Last month my service was better than 18% of others, but it seems it just gets better and better! 
From the latest Honesty Box report
Your Internet is better than 14.8% of other Australians on nbn⢠SATELLITE(LTS) 25/5
3 Likes
We could suggest putting a third satellite into the system, a 50% capacity increase, which might see you back up from 14.8% to being better than 22.2%! Either result is still 100% better than our NBN, however our ADSL2 for as long as it remains is likely better than either option!
Unlikely, although the government has completed an inhouse upgrade, with the Fed Minister for Communications, Cyber Safety and the Arts (previously Mitch Fifield) being superseded by Paul Fletcher.
It is not yet clear if the upgrade is intended to improve the customer performance of the NBN or facilitate an improved market valuation for the sale of the NBN/NBN Co?
The Hon Paul Fletcher has a background suited to selling IT infrastructure, as well as prior experience in a governance role with Optus. There is even a Telstra connection and a published book âThe wired Brown Landâ. All are well supported by First Class Honours degrees in Economics and Law.
For anyone in regional Australia wondering about where to take their NBN performance data for a second opinion there is also the Minister for Regional Communications, The Honourable Bridget McKenzie. As a senior member of the Nats, perhaps there is some hope for a better solution in the regions, although the federal electoral results in regional Qld and elsewhere suggest the NBN is not a deal breaker!
2 Likes
A month ago I asked the ACCC some questions about broadband speed reporting, here they are:
Regarding the MBA assessment of broadband internet speed.
Why is fixed wireless and satellite not measured?
What evidence do you have that the sampling of performance done by white boxes is an accurate statistical representation of the whole population of those services?
Since the connections sampled are volunteers this necessarily gives a different sample compared to a random selection. How did you assess and deal with the bias introduced by this factor, and any other factor, leading to a non random sample?
Have the statistics published been adjusted or corrected in any way other than the stated removal of underperforming services?
I have just got their reply:
When the MBA program was originally developed, most of the international experience in monitoring broadband performance was limited to fixed-line services and we accordingly limited the scope of the MBA program to these connections. There has since been a greater uptake of fixed wireless and satellite services, and we are working with the Department of Communications and the Arts and the Australian Communications and Media Authority to explore options for monitoring broadband performance on satellite and fixed wireless networks without impacting consumersâ quality of service.
The MBA program gives consumers independent information on which to base their fixed-line broadband purchasing decisions and promotes effective competition on the basis of service performance between Retail Service Providers (RSPs). It is a snapshot of performance during a limited monitoring period, and encourages RSPs to maintain and review their entire networks as they do not know the locations of Whiteboxes. Since Report 2, download and upload speed results have been presented with 95 per cent confidence intervals to indicate where the true average is likely to be. We have also been able to report on the download and upload speeds by access technology since Report 4, and these speeds are also presented with the 95 per cent confidence intervals.
While we have had a high number of prospective volunteers, we prioritise those that allow us to increase the scope of the MBA reporting across RSPs, speed tiers, access technologies and locations. We have been able to report on eight RSP groups since Report 4, and we continue to seek volunteers who have purchased lower speed plans or services from smaller RSPs.
SamKnows was appointed to measure the quality of fixed-line broadband services and produce quarterly reports on these measurements. SamKnows exclude any data generated by failed tests or where results are significantly inconsistent, as applied in the outages results presented in Report 5 (p.7). SamKnows also considers whether insufficient data has been collected and if monitored services have changed RSPs or speed plans during the monitoring period as part of their data preparation process.
Regards
blank
Measuring Broadband Australia
The replies to Q1 and Q4 are sensible with one exception. From the whitebox point of view no data can be collected during outages or failures. From the consumer point of view these failures represent a loss of service. Thus the sample is immediately biassed as it is not of the whole performance of the system 24/7 but only of the system while it is working.
The replies to my questions about bias in the samples are revealing.
Regarding Q2 their para 2 does not really address the question. Having a sufficient sample to provide 95% confidence limits does not mean the sample is unbiased, it just means it was big enough to specify the limits of accuracy due to measuring a sub-set rather than measuring the whole population. Put another way the analysis they have performed says they are 95% sure the reported statistics of the sampled group falls within the specified range, thay have no idea if the sampled group represents the whole population.
