ACCC’s Final Broadband Report: What It Really Says About NBN, Starlink and 5G

In June 2026 the ACCC published the 33rd and final edition of its Measuring Broadband Australia program. The headline — fixed-line NBN connections delivering 99.4% of plan speed in busy hours — is real. So are three traps that make most coverage of the report misleading.

In brief

• The ACCC published its 33rd and final broadband report in June 2026. What it really says about the fastest providers — and the traps hiding in the headlines.
• What the program actually measured: Measuring Broadband Australia was built on hardware, not surveys.
• The provider table trap: there is no meaningful "fastest provider": The most-reported detail of the final report was its provider ranking: Dodo and iPrimus — reported by the ACCC as a single combined result — tied with Leaptel in first…
• Fixed wireless: strong headline, read the plan names: The final report's fixed wireless results are genuinely good news for regional Australia.
• The satellite section: two services, two different rulers: The satellite section of Report 33 is where the most misleading coverage appeared.

After eight years, the Australian Competition and Consumer Commission has stopped publishing its independent broadband speed reports. Report 33, released on 17 June 2026 and based on measurements taken during March 2026 (1–31 March), is the last in the Measuring Broadband Australia program that began in 2017. Its headline finding is that Australian households with fixed-line NBN connections received, on average, 99.4 per cent of their plan’s download speed during the peak evening period. That number is accurate — and it is the least interesting thing in the report. What matters more is what the report says about provider comparisons, satellite services, and what replaces the program now that it has ended.

What the program actually measured

Measuring Broadband Australia was built on hardware, not surveys. Volunteer households hosted a small device called a Whitebox, manufactured by the program’s testing provider SamKnows, which ran speed tests when the household’s connection was not in use. That matters because it removes the two biggest distortions in self-reported speed tests: people mostly test when something feels slow, and a test over Wi-Fi measures the home network as much as the broadband service.

The final report drew on 1,156 fixed-line services, plus separate samples for fixed wireless, very-high-speed services, and satellite. “Busy hours” in the report means 7pm to 11pm on weekdays — the period when networks are under the most load and when the gap between advertised and delivered speeds is widest. Every percentage in this article refers to that definition unless stated otherwise.

The program’s arc is the story of Australian broadband growing up. In the ACCC’s early reports, fixed-line connections averaged between 80 and 85 per cent of plan speed in busy hours. By the final report that figure was 99.4 per cent, up from 98.5 per cent in the previous report (Report 32, released 2 April 2026). The share of underperforming services — defined by the ACCC as a service recording more than 75 per cent of plan speed in fewer than 5 per cent of download tests — fell from 13.9 per cent in May 2018 to 5.6 per cent in the final report.

The ACCC credits part of that improvement to NBN Co’s wholesale changes, including the removal of capacity-based charging on all fixed-line and fixed wireless services. When providers no longer pay per unit of capacity, the incentive to under-provision busy-hour capacity largely disappears.

The provider table trap: there is no meaningful “fastest provider”

The most-reported detail of the final report was its provider ranking: Dodo and iPrimus — reported by the ACCC as a single combined result — tied with Leaptel in first place at 102.3 per cent of plan speed in busy hours, followed by TPG at 101.9 per cent and Exetel at 101.6 per cent. Across the nine reported rows covering ten providers, busy-hour results spanned 95.2 to 102.3 per cent. Treat that ranking as entertainment, not information.

Three reasons. First, the confidence intervals overlap: with samples of this size, the statistical uncertainty around each provider’s average is wider than the gaps between them, so the ordering could shuffle on a different month’s data. Second, providers serve different mixes of plans and connection technologies, so the table compares unlike with unlike. Third, NBN Co overprovisions download speeds — it supplies a slightly higher data rate than the wholesale plan speed to cover protocol overhead — which is why averages above 100 per cent are possible at all, and why a provider with more overprovisioned connections can look “faster” without delivering a better experience.

The ACCC said as much in its own reasoning for ending the program: recent results show far less differentiation between providers’ performance than when the first report was published in 2018. When the gap between best and worst is a few percentage points inside the margin of error, a league table stops being a buying guide. What still matters for choosing a provider is the typical evening speed on the specific plan you are considering, the upload speed, the first-year price, the connection technology at your address, and support — not a headline ranking. Our guide to the best NBN providers weighs those factors instead.

Fixed wireless: strong headline, read the plan names

The final report’s fixed wireless results are genuinely good news for regional Australia. The NBN Fixed Wireless Home Fast plan averaged 160.5 Mbps download and 12.6 Mbps upload in busy hours. The Fixed Wireless Plus plan — the most common fixed wireless plan in the ACCC’s sample — delivered download speeds close to 100 Mbps during peak evening periods.

The caveat is the same one that applies to every fixed wireless comparison: plan names are not speeds. “Home Fast” and “Plus” are product labels, and the wholesale tiers behind them have changed over time as NBN Co upgraded the fixed wireless network. When comparing a fixed wireless quote against a fixed-line or 5G offer, compare the typical evening speed on the specific plan, not the label. Our fixed wireless guide explains which tiers suit which households.

