The UK and Australian governments run programs to measure real-world performance of broadband plans. The data is collected from volunteers who install a monitoring device called a Whitebox on their broadband connection. The collected data is then used to produce a report that informs decision makers about the state of the nations internet.
The average speeds are published on an ACCC dashboard.
According to the ACCC dashboard, Australian internet speeds range from 95.4% to 103.5% of the advertised plan speed. However, a 2023 survey of 1600 internet users found that 31% of respondents had issues with their internet provider relating to slow speeds. So who is lying, the government or the people in the survey. The raw data tells the full story.
To get to the bottom of it, we start by reading through the reporting methodologies. The first red flag is ADSL connections are filtered out because they are slow, and the government wants the report to be “future focussed”. That leaves only data from NBN connection types. The slow NBN connections are labelled “Under performing services”, and these are also filtered out. Now the average speed is calculated using only the fast and very faster NBN connections.
The government categories an underperforming service as “services which rarely or never attain plan speed” and claim that underperforming services represented 5.6% of the services tested. However, when you filter the services with a maximum download speed below the advertised plan speed, we can see there’s just over 20% of customers who aren’t getting what they paid for.
Who are these customers? Where are they? and what connection types do they have?
Unsurprising to anyone, its fibre to node connections that have the highest number of under performing connections. The small sample size of fixed wireless (and my own bias from being a happy fixed wireless customer) means that this result probably isn’t indicative of the greater population.
Vodafone, IInet, and ‘Other RSPs’ were the three providers with the highest number of under speed connections. IInet kind of get a pass because 65% of their customers are using fibre to node, the rest of the services they provide have good speeds. Vodafone and ‘Other RSPs’ on the other hand are providing under speed connections across multiple connections types.
The ACT was the worst performing state. However, they also have the highest population density of 171 people per square kilometre. This is 6 times higher than the next state. 47% of ACT connections are fibre to node, so it’s understandable that congestion would be an issue.
Figure 4 makes it clear Fibre to node, and ‘Other RSPs’ are giving Australia’s NBN a bad rep. Who are these ‘Other RSPs’? I assume it’s a collection of smaller services providers, but I couldn’t find any further information on them. All I know is they need to lift their game.
Selecting an internet plan is a significant financial decision. Often once the decision is made, the customer is locked into the service for 12 to 24 months. It is essential that people have the ability to select a plan that suits their own personal needs, and be confident that the plan they purchase will function as advertised. A fast internet is vital to a countries trade and commerce. Everything from currency exchanges to online banking to booking a table at a restaurant relies on being able to send and receive data.
The average speeds calculated in the Australian report include speeds which are far above the plan speed, using the mean connection speed (rather than median) skews the report favourably, but obfuscates problems. The report also excludes both ADSL connections and connections that are performing poorly. which further helps paint a rosy, but inaccurate picture.
To sum up, if you can’t get fibre to the premise or curb, choose fixed wireless or Starlink, never choose fibre to the node. Choose a big service provide that’s not Vodafone, and don’t trust the published figures.