Every voice process has the same daily balancing act. Too few agents and customers wait, abandon calls and complain to the client. Too many and the BPO pays for people sitting idle. Workforce management (WFM) teams solve this with a chain of calculations that is worth understanding even if you never run a WFM tool yourself, because it explains a lot about why rosters, breaks and leave approvals work the way they do.
Step 1: forecast the volume
Planning happens in intervals, usually 30 or 15 minutes. Using history, WFM forecasts how many contacts will arrive in each interval, adjusted for known patterns: Monday mornings, salary days, festival weeks, a client’s marketing campaign or a billing cycle that always triggers queries. A forecast that misses badly makes everything after it wrong, which is why forecast accuracy is tracked as a metric in its own right.
Step 2: convert volume into workload
Multiply contacts by average handle time. Using an illustrative interval: 240 calls in 30 minutes with an AHT of 300 seconds is 72,000 seconds of work. A 30-minute interval has 1,800 seconds, so the workload equals 40 agents’ worth of time, a unit WFM teams call Erlangs, after the Danish engineer A. K. Erlang who developed the underlying queueing theory for telephone exchanges.
Step 3: why 40 agents is not enough
If exactly 40 agents are logged in, every one of them would need to be busy every second, and calls would have to arrive perfectly spaced. In reality calls arrive randomly, bunching up and then going quiet. With exactly enough capacity on average, queues grow without limit during the bunches. You need some spare capacity to absorb that randomness.
How much spare capacity depends on the service level target, often written as 80/20: answer 80% of calls within 20 seconds. The Erlang C formula, built into most WFM tools and available as free spreadsheet calculators, estimates the probability that a caller waits and the share answered within the threshold, for a given number of agents. For our illustrative workload of 40 Erlangs with a 300-second handle time, 45 agents would answer only about 76% of calls within 20 seconds, while 46 agents reach about 82%. So the answer is 46. Tighter targets need more agents, and the extra agents needed rise steeply as the target approaches 100%.
Erlang C has known simplifications. It assumes callers never hang up while waiting, which overstates required staff somewhat. Variants such as Erlang A account for abandonment, and simulation is used for complex multi-skill environments. For a single queue, Erlang C remains the standard starting point.
Step 4: occupancy as a sanity check
Occupancy is the share of logged-in time agents spend actually handling contacts. With 40 Erlangs of work spread across 46 agents, occupancy is about 87%. Sustained occupancy that high is tiring, and many operations try to keep it in the 80s for voice. In larger centres, high occupancy is easier to reach at the same service level. This is the economies-of-scale effect that makes merging small queues attractive.
Step 5: shrinkage turns seats into headcount
The 46 agents must be on the phones. But rostered staff are not available all the time: they take breaks, attend training, coaching and team huddles, go on leave or are absent unexpectedly. All of this is called shrinkage. If shrinkage is 30%, then:
Rostered staff = required agents ÷ (1 − shrinkage) = 46 ÷ 0.70 ≈ 66
A common mistake is to multiply by 1.30 instead, which gives about 60 and leaves the floor short. Dividing by the available fraction is correct because shrinkage is a share of rostered time, not of required time.
Why this explains floor life
Once you see the chain, many operational rules make sense. Breaks are staggered because a whole team going at once collapses the service level in that interval. Leave is approved against a shrinkage budget for the week. Real-time analysts move people between queues or delay training when volume spikes above forecast. Every one of these decisions is an attempt to keep the actual number of available agents close to what the calculation said each interval needs.
Written by the RP Strativa Team. Volumes, handle times and shrinkage in the examples are illustrative.
Leave a Reply