Average speed of answer formula explained (plus benchmarks for modern contact centers)

Your Monday dashboard shows the average speed of answer (ASA) at 25 seconds, comfortably inside the target. The same report shows abandonment up for the third month running, and this week's customer satisfaction score (CSAT) verbatims mention long hold waits twice.

Those numbers cannot all be telling the truth.

ASA is the metric you staff against, report to the executive team, and write into business process outsourcing (BPO) contracts. It may also be describing a queue that no longer exists in the form the formula assumes. Before you defend that 25-second figure, ask a precise question: what does the number actually count, and what does it quietly leave out?

What ASA measures (and what it doesn't)

Average speed of answer is the metric that reports how long, on average, a caller waits in the queue before a human agent picks up. It measures the average time it takes for an agent to answer an incoming call, including the time a customer spends waiting in a queue before their call is answered by a live agent, providing a clear indication of how quickly your team responds to customer inquiries. It is a queue-side metric: ASA describes what the caller experiences after routing.

ASA is closely tied to two adjacent measurements that get conflated with it and should not be. Average handle time (AHT) measures the human agent's side of the interaction, from pickup through wrap-up, and belongs on the same dashboard because it shows how long answered interactions take after pickup. Service level uses the same queue data to report the share of calls human agents answer within a set number of seconds. ASA tells you only the average wait for answered callers, not individual wait times or the experience of callers who never got through.

The average speed of answer formula and its measurement boundary

The formula to calculate the average speed of answer is the total waiting time for answered calls divided by the total number of answered calls. If your contact center answered 500 calls yesterday and those callers waited a combined 5,000 seconds, your ASA is 10 seconds. The arithmetic is not the hard part; deciding what counts as "wait time" and which calls count as "answered" is.

The ASA clock has a precise start and stop, and four rules define it:

• Queue time: ASA includes it. The clock starts when the call enters the queue after routing completes.

• Hold time before first answer: ASA includes it. The time a caller spends holding before a human agent first picks up counts toward the total.

• Interactive Voice Response (IVR) and routing time: ASA excludes menu and routing seconds from the queue time.

• Abandoned calls: ASA excludes them from the denominator. ASA does not count queue hang-ups as answered contacts; automatic call distributor (ACD) settings determine numerator treatment.

The IVR, routing, and abandonment exclusions are where ASA loses the caller experience, and they matter more than the arithmetic itself. A caller who spends 45 seconds in the IVR and then 20 seconds in the queue experiences a 65-second wait; ASA reports 20 seconds. That gap between felt wait and reported wait is the reason a healthy-looking ASA can coexist with rising abandonment and complaints about long holds.

ASA benchmarks vendors cite (and the numbers operations see)

Ask what a good ASA looks like, and you will often hear software vendors cite low fixed-second cross-industry averages. A Head of Customer Experience (CX) should ask whose operation that benchmark describes.

The average speed of answer is commonly cited at 28 seconds industry-wide, and a good ASA typically ranges from 20 to 30 seconds, a benchmark widely accepted across various industries. But those blended figures obscure a wide range once you split by sector. Large, complex service operations can report ASA well past a minute, especially in high-volume regulated queues with complex work. Lower vendor-cited averages and higher real-world figures can both be directionally useful because they describe different populations.

High-volume regulated operations should not measure themselves against a cross-industry average, because that average blends in low-complexity, low-volume desks whose queue math has nothing in common with a regulated claims or fraud line. Choosing the correct reference class means matching call complexity and service-level commitment before setting a number.

One ratio-based target scales with call complexity by tying ASA to AHT. It predates AI agent deployment entirely, but the logic holds: a queue of complex claims calls earns a longer acceptable wait than a queue of password resets. The practical sequence for a Head of CX is to choose the reference class before setting the target, then read ASA alongside the rest of your call center efficiency metrics.

