What is average speed of answer (ASA)?

Average speed of answer (ASA) is the average time a customer waits in the queue before reaching a person, measured from the moment their contact joins the queue to the moment an agent picks it up. It is usually reported in seconds and is one of the oldest standard measures of contact center responsiveness.

ASA is a queue metric. It says nothing about whether the conversation went well, whether the issue was resolved, or whether the customer was happy at the end of it. What it tells you is how long people waited, on average, before anyone spoke to them.

This page covers what ASA includes and excludes, how to calculate it, what a good ASA looks like, why the average hides more than it reveals, how ASA relates to service level, and what to do about a rising number.

ASA in one sentence

ASA is the average wait before a customer reaches a person.

What average speed of answer includes and excludes

The boundaries of the measurement matter more than the arithmetic, and they are where most disagreements about ASA come from.

Included: time the contact spends in the queue waiting for an available person, plus the ring time before the agent connects.

Excluded, in most implementations: time the customer spends navigating an interactive voice response (IVR) menu before they enter the queue. This is the exclusion that causes the most trouble, because from the customer's point of view the clock started when they dialed. A center reporting a 25-second ASA behind a 90-second IVR menu is describing a 115-second experience and reporting a quarter of it.

Also excluded: abandoned contacts. Because ASA averages only the contacts that were actually answered, every customer who gave up waiting disappears from the calculation. This matters more than it sounds, and it is covered below.

How to calculate average speed of answer

ASA = total wait time for answered contacts ÷ number of contacts answered

A worked example. In one hour, 200 contacts are answered. Their combined queue time is 4,800 seconds.

ASA = 4,800 ÷ 200 = 24 seconds

The arithmetic is trivial. The judgment is in the two exclusions above and in one more thing: what happens to the contacts that never got answered.

The abandonment problem

Because ASA only counts answered contacts, it improves when customers give up. Consider an hour where the queue is badly under-staffed. The customers who wait longest are the ones most likely to abandon, and when they abandon they leave the calculation entirely. What remains is a set of shorter waits, and the ASA looks better than the hour actually was.

The effect is real enough that ASA should never be read without the abandonment rate next to it. An ASA that improves while abandonment climbs is not an improvement, it is the same problem measured differently.

What is a good average speed of answer?

Most published ASA targets are single-digit or low double-digit seconds. Measured performance is not close to that.

ContactBabel's UK Contact Centre Decision-Makers' Guide, surveying more than 200 UK contact center operations, put the average speed to answer at 116 seconds in 2024, its second-highest level in 20 years. In Australia, the ACXPA Australian Call Centre Rankings, which use independent mystery shopping rather than self-reported figures, recorded a national average of 109 seconds in 2025, down from 117 seconds in 2024 and 152 seconds in 2023.

Both of those are minutes, not seconds. A target of 20 or 30 seconds is aspirational for most operations rather than typical, and the widely repeated claim that the industry average is around 28 seconds does not survive contact with either dataset.

What actually sets a reasonable ASA target:

The channel. Voice and chat are synchronous, so the customer is waiting in real time and seconds count. Email and messaging are asynchronous, and applying a seconds-based target to them is a category error.

The issue. Someone reporting a fraudulent transaction has a different tolerance for waiting than someone checking a delivery date. Centers that route by topic can set different targets by queue and should.

What was promised. If the site says "typical wait under a minute," that is the target. Consistency with the stated expectation matters more to customers than the absolute number.

ASA vs. service level

These two are measured from the same data and answer different questions. Confusing them is common and consequential.

The example that makes the difference concrete: a queue where half the callers are answered instantly and half wait 48 seconds has the same 24-second ASA as a queue where everybody waits 24 seconds. The first is a much worse experience for half the customers, and ASA cannot see the difference.

This is why most contact centers report service level as their commitment metric and use ASA alongside it for trending. Neither is sufficient alone. ASA gives you a single number that moves sensibly with staffing; service level tells you how many people actually got the experience you promised.

Note also that a service level is an internal operating target. It is not the same thing as a service level agreement (SLA), which is a contractual commitment between two parties.

Why ASA matters

It is the most direct read on staffing adequacy. ASA responds almost immediately to the gap between arriving volume and available people. When it moves, staffing is usually why.

It correlates with abandonment and with satisfaction. Long waits produce abandoned contacts and irritated customers, and the irritation arrives before the conversation starts. An agent picking up a contact after a four-minute wait is beginning at a disadvantage that has nothing to do with them.

It is simple enough to be understood outside the operation. ASA is one of the few contact center metrics that means something to a finance director or a board without explanation, which makes it useful for arguing the case for headcount.

Where ASA misleads

The average conceals the tail. Covered above, and it is the central weakness. Reporting the median and the 90th percentile alongside the mean costs nothing and reveals a great deal.

It improves when customers give up. Also covered above. Always read it with abandonment.

It ignores the IVR. A center can hold a flat ASA while adding menu layers that make the actual customer experience steadily worse. If the IVR is excluded from the metric, someone should still be measuring total time to a person.

It says nothing about resolution. Answering quickly and then failing to solve the problem produces a fast ASA and a repeat contact tomorrow. ASA belongs next to first contact resolution, not instead of it.

Chasing it can distort behavior. Teams pushed hard on ASA sometimes answer contacts faster than they can handle them properly, or move people off complex work to clear a queue. The wait gets shorter and the work gets worse.

How to improve average speed of answer

Forecast and staff at the interval level. Most ASA problems are a staffing mismatch in specific half-hour windows rather than an all-day shortfall. Daily averages will not find them.

Route on the first attempt. Contacts that reach the wrong queue and get transferred wait twice. Routing based on customer history and topic rather than menu selection removes a large share of that.

Reduce the contacts that did not need to happen. The most durable way to shorten a queue is to have fewer people in it. Self-service that genuinely resolves common questions, and proactive updates on the things customers most often chase, both take volume out of the queue rather than moving it faster.

Give people the context to start immediately. A meaningful share of wait time in a poorly integrated operation is the previous contact still being written up. A workspace where history and customer detail are already present shortens after-contact work and returns people to the queue sooner.

Set the expectation. Queue position and estimated wait, offered honestly, reduce abandonment even when the wait itself does not change. A callback option converts a wait into a scheduled contact and is one of the few interventions that improves the customer experience and the metric at the same time.

ASA when AI answers first

When AI handles the front of the queue, ASA needs redefining before it can be read at all.

If AI responds immediately to every arriving contact, the naive ASA calculation collapses toward zero and stops carrying information. That number will look excellent in a board pack and will describe nothing. The measurement that still matters is the wait for the contacts that need a person, which means tracking ASA separately for escalated contacts and treating the moment of escalation as the queue entry point.

There is a second thing worth watching. A customer who has spent four minutes with an AI assistant before asking for a person has already invested four minutes, and their tolerance for a further wait is lower, not higher. The escalation queue deserves a tighter target than the general queue, not a looser one, which is the opposite of what most centers do when they deploy automation.

The short version: report ASA for AI-handled and human-handled contacts separately, measure escalation ASA from the point of escalation, and resist the temptation to publish the blended figure.