Service level 80/20: what it means and how to choose your target

Ask a contact center manager for their key metric, and most will say service level. The 80/20 numbers are so familiar that teams set them by default without checking whether they fit the business. Here is what the metric means, how to choose a target for your queues, and what to watch alongside it.

What 80/20 means

Service level is the share of calls answered within a threshold time. 80/20 means 80% of calls answered within 20 seconds. The other 20% wait longer, and some of those callers hang up.

• The first number is the share of calls: 80%.
• The second is the threshold in seconds: 20.
• Measure it by interval and by queue, not just for the whole day: 80% for the day can hide 50% in the morning peak and 99% on a quiet evening.

Agree on the formula before you compare numbers. Some teams exclude calls abandoned in the first few seconds – often misdials. Others count every abandoned call as unanswered, and still others drop them from the calculation. The same day can come out at, say, 78% or 85% depending on the rule. Abandoned calls are usually counted, except very short ones – but whatever rule you choose, state it in the report.

Why 80/20

80/20 is a sensible balance for many sales and service lines: most customers get through quickly, and the team isn't overstaffed for rare spikes. It is the most common benchmark, but there is no single standard: depending on the line, teams use 80/20, 80/30, or 90/10. The benchmark fits a typical queue, and a typical queue isn't necessarily yours.

How to choose your target

Start with the customer and the cost of waiting, not with the familiar number.

1. Work out what a lost call costs. On a sales line, it's a lost order; on an information line, it's a repeat call later.
2. Find out what customers expect. Emergency and medical lines often need 90% in 10-15 seconds, while internal queues can do fine with 80% in 60 seconds or more.
3. Check whether customers have other options. If they can request a callback or solve the issue in the IVR, you can relax the voice target.
4. Assess what you can afford. The last few points of service level are the most expensive – we show the numbers below.
5. Set different targets for different queues and times of day instead of one number for everything.

Service level, average speed of answer, and abandoned calls

Service level doesn't tell the whole story. Three more metrics fill the gaps.

• Average speed of answer (ASA) – how long customers waited on average. A few very long waits hide behind a good average.
• Abandonment rate – customers who hung up before an agent answered. These are lost customers, and service level alone won't show them.
• Longest wait time – a simple check that no one waited unreasonably long.

An example. 120 calls an hour, an average handle time of 4 minutes, and 11 agents on the line. By the Erlang C formula, service level is 81%, the 80/20 target is met, and ASA is 20 seconds. But while three-quarters of callers reach an agent right away, the rest wait 80 seconds on average. About 12% of calls wait more than a minute, and about 5% more than two. That's where abandoned calls come from even with a "green" service level.

That's why service level, ASA, and abandoned calls are shown together, by interval. There is no single norm for abandonment rate: it depends on the line and the customers. Watch its trend and the intervals where it climbs. If service level looks good but abandoned calls keep growing, check whether they are being dropped from the calculation.

What a higher target costs

Service level doesn't grow in proportion to headcount. Near the target, each person adds a lot; well past it, very little. Take the same volume – 120 calls an hour at 4 minutes each – on a line open 9:00 AM-9:00 PM, seven days a week, with 30% shrinkage.

• 80/20 – 11 agents on the line, 16 people per shift, 45 on staff.
• 90/10 – 13 agents on the line, 19 people per shift, 54 on staff.
• 95/20 – 14 agents on the line, 20 people per shift, 56 on staff.

Why the last few points cost the most. Calls don't arrive on schedule: one minute may bring five calls, the next none. You keep free agents "in reserve" to absorb these spikes. While the reserve is small, each new person noticeably shortens the queue. Once the reserve is large, few spikes are left for it to absorb, and the next agent mostly waits for calls. In our example, the 11th agent raises service level by 15 points, the 12th by 9, the 13th by 5, and the 14th by less than 3. Yet each one draws a full salary.

Moving from 80/20 to 90/10 adds three people to every shift and nine to the headcount – 20% more payroll, at 160 hours per agent per month (as in the AI agent savings calculator). What do you get in return? Using the Erlang A mode of the calculator, if callers are willing to wait 2 minutes on average, abandoned calls drop to about 1% from 4% – roughly 1,240 fewer a month out of 43,200.

Now say, as in the missed call cost calculator's defaults, every other call is about buying, one in ten of those ends in a sale, and the customer buys once more within a year. In our example, the revenue that 90/10 wins back covers only about 70% of the extra payroll – before you even count the cost of goods. On a line like this, 90/10 pays off only if the average order is at least 40% larger than the calculator's default. And customers on a regular line will barely notice the difference between 20 and 10 seconds. Run this calculation with your own numbers.

1. Calculate agents for two or three target options in the Erlang calculator.
2. Turn the difference into monthly payroll, accounting for shrinkage.
3. Compare it with the value of the calls you will stop losing – the missed call cost calculator is built for this.
4. Pick the target where the extra cost still pays for itself.
5. Revisit the choice when call volume, prices, or customer expectations change.

This turns the debate about the "right number" into a calculation that both finance and the contact center manager can agree on.

Common mistakes

• Looking only at the daily average and missing dips in peak intervals.
• Quietly dropping abandoned calls – the number looks better than the customer experience.
• Setting one target for sales, support, and complaints, although the value of these calls differs a lot.
• Chasing 95% everywhere: staffing costs grow much faster than customer satisfaction.
• Forgetting about chat and messaging. Customers there don't hold on the line, so teams usually track time to first response instead of service level.

How to raise service level without hiring

• Sharpen the forecast by interval and move breaks out of peaks.
• Offer a callback instead of a long wait – the customer isn't lost, and the queue gets shorter. For those who hang up anyway, use automatic callback for missed calls.
• Cut handle time with a knowledge base, suggested replies, and the customer record from your CRM – details in our article How to reduce AHT.
• Hand simple questions like order status to an AI voice agent or the IVR.
• Set up skills-based routing: the call goes straight to an agent who knows the topic, without bouncing between transfers.

Putting it into practice

1. Lock in one service level formula and make sure everyone who reads the reports knows it.
2. Set a target for each queue based on call value and customer expectations – and calculate what each one costs in people and money.
3. Show service level, ASA, and abandoned calls by 15- or 30-minute intervals in one report.
4. Every week, review the worst intervals and decide what to change: headcount, schedule, callbacks, or self-service.
5. Recalculate staffing needs whenever the target or call volume changes.

In CONTACT.WORK reports, service level, ASA, and abandonment rate are calculated by queue and interval from the same call records. The numbers never contradict each other, and the problem hour stands out right away.