In reply to Q3 their para 3 actually says the sample is not random. They had more volunteers than they could measure and reduced their numbers in a way to give them representation in different categories that they were interested in. This most probably has the effect that smaller categories are overrepresented and larger categories are underrepresented. The respondent does not approach the issue of assessing the bias introduced by either the volunteering by respondents or by the ACCC deliberate selection of participants from that group of volunteers, so (unless there is lying by omission) my conclusion is that such bias was not assessed.
My overall conclusion is that the figures published are not reliable and under this methodology can never be a reliable gauge of what is happening across the system. We have no idea at all whether the data points are drawn more from those with particularly good performance ,or bad, or neither.
Since the biased nature of the sampling is not mentioned in any of the publicity the ACCC are misrepresenting their measuring program. Whether they know this or not is speculation. In their defence the results that we have may be the best that they can do under the circumstances but if that is the case they ought to say so rather than allow everybody to assume levels of accuracy that simply are not there.
5 Likes
Or you could wait until Elon Musk has finished launching his new nebula of TWELVE THOUSAND low earth orbit Internet-enabled satellites. (The first sixty resulted in widespread UFO reports, along with some complaints from astronomers who would like to be able to see the sky.)
I would suggest that youâre asking the wrong people. While the ACCC is responsible for the monitoring, it has no idea about statistics and so doesnât even understand how badly it has messed up. That is clear from the response you received, which would have gone through several staff before being finalised and sent.
If you asked the ABS to evaluate the program and give a confidence interval it would likely be very different from the claimed 95%.
1 Like
How do you know this?
How do you get another agency to look into the ACCC and their statistical competence? You should try it and ask the ABS, tell us what reply you get.
Which bit - that the ACCC doesnât do statistics very well or that it is responsible for the monitoring?
- The ACCC is responsible for providing consumer outcomes. It is not responsible for producing statistics, and while it does produce some (such as the number of cases it deals with) these are generally low-complexity amateur stuff. Statistical analysis can get very complicated very fast - I know, because I donât do it!
- The ACCC itself acknowledges its responsibility for monitoring NBN performance.
Hypotheticals are hypotheticals.
It would be better handled by the Federal Auditor-Generalâs office, the Australian National Audit Office (ANAO), to either support or find lacking the ACCCâs Broadband monitoring methods and outcomes.
I think they only take requests only from members of both Houses as to what they should audit and not from the general public. I do know that they accept public submissions to any current audits they are undertaking, so if one was started by them about the ACCCâs program the concerns raised here could be submitted to that audit. A person can subscribe to email alerts so they can be informed about audits and reports at the following webpage
https://us9.list-manage.com/subscribe?u=81413a7f32ad4e90da9c0ce37&id=b7cf473752
For any request we as members of the public might have we would first have to convince either a Senator or Member of the House of Representatives to submit a request, then ANAO would have to agree with that request and then it would have to be given a priority and so if accepted it may take some time before it was undertaken.
2 Likes
Count me as one of them, Iâve seen some deep sky photos ruined by them, and there are only 60 up there so far!
4 Likes
It was âit has no idea about statisticsâ that I referred to. Just because stats isnât their primary business doesnât mean they are incompetent or that they should be allowed to be incompetent All government agencies are required to produce stats about their responsibilities. They should all produce them professionally so that the public are well informed.
How do you come to the conclusion that the ABS would compute a different confidence interval? Confidence intervals are a result of a calculation that crawls out of the raw figures. The ABS could very easily come up with the same one but that obscures the problem.
Picking a figure out of the air to illustrate; if the mean download speed on a 25 plan is 20 Âą 2 it means we are 95% sure the real figure for the sample is between 18 and 22.
The problem isnât the confidence interval or the computational accuracy it is that the sampling is biassed because it is not a random selection of the total population. It is non-random because:
- participants volunteered, they were not selected randomly and
- of the volunteers a subset was chosen by the ACCC according to criteria of their devising.
So in the above example we can be fairly sure the mean download of 20 is representative of the connections measured, what we donât know is how the connections measure represent the whole population.