The satellite section: two services, two different rulers

The satellite section of Report 33 is where the most misleading coverage appeared. The report measured 48 Starlink Residential Max services (plus 3 Residential 100 services) and 46 NBN Sky Muster services across several speed tiers. The headline figures, as unpacked from the report’s appendix tables, were:

Measure	Starlink Residential Max	NBN Sky Muster
Average download, all hours	230.4 Mbps	89.0% of plan speed
Average download, busy hours	199.1 Mbps	78.5% of plan speed
Average upload, all hours	47.5 Mbps	119.1% of the ACCC benchmark
Average upload, busy hours	45.5 Mbps	111.7% of the ACCC benchmark
Average latency, all hours	26.2 ms	658.3 ms
Average latency, busy hours	26.8 ms	662.7 ms

Read that table carefully, because it contains a trap. Starlink’s speeds are reported in megabits per second. Sky Muster’s are reported as a percentage of plan speed across several tiers. Those are different units, and they are not directly comparable. You cannot line up “199.1 Mbps” against “78.5%” and conclude anything about which service is faster for a given household, because the Sky Muster percentage spans 25/5, 50/5 and 100/5 Mbps tiers. A Sky Muster service at 78.5 per cent of a 100 Mbps plan delivers a very different experience from one at 78.5 per cent of a 25 Mbps plan.

The upload row contains a second trap. Sky Muster’s upload performance above 100 per cent looks like a mistake; it is not. NBN Co lists Sky Muster wholesale tiers at 25/5, 50/5 and 100/5 Mbps, but the mid- and high-tier services can reach burst upload speeds of up to 10 Mbps in optimal conditions. Because monitored services frequently recorded uploads around 10 Mbps, the ACCC used 10 Mbps as the upload benchmark for those tiers. The above-100 per cent result is an artefact of that benchmark choice, not evidence that Sky Muster uploads beat its own specification.

Latency is the one comparison that is genuinely like-for-like, and it is not close: about 26 ms for Starlink against about 660 ms for Sky Muster, in both all-hours and busy-hours measurements. That gap is physics. Starlink’s satellites orbit roughly 480 to 1,200 kilometres above Earth; Sky Muster’s two satellites sit about 35,786 kilometres above the equator in geostationary orbit, so every packet makes a far longer round trip. Latency does not stop browsing, email or buffered streaming, but it is noticeable on video calls, voice calls, online gaming and any application that needs fast back-and-forth responses. Our Starlink Australia guide covers Starlink’s plans, pricing and hardware in full.

The outage outlier: one service moved the average by 30x

The report’s reliability section is the best illustration of why averages need their distributions alongside them. The ACCC counted outages lasting at least 30 seconds, excluding outages between midnight and 5am:

Reliability measure	Starlink	NBN Sky Muster
Average outages per day	2.05	0.66
Households with fewer than one outage per day	98%	80%
Outages lasting no more than three minutes	91%	64%

The first row, read alone, says Starlink is three times less reliable than Sky Muster. The next two rows say the opposite: 98 per cent of Starlink households had fewer than one outage a day, against 80 per cent for Sky Muster, and 91 per cent of Starlink outages were over within three minutes, against 64 per cent.

The reconciliation is a single outlier. One monitored Starlink service experienced a disproportionate number of outages; when that service is excluded, Starlink’s average falls from 2.05 outages per day to 0.07. The headline average was, in effect, measuring one faulty installation (or one badly obstructed dish) rather than the typical Starlink household. Sky Muster’s lower unadjusted average is real, but the distribution rows show that the typical Starlink household in the sample had fewer, shorter outages.

This is the pattern to watch for in every broadband statistic: a single average can be dragged far from the typical experience by a small number of bad cases. Where a report publishes the distribution, read it. Where it does not, treat the average with suspicion.

What replaces the program

Report 33 was the final publication in the program, which concluded at the end of June 2026. The ACCC says it will keep monitoring broadband performance using data collected from service providers under its telecommunications record-keeping rules, including the NBN service quality and network performance rule, and that it will keep taking enforcement action where it sees evidence of misleading speed claims.

That is not a like-for-like replacement, and it is worth being clear about what is lost. Record-keeping data comes from the providers themselves rather than from independent hardware in volunteers’ homes, and it is collected for regulatory oversight rather than published as consumer-facing comparisons. The practical consequence is that there will be no new independent provider league table, and no new independent satellite comparison, unless someone else funds one. The March 2026 snapshot in Report 33 is therefore likely to remain the freshest independent benchmark for some time — another reason to read it carefully rather than skim the headline.

The ACCC’s enforcement role matters more than ever in that gap. Providers still have a responsibility to deliver the speeds they advertise, and the ACCC has said it will continue to act on misleading claims. For shoppers, the defence is the same as it always was: check the typical evening speed on the specific plan, not the headline maximum; check the upload speed if you work from home; and test your own service over Ethernet before deciding it is the provider’s fault.

What this means for you

• Do not pick a provider from the Report 33 league table. The gaps between providers sit inside the statistical margin of error, and the ACCC itself ended the program partly because providers’ performance had converged.
• Compare speeds in the same units. Starlink’s Mbps figures and Sky Muster’s percentages of plan speed answer different questions. For Sky Muster, ask the retailer for the typical evening speed on the specific tier you would buy.
• Latency is the real satellite divider. About 26 ms for Starlink versus about 660 ms for Sky Muster is a difference you will feel on calls and games, whatever the download figures say.
• Read distributions, not just averages. One outlier service moved Starlink’s outage average from 2.05 to 0.07 per day. The distribution rows — 98% of households under one outage a day — describe the typical experience better than the average.
• Budget for the monitoring gap. With no new independent reports coming, treat Report 33 (March 2026 data) as the last independent benchmark, and weigh provider-published typical evening speeds and your own tests more heavily.
• Putting the report to work: our Telstra vs Optus vs TPG comparison applies these findings to the three biggest providers.

Sources

• ACCC — Measuring Broadband Australia, final performance report (primary source)
• ACCC media release — households receive consistent broadband speeds