Why a healthy ASA can hide a deteriorating caller experience

A green ASA next to rising abandonment is the expected output of how the formula is built, because ASA has three predictable blind spots:

• Abandoned calls are excluded from the calculation: callers who give up waiting never count toward ASA, which introduces survivorship bias into the metric. The longest, angriest waits are the ones the number cannot see.

• Zero-wait calls dilute the average: Every call answered instantly adds a zero to the numerator and a one to the denominator. A strong headline figure can mask a long tail of callers waiting many multiples of the reported average.

• The average conceals the distribution: A healthy ASA can still leave a small share of callers waiting far longer than the headline figure suggests. ASA does not represent what happens to individual callers.

Each blind spot widens as volume grows: abandoned-call exclusions multiply, and the long tail becomes harder to see in the average. If abandonment drifts up while ASA holds steady, your call abandonment rate is the number telling the truth. Volume volatility compounds the problem. Forrester predicts that at least three major brands will see single-day call volume spikes 100 times above normal on six separate occasions in 2026, and an ASA computed across a spike day is nearly meaningless to the callers within it.

How AI agents change ASA inputs

Voice AI at the front of the phone channel changes the inputs to the average speed of answer formula.

The mechanics change in several places:

• Automated contacts answer in effectively zero seconds: BER Airport achieved zero wait times across 4 languages: the AI voice agent picks up on the first ring, whether one call or a thousand arrive. Putting AI agents at the front of the service line can reduce caller wait time by answering at the queue level, instantly, at whatever volume arrives.

• Deflection reshapes the residual human queue: Simple, high-frequency contacts leave the queue. Residual human ASA measures the longer, harder calls that remain and needs its own target.

• Concurrent capacity decouples spikes from queue time: Human queues convert volume surges directly into wait. AI agents handle calls in parallel, so a surge increases throughput rather than hold times.

• Escalation wait becomes its own clock: The measurable wait shifts to the handoff from the AI agent to the human agent. Escalations need separate reporting inside the residual human ASA.

Report the handoff wait from the AI agent to the human agent on its own clock within the residual human ASA, rather than folding it into the blended figure. A Head of CX who keeps reporting a single blended ASA after deploying AI agents is averaging two different operations into one number. The instant-answer automated volume drags that figure down, hiding whatever is happening in the harder human queue.

Recalculate what the average speed of answer formula means now

If AI agents answer part of your volume, your ASA target and reporting structure need a new reference benchmark: one figure for AI-answered contacts and one for the residual human queue.

Parloa's AI Agent Management Platform separates reporting for AI-answered contacts and the residual human queue. Its AI agents pick up instantly at enterprise call volumes in 140+ languages, managed across Design and Integrate, Test and Iterate, Deploy and Scale, Monitor and Improve, and Secure, so answer-speed results hold in production after pilot.

Book a demo to see how AI agents cut your average speed of answer at enterprise scale. Every caller who hangs up unanswered is the distance between what they needed and what your contact center delivered.

FAQs about average speed of answer

What is the average speed of answer formula?

Divide the total wait time for answered calls by the total number of answered calls. A center that answers 500 calls, totaling 5,000 seconds of wait time, reports a 10-second ASA.

What is a good average speed of answer?

The typical average speed of answer per industry standards is 28 seconds, with a good ASA generally falling in the 20 to 30-second range. That said, the right target varies by industry; pick a reference class that matches your call mix, volume, and service-level commitment before choosing a number.

Does ASA include IVR time?

No. The ASA clock starts after the IVR menus and routing are complete. A caller feels wait; therefore, it runs longer than the reported figure, because everything before the queue is invisible to the metric.

Do abandoned calls count toward ASA?

No. ASA removes callers who hang up in the queue from both sides of the formula, which makes this exclusion the metric's biggest blind spot. The worst waits are the ones ASA never records.

What is the difference between ASA and service level?

Service level is a threshold metric: the share of calls human agents answer within a set number of seconds. ASA is an average computed from the same queue data, so the two move together but answer different questions.