Search this term and you’ll find the same number repeated everywhere: 28 seconds, described as the global standard. Independently measured data tells a different story, and the gap between those two figures is the most useful thing on this page.
What Is Average Speed of Answer?
Average speed of answer, usually shortened to ASA, is the average time a caller spends in queue before an agent answers. Divide total waiting time across answered calls by the number of calls answered. If 3,000 calls were answered with 300,000 seconds of combined waiting, ASA is 100 seconds. The clock starts at queue entry and stops on connection, so IVR navigation happens outside the measurement window.
That definition matches the one used by the Australian Customer Experience Professionals Association, an independent practitioner body that publishes vendor-neutral research (ACXPA glossary).
Worth flagging early: the queue-start point isn’t standardized across platforms. Check your own vendor documentation, whether you’re on Genesys, a legacy PBX, or a cloud system, because a few seconds of definitional difference compounds across thousands of contacts.
How to Calculate Average Speed of Answer
ASA = total queue waiting time for answered calls ÷ number of answered calls
Small worked example: five calls answered after waits of 10, 15, 20, 25 and 30 seconds. Total 100, divided by five, gives ASA of 20 seconds.
Notice what the formula requires. Only answered contacts count. Every caller who gave up before connection sits outside the calculation entirely, which turns out to matter enormously.
What ASA Leaves Out
Two exclusions are built into the metric, and both push the reported figure downward:
- Abandoned calls are excluded. ASA is computed on answered contacts only. The people who waited longest, lost patience and hung up are systematically removed from the average.
- IVR time is excluded. Somebody navigating three minutes of menus before reaching the queue, then waiting thirty seconds, appears in your reporting as a thirty-second wait. Their actual experience was three and a half minutes.
ACXPA identifies both as critical limitations that most operations report without caveat. The second one scales with menu complexity: the deeper your IVR, the further your reported answer speed drifts from what customers lived through.
There’s a practical consequence worth sitting with. Two organizations can post identical figures while delivering wildly different experiences, simply because one funnels callers through a four-level menu and the other picks up on the second ring. Neither is lying. Both are reporting the metric correctly. If you want the real number, measure from call connection through to agent answer and treat queue-only timing as a subset of that.
What Is a Good Average Speed of Answer?
Here’s where published figures and measured reality part company.
Call Centre Helper puts the global average at 28 seconds while noting that variation between centers is wide enough that few universal rules survive contact with an actual operation (Call Centre Helper). Fair enough. But ACXPA’s mystery-shopping program, which measures real waits rather than asking managers to self-report, produced these 2025 sector figures:
| Sector | Measured Wait |
| Internet providers | 45 seconds |
| Energy retailers | 1 min 12 sec |
| Car insurance | 1 min 15 sec |
| Education | 1 min 22 sec |
| Aged care | 1 min 38 sec |
| Local councils | 1 min 47 sec |
| Banks | 3 min 48 sec |
Self-reported data from the same market lands at roughly 97 seconds for inbound sales, 115 for customer service and 130 for technical helpdesks. National measured averages have been falling year on year: 152 seconds in 2023, 117 in 2024, 109 in 2025.
So the honest answer to “what’s good” is that 28 seconds would put you far ahead of most organizations actually being measured. Treat it as an aspiration rather than a floor.
Sector context bends the target further. Research cited by Call Centre Helper indicates callers tolerate longer delays on technical assistance lines than on routine service lines, which fits intuition: somebody troubleshooting a broken connection has already accepted the problem is complicated. Somebody checking an order balance has not. Setting one target across every queue you operate ignores that difference, and in my experience it’s the queues with the shortest patience curves that get neglected first.
COPC makes a related argument about targets generally, cautioning that over-servicing raises cost and depresses occupancy without necessarily buying satisfaction, and recommending a performance band instead of a fixed number (COPC). I find that framing more useful than any single figure.
Where ASA Fails as a Standalone Measure
Three failure modes, roughly in order of how much damage they cause:
- Survivor bias. Because abandons drop out, worsening service can improve your reported answer speed. Callers grow impatient sooner, leave sooner, and the remaining answered contacts show shorter waits. A falling ASA alongside a rising abandon rate is a warning, not a win. This is the failure mode I’d watch hardest, partly because it rewards the wrong thing: a manager under pressure to hit a number gets one for free when the queue deteriorates.
- Averaging hides distribution. A center reporting 60 seconds might be answering half its callers in five seconds and making the rest wait three minutes. Report the median and the 90th percentile alongside the mean and the shape becomes visible.
- Daily figures hide intraday peaks. Monthly reporting is close to useless here. Queue pain concentrates in specific half-hour intervals.
Pairing matters more than precision here. Answer speed read next to abandonment tells you whether short waits reflect good coverage or early surrender. Read next to occupancy, it tells you whether the figure was bought with idle capacity. Read alone, it tells you very little that survives a follow-up question from anyone numerate.