1 Like
The latter leads to the former. A confidence interval is only as meaningful as the data that is fed in, and a competent statistician would see and allow for the statistical bias. It is not possible to produce a meaningful confidence interval when your underlying data is biased (although you could produce an interval with the caveat that the underlying data is biased, I suppose - this is presumably what you mean).
I have produced government statistics, and am not a statistician. What I produced was for a small agency that no longer exists, and involved very simple numbers that most people with an understanding of basic mathematics could reproduce.
Most government agencies have a lot of experts in their core business, a few experts in managing the business, and no experts in areas that are not their focus. Hence the Treasury has economists, accountants and statisticians by the score - but the Department of the Environment has a lot of scientists, a few accountants and economists, but probably out-sources any need for detailed statistical analysis where its own scientists do not have the requisite skills. These are entities that operate within constrained budgets that have been cut by around 33% since 1987 simply by the blunt hammer of the âefficiencyâ dividend - they canât afford to have full-time staff working on issues that are outside their area of expertise and responsibility.
1 Like
This conversation is getting too academic - most readers have drifted off already.
I just wanted to show that the publications sound much more reliable than they are. The methodology or the descriptions of the results (or both) need work for the results published to actually mean what they purport to mean. So when some poli gets up in the house and says the NBN performance is thus and so we actually have no idea if that is true or not.
âThere are three kinds of lies: lies, damned lies, and statistics.â [unknown]
1 Like
My ISP advised last week that I am able to connect to the NBN (FTTN), and so Iâm planning to do a little stroll through the neighbourhood this afternoon to see if I can locate said node and get some idea of the distance between it and my home.
It appears that node location was once freely published as part of Commonwealth mapping data, but has since been taken down. Itâs a secret!
Iâm not sure how accurate it is, but the Finder website has a map that seems to indicate where the nodes are located. My first stop in this afternoonâs drive will be where that map suggests a node. Nope, no node there. Couldnât see one anywhere I went 
.
4 Likes
Of course it is a secret - knowledge is power, itâs just one way the blunt end of Government is depowering the masses 
or so I am led to believe by the voices in my head ⌠
5 Likes
Without starting another advanced class in statistics, @postulative steps to locate his FTTP node deserves consideration?
Not all NBN connections are created equal. The ACCC reports may suggest to some that FTTN performs just as well as FTTP and HFC (CVC issues excepted)? Educated users know otherwise!
FTTN is the most common connection type. The available service speed depends on distance and copper condition. The NBN performance results do not relate the speed data collected accordingly.
Statistically the data analysis mess side steps the bias this might introduce. Within tiers both line condition and CVC have effects on performance. And the question of whether the sample points selected (distance and copper condition variable) are truely representative remains untestable?
Notes:
Perhaps FTTN should be charged the same way that ADSL is? One fee, speed not charged by tiers. All users are treated equal paying for total data consumed (variable) and the line connection (common fixed cost across every user)? The ACCC results would take on a whole new meaning in that scenario!
All the discussion on the published NBN performance statistics has been insightful, but more worthy of a 201 level class than a 101, or basic high school maths. Does the degree of ignorance raise this is to 301. String theory appears simpler to explain.
If node location is now a state security secret, with some possibility of seeing reporters raided and whistle blowers jailed, is there anything else that might help? At least for now the NBN wireless tower locations are available on line. Those in our area even have published local council development applications. If FW users can know where and how far away a tower is (determining as a result likely speed tiers and performance outcomes) surely knowing where your local FTTN node is and copper distance is no greater a risk to state security?
Certainly âDial Before You Digâ will know the locations. Although a good guess is any one with a node nearby the front gate will not need a permit to find a node location. It would seem irresponsible for DBYD to not provide this advice, or to require the recipients of their services to sign onto a state secrecy agreement.
3 Likes
It is possibly bundled into copper condition but another angle Iâve recently become aware of is line power (the figures quoted as dBm). My area was RFS in April 2018 and weâve had the service since then via Telstra. In the first 12 months line sync was usually 50 - 60 down, 20 - 25 up. The Telstra plan has some statement of expected speed like 45 down and 17up which has delivered in that time.
