We called 100 radiology practices. Here is what we found.

On Friday 21 August 2026 we rang 100 radiology practices across every Australian state and territory and asked the same thing a patient with a referral would ask: to make a booking. We timed every call from dialling to the moment a person who could take that booking was on the line.

Eight calls out of one hundred were answered directly by a person. For everyone else, the road to a booking ran through a recorded menu, and sometimes through a very long queue behind it. This article is the Friday daytime data in full: the wait times, the state-by-state results, and the pattern inside the big networks that we did not expect to find.

Methodology

Research led by Nathan Sri, with the Triagents Research Team.

One caller placed one call to each of 100 Australian radiology practices on Friday 21 August 2026, starting at 11:38am AEST, using a consistent scenario: a patient holding a referral, asking to book. The sample spans all eight states and territories (NSW 28, VIC 24, QLD 18, WA 11, SA 10, ACT 4, TAS 4, NT 1) and mixes the major national networks, mid-size groups, hospital departments and independent single-site clinics.

For each call we recorded whether a person answered directly or a recorded menu answered first, and the elapsed time from dial to a person able to take the booking. Whenever a real person answered and engaged, we identified the call as a test call once the timing was recorded, rather than holding a fake booking on their line. We also capped our patience at 15 minutes per call: three calls were still on hold at that cap and are recorded as 15 minutes plus, a floor rather than a measurement. One call could not be completed, leaving 99 connected calls: 96 produced a timed handover to a person, and the other three are the calls capped at 15 minutes. Every chart below counts all 99 connected calls, with the three capped calls in the longest band. This is Triagents first-party data; limitations are stated at the end.

The headline numbers

Eight practices answered directly. No menu, no hold music, a person inside the first minute. Our caller’s note on the first of them reads, in full, “AMAZING”, which says something about how rare the experience had become by call five.

Ninety-one calls landed in a recorded menu first. Most menus eventually handed the call to a person.

That accounts for all one hundred: 8 answered by a person directly, 91 answered by a menu, and 1 call that never connected. Every chart and average below is built on the 99 calls that did, and the wait for a person is the number worth staring at.

• Under one minute 35.4% 35
• One to three minutes 33.3% 33
• Three to five minutes 12.1% 12
• Five to ten minutes 11.1% 11
• Ten minutes or longer 8.1% 8

Triagents first-party research. 100 practices called, Friday 21 August 2026, from 11:38am AEST. One call per practice; 99 calls connected, of which 96 yielded a timed handover and 3 were capped at 15 minutes and counted in the longest band.

The median was 1 minute 30 seconds. More than a third of practices produced a person inside a minute, which shows the standard is achievable on an ordinary Friday. But the average was dragged to 3 minutes 3 seconds by the tail: 19 of the 99 connected calls took five minutes or longer, and 8 took ten minutes or longer. The fastest was 11 seconds. The three slowest calls still had us on hold when we gave up at our 15-minute cap, so for them 15 minutes is a floor, not a wait.

What the menu does with your wait

The notes behind the numbers are more instructive than the averages.

Eight of the menus used the hold time to push the caller somewhere else: book online, use the patient portal, press one for the automated phone booking. One practice’s queue offered to take our number and ring back when we reached the front. Two calls, at sister sites of the same network, still had us on hold when we gave up at 15 minutes.

And one menu never delivered a person at all. Its automated questions looped, repeated, and did not transfer. A patient on that line does not file a complaint. They hang up and dial the next practice on the map, and the referral in their hand is valid there too.

None of this is a criticism of the receptionists. Every person who eventually answered was doing exactly what the staffing allowed. The queue exists because phone demand and the people available to meet it are two lines that stopped touching, which is the same gap the lost-call data shows from the other side.

Outcomes by state

The table shows both the median and the average deliberately: the median is what a typical caller experienced, and the gap between the two is the size of each state’s long tail. Calls we abandoned at the 15-minute cap enter the table as 15m+, and any average that includes one is a floor.

Victoria was the fastest phone in the sample: the shortest median wait, the shortest average, and half of all the direct answers in the country. Every direct answer we recorded came from Victoria, New South Wales or Western Australia. Across Queensland, South Australia, the ACT, Tasmania and the Northern Territory, 37 practices, not one call was answered without a menu first.

