state data behind them. A few patterns are common enough to watch for.
Agents sometimes stay in ACW to avoid the next call. That inflates handle time and occupancy at the same time, making the floor look busier than it is. A long tail of ACW durations, or ACW that is much longer for some agents than others on the same call types, is the signal. Many processes set an auto-ready timer that pushes agents back to available after a fixed wrap period.
Aux codes get misused when the code list is long and confusing. If “coaching” and “meeting” mean the same thing on the floor, the reports will not tell you anything useful about either. Fewer, clearly defined codes usually produce better data than an exhaustive list.
Back-office and non-voice processes need a different approach. In email or claims processing there is no ACD pushing work at the agent, so “idle” is not measured the same way. Utilisation is usually estimated from transactions completed multiplied by a standard time per transaction, divided by available hours. That works only if the standard times are refreshed whenever the process changes.
Finally, login time is not the same as paid time. An agent paid for a 9-hour shift with a 1-hour unpaid meal break, who logs in 10 minutes late, has a login of 7 hours 50 minutes against 8 paid hours. Whether that gap shows up in utilisation depends on which denominator the report uses, which is one more reason to write the formula down.
Written by the RP Strativa Team. Figures in the examples are illustrative.
Leave a Reply