Over the last couple of weeks I noticed download speed was down and upon checking modem found sync was 11 down and 19 up. Exploring in another forum as to what the issue might be it was identified that at that time the line power was running around 7 dBm down when it should be 14. It was mention that this is can be done if there is interference with other services. So I donât expect I am an orphan in this situation but it appears down the track as more customers migrate to FTTN these issues crop up and the service is degraded. So far up to 4 levels in the Telstra help system and still getting âitâs your distance to the nodeâ or is because âmany people are streaming lots of dataâ causing the slow down. Waiting for a âcase managerâ to call me, hopefully that is up far enough to speak to someone that is more savy.
5 Likes
Just musingâŚBased on your scenario it could be so simple as they changed your port profile because they could, or possibly because that is your plan, which you did not mention but implied it was 50/20 per the first 12 months speeds. Could it be that you inadvertently bought a plan that was âupgradedâ to 50/20 for a year and they dropped it back to 12/1 at the end of the original offer?
Going a bit off your problem, Iâll reference my old post for something probably unrelated but that could be helpful (or not) to others.
Some FTTN modem-routers might also have an SNR setting. Check the manual. If you are not getting the speeds you expect experiment with it. It might not be documented but could be worth a google search to see if it has been âdiscoveredâ or is not supported as a configurable setting in the product.
Also, it is my understanding that ADSL/VDSL copper ports have profiles at the ISP/RSP end that are configured for stability or to a policy. Sometimes a friendly tech can be persuaded to set a particular line to âperformanceâ but they donât easily do it because their end will also run hotter, use more power, and they have their directives. All that is secondary to the plan, and that will limit the max regardless of whatever else is âin the wayâ.
3 Likes
Look for a pillar first as the node shouldnât be too far away from that. Some nodes are placed behind hedges and similar types of view obstructions and some are quite some distance from where you might expect them to be. The two for areas where we had been were about 3 streets away but the copper run was fairly short allowing 100 Mbps plans to the current residents of those houses.
The mapping done privately at http://nbnmtm.australiaeast.cloudapp.azure.com/nbnmtm.html linked by @peterjvwatt in the âWith NBN, distance from the Node determines speedâ topic while it may not be being updated might be worth checking just in case.
I think the normal setting they employ is âStableâ but âStandardâ could also elicit improvements and may be an acceptable alternative to âPerformanceâ.
5 Likes
No anticipated changes in contract with RSP, in middle of 24 month arrangement, yes would have been 50/20 but at the time it was struck people were complaining about congestion and Telstra quoted speeds may drop to 45/17 in peak. However as I understand it the modem sync is independent of âthe planâ. Modem will sync at what ever speed it can, if it is above the plan it will still sync there, if it is below well that is the maximum that can be achieved.
Interesting on Saturday afternoon I was working at PC and internet dropped out, so did local network and looking at modem/router lights it was rebooting. When it came back on line sync speeds were very different at 35/13. Line down power was up from 7dBm to 12 (but not 14 that others have said is usual). All I can assume from that is the modem was rebooted remotely after techs had changed settings their end which I didnât know was part of the system.
I donât know if what I am experiencing is going on where customers are stealthily degraded because NBN infrastructure is not up to it or RSP trying to cut costs. On my first contact with Telstra the rep said âdonât worry we will definitely refund your accountâ at the start like they know but donât want to admit, I hadnât mention refund I just wanted the service I had and was contracted to receive.
5 Likes
Yes Sync Speed will be the maximum speed your line can handle for that connection, but it can change on a reboot as the speed is negotiated between the home modem and the VDSL modem in the node depending on levels of interference and packet loss at the time of the negotiation. Normally it will only be renegotiated on a reboot or reset. Hence if you have a poor link speed a reboot may improve your outcome for that new session.
If you have had wet weather your link speed may also be degraded due to water infiltration of any joins and cracks in insulation of the wires. This can persist for some time after the wet weather has ceased. Checking your SNR can be useful in determining if you have line issues as the SNR will decrease if there is line interference.
5 Likes
A resource that may help is here
https://forums.whirlpool.net.au/forum/142
There is likely a thread dedicated to your locale and can advise locations of those secret cabinets.
5 Likes