Queensland is the state to watch in the table, and it is the reason the average column is there. Its median of 1 minute 17 seconds is the third best in the country, behind Victoria and Western Australia, yet its average blows out to 3 minutes 57 seconds, because when a Queensland queue was slow it was very slow: its two slowest calls still had us on hold when we gave up at the 15-minute cap, the equal longest waits in the study. South Australia was the slowest mainland state on both measures. The ACT, Tasmania and Northern Territory samples are too small to rank fairly, though it is worth recording that the single NT practice we called still had us on hold at the 15-minute cap, matching those Queensland calls for the longest wait in the study.

The consistency problem

The sharpest pattern in the data is not between states. It is inside the brands.

Grouping the calls by operator, the five independent single-site clinics in the sample were the fastest phones in the country: a median of 15 seconds to a person, and none slower than 90 seconds. At an independent clinic, the person who answers the phone is often the person who takes the booking, and it shows.

The large networks ran medians between two and sixteen times longer, and the spread within a single brand was wider than the spread between states. Inside one national network, one site produced a person in 40 seconds while two sister sites still had the caller on hold at the 15-minute cap. Same brand, same website, same booking promise, and a completely different phone depending on which suburb you dialled from.

That spread is the operational finding of this study. A network’s phone reputation is set by its slowest queue, because the patient who held for 15 minutes does not report the brand average. Consistency across sites is not a branding exercise. On this evidence it is the difference between a one-minute phone and a ten-minute one under the same logo, and it decides where the advertising money each brand spends actually lands.

Credit where it is due

We kept every practice in this study anonymous, with one exception, because praise should be named.

The first direct answer of the study was Castlereagh Imaging Kogarah. A receptionist picked up in 40 seconds, worked through the booking without a menu in sight, and our caller’s note against the entry reads, in full, “AMAZING”.

To check that Friday was not luck, we rang Kogarah again this morning, Monday 24 August, at 10:01am. The same receptionist, Emily Edwards, medical receptionist at Castlereagh Imaging, answered in 25 seconds, faster than Friday, and every bit as polite the second time. We told her so on the call: her practice was among the fastest to answer the phone of the hundred we tested.

The detail worth copying is not only the speed. It is that she answered ready to help. Good is an answer within two rings. Better is what came next: her first move was the booking, not the hold button. A number of practices in this study technically answered within a reasonable time, but with the familiar “Thank you for calling, please hold the line”, which is a human answering a phone while the patient keeps waiting. An answer that opens with hold is a queue with a voice. Kogarah’s opened with help, twice.

One caveat, stated plainly: we called one Castlereagh Imaging site, not the network. This is praise for Kogarah’s phone on two separate mornings, not a finding about every Castlereagh location.

Nothing about that call needs software. That is simply what a well-run phone sounds like, and it sets the bar for everyone else in this dataset. The honest difficulty is that most networks cannot staff an Emily at every site, at every hour, through every Monday surge, and the 15-minute holds elsewhere in this study are what that gap sounds like. Holding the bar Kogarah set, at network scale, around the clock, is exactly the job TriageVoice exists to do.

What this study is not

A study of 100 calls deserves the same honesty we ask of any dataset.

It is one call per practice on one Friday, not a longitudinal average; a practice’s queue at 11:38am says nothing certain about its queue at 8:05am on a Monday, which is usually worse. The practices were drawn from major networks, mid-size groups, hospital departments and independents, but the sample is not a census, and the ACT, Tasmania and NT results are indicative only. Hold times were recorded by a person, not a switch, so treat the odd half-minute as noise. And our caller was a researcher with patience. A patient’s tolerance for minute eleven of hold music is shorter, which means these numbers flatter the real experience.

Treat the study as a direction, not a decimal. The direction is not subtle.

Run the test on your own practice

The method transfers in ten minutes. Ring your own main line at 11:30 on a weekday from a mobile, and time the gap between dialling and a person who could take a booking. If the answer is under a minute, your phones are in the best third of the country. If it is over five minutes, roughly one caller in five is getting the same experience, and they are holding a referral that any competitor can accept.

One more thing. We also rang the same 100 numbers again after hours, and what answers a radiology phone at 7:30 on a Sunday evening is a study of its own. That data is next.

© 2026 Triagents.AI. This study is Triagents first-party data. You are welcome to cite the findings with attribution and a link to this page.