None of this makes the metric worthless. It makes it a measure that needs company, which is why TabaTalk’s overview of inbound contact center metrics treats it as one input among several rather than a headline.
ASA vs Service Level vs Average Wait Time
These three get used interchangeably in meetings, which causes more confusion than almost anything else in queue reporting.
| Measure | What It Captures | Main Weakness |
| Average speed of answer | Mean queue wait across answered calls | Hides distribution, ignores abandons |
| Service level | Share of calls answered inside a threshold, commonly 80% in 20 seconds | Says nothing about the calls that miss it |
| Average wait time | Often used more broadly, sometimes including IVR navigation | Definition varies by platform and by team |
A queue can hit its service level target and still carry a poor answer speed, provided the missing 20% wait long enough. Both belong in the same report.
How to Reduce Your Answer Speed
Causes first, briefly. High ASA usually traces back to forecasting that missed, handle times that crept up, routing that sends contacts to the wrong skill group, or arrival patterns nobody modeled. In Gulf operations that last one is common: Ramadan reshapes the daily curve, prayer times create recurring troughs, and rosters built on imported assumptions leave predictable gaps.
What actually moves the number:
- Forecast at interval level, then staff to it. Erlang C modeling tells you whether a roster can hit a target before you commit to it. Guessing does not.
- Trim average handle time. Agents free up faster, queues drain faster. Thirty seconds off AHT across a team shifts answer speed without adding headcount.
- Route by skill and language on the first prompt. Fewer transfers, less wasted talk time. Arabic and English separation matters here more than most planning documents admit.
- Cut IVR depth. Two levels wherever you can manage it. TabaTalk’s Call Flow Builder makes routing changes without developer time, which suits the small iterative adjustments this work actually requires.
- Offer callbacks. These don’t reduce ASA directly, and some platforms handle the logging oddly, but they reduce abandonment sharply and callers generally prefer them.
- Watch queues live. Real-time contact centre dashboards let a supervisor move people during the peak instead of reviewing it later. Teams managing high call volumes get most of their gains here.
TabaTalk recommends publishing answer speed next to abandonment and the 90th percentile wait, always. One number alone invites the wrong conclusion.
Frequently Asked Questions
What is the average speed of answer formula?
ASA = total queue waiting time for answered calls ÷ total answered calls. If 500 contacts were connected with 50,000 seconds of combined queue time, ASA works out to 100 seconds. Only answered calls enter the calculation, so abandoned contacts are excluded by design. Measurement periods are usually daily or half-hourly, and shorter intervals reveal far more than monthly figures, which flatten out the peaks where problems concentrate.
Does average speed of answer include hold time during a call?
No. ASA covers the period between queue entry and agent connection only. Time a customer spends on hold after the conversation begins, while an agent checks an account or consults a colleague, belongs to average handle time instead. Confusing the two produces reporting that looks fine while callers experience long mid-conversation silences. Track hold duration separately if your agents place people on hold frequently.
Is a very low ASA always good?
Not necessarily. Answer speeds below ten seconds usually indicate capacity sitting idle, which costs money that may buy nothing. COPC has argued that faster answering does not reliably raise satisfaction and that over-servicing carries real expense. Emergency lines and trading desks justify it. Most support operations do not. Set a range that fits your tolerance for cost and risk rather than driving toward zero.
How does ASA relate to service level agreements?
ASA frequently appears in contractual SLAs, particularly for outsourced operations, expressed as a maximum average wait across a reporting period. Read those clauses carefully, since an average target is easier to meet than a percentile one: a supplier can miss badly on a subset of contacts while the mean stays compliant. Percentile-based commitments, such as 90% answered within a set time, close that gap.
What causes ASA to spike suddenly?
Unplanned absence, system outages, marketing activity nobody flagged to operations, and public holidays landing differently than forecast. Sudden spikes usually indicate a demand event rather than a staffing trend. Queue distribution logic can also cause them: if an ACD holds contacts for a specialist group with thin coverage, waits climb in that group while the wider floor looks fine.
Can workforce management software fix a high ASA?
Partially. Workforce tools sharpen forecasting and scheduling, which addresses the most common cause. They cannot fix routing that sends contacts to the wrong queue, handle times inflated by slow systems, or demand that self-service should have absorbed. Treat scheduling as one lever among four or five. Teams that buy workforce engagement software expecting it to solve queue performance alone are usually disappointed within a quarter.
Talk to the TabaTalk Team
Reported answer speed and lived customer experience are rarely the same thing. If you’d like a clear reading of what your queues are doing, including the parts the average hides, we can walk through it with you.
TabaTalk provides cloud contact center software built for Gulf operations, with local numbers, BYOC support and deployment measured in hours. Contact our sales team to review your inbound setup, or ask us to model what interval-level staffing would do to your numbers before you change anything.
