Setting the right service level target

Leadership · Forecasting · ~7 minute read

The target you inherited probably wasn’t set deliberately

Most contact centres run against a service level target that was set so long ago that nobody currently in the operation remembers the analysis behind it. Eighty percent of calls answered in twenty seconds — the famous “80/20” — is the most common example, and almost none of the operations using it can explain why those specific numbers fit their business better than any other pair. The target is, in effect, traditional. The cost of running to a traditional target is real: the wrong SL target staffs the operation either too generously or too sparingly relative to what customers and the business would actually choose, and the gap can run to a significant fraction of the wage bill in either direction. This article walks through what a service level target really measures, the trade-offs that should drive it, and how to set one deliberately for your operation.

What service level actually measures

Service level is the percentage of contacts answered within a defined target time. It is a queue-experience metric, not a quality metric: it says nothing about whether the conversation that followed was useful, fast, or pleasant. It also says nothing about the customers who hung up before being answered — that is abandonment, a related but distinct measure. Some operations report a blended “service level after abandons”; some exclude abandons; some weight short and long abandons differently. The definition matters less than the consistency: whatever you choose, apply it the same way every time.

The trade-off triangle

A service level target sits at the intersection of three things: cost, customer experience, and competitive position. Cost is the most directly modellable — using Erlang or simulation, you can map service level against required headcount and produce a curve that shows the cost of every additional percentage point of SL. The curve is non-linear and steepens sharply at the high end. Moving from 70/20 to 80/20 might cost ten percent more agents. Moving from 90/20 to 95/20 might cost twenty percent more. The shape of the curve is the most important fact in the target-setting conversation, and most operations have never actually plotted it.

Customer experience is the harder side of the trade-off. The relationship between SL and customer satisfaction is not linear either: customers tolerate short waits well, tolerate medium waits with increasing irritation, and abandon or complain past a threshold that depends on the service, the brand, and the alternative. The threshold is researchable through customer interviews, satisfaction surveys, abandonment patterns, and the “sweet spot” analysis that some BPOs sell as a service. Without that research, the target is a guess.

Competitive position is the third corner. In some sectors, customers compare wait experiences across providers and switch when one is consistently worse. In others, the operation is the only viable provider and the customer has nowhere to go. A telecoms challenger needs an SL target that protects against switching. A government service department, with a captive audience, can rationally set lower SL targets because the customer cannot vote with their feet. Knowing where your operation sits on this spectrum is part of setting the target.

80/20 isn’t sacred — where it comes from

The 80/20 target is widely attributed to a 1970s standard adopted by the telecoms industry and propagated outward through Erlang training materials and WFM platform defaults. It is a reasonable point on a curve, but it is not a law of nature. Operations that match their target to their actual cost/CX curve typically land somewhere else entirely — a low-margin, low-CX-sensitivity operation might rationally target 60/30 and run profitably; a premium financial-services operation might target 90/15 because the cost of customer impatience is high. Setting your target to 80/20 because everyone else does is the same kind of analysis as setting your salary band to whatever the recruiter quoted — it works, but it leaves money on the table.

Different targets for different channels

A common mistake is treating service level as a single concept across all channels. Voice customers tolerate short waits poorly. Chat customers tolerate slightly longer waits as long as the conversation moves once started. Email customers expect a response in hours, not seconds. The right target for each channel is a function of the customer’s expectation, which is itself a function of the medium. Operations that apply a single SL target across channels either over-staff the asynchronous channels or under-staff the synchronous ones, and rarely get either right.

How to set a defensible target

A target-setting exercise that consistently works has three steps. The first is to model the cost curve for each channel: produce a table of agents required and wage cost at every SL target from 60/20 to 95/20, holding volume and AHT constant. The numbers come out of an Erlang calculation, easily run from a spreadsheet or the Erlang C calculator on this site. The shape of the curve is the conversation. The second step is to overlay the customer-experience evidence. Abandonment rate, CSAT, and complaint volume against wait time tell you where the customer pain lies. If 30 seconds is the inflection point in CSAT, the target should land where the curve still protects that experience — not five seconds further out because it is cheaper. The third step is the business framing. The target you propose, and the cost it implies, is presented as a trade-off explicitly. Senior management may choose to accept a worse SL to reduce headcount, or to invest in better SL to protect retention. The planner’s job is to make the trade-off visible, not to make the decision.

Reviewing the target periodically

Service level targets are not set once and forgotten. As the operation grows, the customer base shifts, the brand position evolves, the competitive landscape changes, and the cost of headcount moves, the right target moves with them. A sensible review cadence is annually, tied to the budget cycle, with a re-modelling of the cost curve and a fresh look at the customer-experience evidence. The review does not have to result in a change; the point is to confirm that the existing target is still the right one rather than just the inherited one.

Common mistakes

Three patterns recur. The first is using a target inherited from a different operation, a different brand, or a different decade, without checking whether the conditions that justified it still apply. The second is treating SL as the only operational metric, when abandonment, ASA, longest wait, and customer satisfaction together carry information that the SL alone does not. The third is locking the target in too aggressively — for example, demanding 95/20 across all queues without modelling the cost, and discovering that the wage bill rose by 18% in pursuit of a target nobody can justify.

Conclusion

A service level target is one of the most consequential numbers in a contact centre. It drives the headcount, the cost, the customer experience, and the planning team’s working life. Setting it deliberately, against real cost-curve analysis and real customer evidence, is one of the most valuable conversations a planner can prompt with senior management. Done well, the conversation becomes a strategic one rather than an operational one, and the operation as a whole becomes more honest about the trade-offs it is making. Done badly — or never done at all — the operation runs to a number nobody can explain, and pays a quiet but real cost for the privilege.

Model your own cost curve using the Erlang C calculator. Pair this with forecasting for managers and leaders for the governance conversation that should accompany the target-setting exercise.