Service Level Calculator

Service level is the percentage of contacts answered within a stated number of seconds , the familiar "80/20" means 80% of calls answered within 20 seconds. Most staffing tools ask for the target and give you agents. This calculator runs the same maths backwards: you supply the agents, and it tells you what the customer experience will look like.

The maths behind it

First it converts demand into workload, measured in erlangs: traffic intensity = (calls × AHT) ÷ interval length in seconds. With 250 calls at 240 seconds AHT in a 30-minute interval, that is 33.33 erlangs, meaning you need at least 34 agents just to keep up on average, before any allowance for randomness.

The Erlang C formula then gives the probability that an arriving contact has to wait at all, and service level follows from it:

• SL = 1 − P(wait) × e^(−(agents − erlangs) × target answer time ÷ AHT)
• ASA = P(wait) × AHT ÷ (agents − erlangs)

Why extra agents help so much at the margin

Service level does not improve linearly. Near the point where agents barely cover the workload, one extra agent can move service level by twenty points or more. Once you are comfortably staffed, an extra agent may add less than one point. This is why chasing a 98% target costs far more than reaching 80%, and why small understaffing in a single interval causes a disproportionate collapse in the customer experience.

Assumptions to keep in mind

• Erlang C assumes no abandonment, real service level is usually slightly better than the model predicts.
• It assumes contacts arrive randomly and agents are interchangeable; heavily skill-routed queues need separate modelling per skill.
• It assumes a steady state within the interval, so use 15 or 30 minute intervals rather than modelling a whole day at once.
• Enter agents actually on the phone. Apply shrinkage before this step, not after.