I am not beating then drum for a a 100% survey or ‘complete’ accuracy. We don’t need total accuracy, an adequate sample would do. It needs to actually represent the thing to be measured (NBN users) and not some other population, to be unbiased (not an unrepresentative subset) and of sufficient size to give acceptable accuracy. The ABS does this kind of thing all the time it isn’t a new and arcane technology. It is very common in industry to trade off sample size (ie survey cost) with acceptable levels of accuracy.
The problem we have is the metrics available are misrepresented as measuring a specific thing when they do not, the problem is not that the sample size is too small.
@PhilT "His solution was an approximation not a solution, "
This is not a very useful case study as numerical methods are commonly used to solve differential equations because some classes of DEs have no known analytic solution. We were not in search of an analytic solution or an exact measurement but an adequate sample. There is nothing wrong with approximations in either solving differential equations or statistical sampling if they are sufficiently accurate for the purpose. There is something wrong with misrepresentation of the meaning of a statistic.
Many discussions go on and on on what stats mean or don’t mean, but my point was not methods, but to support @grahroll’s assertion about uncertainty. Sometimes one has to take the available information assuming it meets the pub test and run with it. Even if not precise it often is close enough for good policy or good planning.
Wilfully misrepresenting it? I’ll not go there since we seem to be up against masters of the craft in the political sphere, and those representing agencies and businesses government owns.
The sample size for the ACCC program measuring BB speeds is fairly small. I don’t know the size of the Honesty Box sample size but would imagine it to is rather small perhaps several 1,000 boxes. These would be the only programs where pure nbn™ BB metrics could be obtained to generate some stats that refer only to nbn™ speeds. Based on those sample sizes for millions of users how accurate would that data be.
The Ookla™ results offer the total BB picture for Australia (rather than a specific section of that BB market eg ADSL or ADSL2+ or nbn™ connections) ie all forms of connecting to the internet whether it be ADSL, FTTC or so on. It does the same for every Country it provides the rankings, and other information, on. This is probably the best of the choices they are able to make. It also is a better marker of where we stand in the World than cherry picking one particular aspect of how we connect in isolation from all the other ways.
Given also that the nbn™ is such a mishmash of technology how do we then pick which elements of that should be the ones we measure and which we should disregard (or should we even be picking or disregarding sections, I don’t think we should). The new scheme that the Govt wants/seems to use takes out from what I understand any FW or Satellite data so the ranking is purely on actual wired (cable, copper, fibre) nbn™ connections. This selection would be/is then just a biased picture of our ranking in the World, and disregards all those others who are connected by other means including those who use dark fibre and other networks outside the nbn™ and who only connect via the fibre backbone.
This is still the case, in a strict sense. Computers can add natively, but that’s all. Subtraction is negative addition, and after that things are past my pay grade. (Feel free to correct me, old-timers.)
The sample size needed for a close approximation can be a lot smaller than you may think seems logical. I am no expert, so will refer you to Wikipedia to enjoy the light reading.
Oh, and of course the Birthday Paradox is a great demonstration of sampling.
Historically some processors had machine level divide units and others left it to reciprocal units. Numerical analysts once upon a time lambasted CDC/Cray accuracy because of its unique lack of a hardware divide to operate faster. FWIW CPUs are also pretty good at multiply/shifting and courtesy of the NSA from w.a.y back, implementing popcounts.
Anyone wanting to delve further into CPU instructions should start with IEEE-754 that caused standardisation, and then seek out specific hardware manuals for the processor of choice.
This is going w.a.y off topic and well beyond consumer issues so I’ll leave it with ‘it depends how strictly you define strictly’.
HonestyBox gives a reasonable report each month but nothing real-time that I’ve seen. The report is a little granular in its graphical rendering without offering drill-down (lots of dot points but unclear of specific values). SamKnows has a very flash(looking) dashboard that displays data over selected time-frames (day, week, two-weeks, month) as well as an instant test feature.
I’m curious to see what the actual testing is and to that end I intend running both on a mirror port and wiresharking the traffic to get a good idea of what they are both hitting. I know some of it from looking at the pi-hole for dns lookups and since the border router is intercepting dns for non-vpn traffic it should be accurate at least for lookups
October 2019 Speedtest Results for Broadband speeds sees Australia on
Fixed at 64th (slipped 3 places from September 2019) Download 42.01 Mbps Upload 19.75 Mbps
Mobile at 4th (gained 1 place from September) Download 67.19 Mbps Upload 16.74 Mbps
This is a result across all Broadband connection types including ADSL, FTTP, HFC, Satellite, Fixed Wireless. So this is not just nbn™ testing and should be read with that in mind. Still a very poor outcome for the dollars spent so far.
In the scheme of things very much splitting hairs over the few percentage point differences between different providers.
It’s interesting to note that ADSL customers see no slow down when compared to the fixed line NBN services during the busy hour. And for overall web browsing the NBN is only twice as fast to load a web page compared to ADSL. On its own not a big incentive to upgrade. Interesting in that there is something else significant that is holding back everyday browsing and Internet performance for NBN users.
Note:
Should we note the lack of reporting of the Fixed Wireless service performance and Satellite Services yet again? Or does their absence ask us to accept that by the ACCC they are not fast broadband or even part of the NBN? There is only one reason these are not reported on each month. Honesty is a hard task master. The NBN remains committed to only achieving a mythical minimum average of 6Mbps download for Fixed Wireless. Although many may see better. Those only seeing low speeds or 6Mbps or worse in the busy hour need to know why they have been singled out when other FW customers get close to full speed? Satellite is equally worse for all, perhaps?
I agree to disagree with their assessment. The hours I use for my online games is in those busy/laggy periods and thus a pita to play. So my online gaming is suffering from an enforced hiatus.
Interestingly, while my NBN connection is pretty rubbish I don’t generally have too much of a problem with my gaming. That said, I don’t play any serious action games - just online action-RPGs like Diablo and Path of Exile.
Lag time for online FPS is critical eg Overwatch. You die fast while waiting for data to catch up next thing have to recover and return…gets old real fast.
The latest. Note they remove or move things from time to time, so get a copy while it works. ‘MBA’ might explain some parts, noting many MBA’s are notable arm wavers who cannot achieve much of anything beyond depositing their handsome pay.
For those unable to see this post in another thread:
As others participating in the Choice-Honesty Box internet measurement program will no doubt also know by now, the program is ending at the end of July. It appears funding is the issue.
It’s been raining for most of the past 3 days here, so I’ve been relying on my phone data for internet, as nbn satellite slows then fails when rainfall intensity increases, so my final Honesty Box report may show some deterioration over this period. However, since it is only ever better than 14-18% of others, it wont make a whole lot of difference!
I hope the program has provided useful insights for those investigating their nbn options, and that Choice has obtained useful data.
It’s been a while since I looked at this issue so I checked the NBN web page to see if there had been any advance. This article is still the latest word from NBN. This was mentioned in this thread back in October when it came out. The report is quite glowing about the wonders of the NBN and puts the slipper into OOKLA etc.
For those who missed it at the time, the Alphabeta report is not based on any kind of measurement, it does not use a hardware box nor a software test using the user’s hardware. It is far superior to all that. It uses … wait for it:
the subscription speed.
Yup. It assumes that the nominal speed of your plan is a fair and accurate measurement of the speed you actually get. I bet we the taxpayers paid a lot of money for that.
So in answer to the question posed by the title of the thread the answer was no when it started in 2017 and it is still no.
Are we suggesting there is a gap between what we know that we need to know and what the NBN Co believes we need to know? This is from June 2020 (this year) with some expanded detail available If you open up the more details.
According to what the NBN needs is to know it’s almost perfect?
Just to be sure I wasn’t dreaming, some other sources of informed opinion provided an alternate view point.
Figures from the consumer watchdog’s Measuring Broadband Australia report show that as many as 20 per cent of NBN users with FTTN connections aren’t getting the speeds they are paying for.
“One in five consumers on these connections are still paying for high-speed 50Mbps and 100Mbps plans that are underperforming,’’ ACCC boss Rod Sims said last month.
Not surprised?
In respect of the now class leading high speed 500-1,000Mbps aka 1Gbps high speed NBN plans. Only 18 per cent of NBN connections have been deemed capable of delivering the speeds,
I was briefly concerned the NBN Co might simply have a different vision of Australia’s future internet needs.
‘Finished’?
There are at least two ways to interpret so many words in English. Kaputt immediately came to mind.
Paul Budde’s synopsis was more comprehensive. Approx $16B was suggested by one more knowledgeable to be the starting cost of putting most of what is wrong right!
