Nation shocked that countries with more full fibre or faster cable beat UK in speed tables

The comparison site[1] has taken the open M-Labs speed test data and has done some number crunching to create a league table on where the UK stands and its stoking the bonfire of bad broadband headlines.



Mean Download Speed

Time to download 7.5GB file/movie

Singapore 1 55.1 Mbps 18m 34 sec
Sweden 2 40.2 Mbps 25m 30 sec
Taiwan 3 34.4 Mbps 29m 46 sec
Denmark 4 33.5 Mbps 30m 32 sec
Netherlands 5 33.5 Mbps 30m 33 sec
Japan 12 24.5 Mbps 41m 51 sec
United States 21 20 Mbps 51m 13 sec
Spain 13 19.6 Mbps 52m 15 sec
Germany 24 18.8 Mbps 54m 28 sec
United Kingdom 31 16.5 Mbps 1hr 2 m 1 sec

So first question is why is the UK average 16.5 Mbps based on the M-Labs data from May 2016 to May 2017, when we report a crowdsourced mean observed speed test of 27.4 Mbps (Q2 2017 – median is 18.9 Mbps) and the Ofcom model says 36.2 Mbps (Nov 2016 data – limited sample size with statistical modelling to create UK average). If you have no idea where you stand on your broadband speeds our speed test[2] will give you an analysis of how you compare to others, simply run the test and click the green analysis button at the end of the test to see how you compare with others in your region of the UK and also how your speeds compare for others on the same type of connection technology. You have to understand the testing methodology and various sample sizes to get a better idea, the M-Labs testing is single thread based and thus is more sensitive to congestion and Wi-Fi connectivity, hence why in July 2017 our main multiple thread mean average for Virgin Media was 67.7 Mbps but a lowly 41 Mbps for single thread, over at BT Consumer this was 30.1 Mbps versus 27.6 Mbps.

This means anyone trying to say the Ofcom results are wrong is comparing two very different measures, the single thread testing is generally good for giving an idea of how streaming video performs, but for download and subsequent playback of any file or video the multiple thread test results are more relevant. The Ofcom model does have some issues in that it is based on just 1 month’s data every year and the sample sizes while carefully selected to give what is said to be a good model for distance limited technologies always carries the risk that some variations due to regional/local congestion are missed or under represented. The choice of a 7.5GB movie to help convey the various speeds is interesting, as at standard movie lengths of 90 minutes to 2 hours this gives a bit rate of 8 to 11 Mbps, but the file size is apparently chosen to reflect the downloads from stores such as Xbox/Playstation which can have higher bit rates than streamed files and also reflecting longer films such as Lord of The Rings which push towards the 3 hour boundary.

Of course most store downloads will let you watch as soon as a proportion has downloaded so you will not usually have to wait the full time to download the file before you can watch, but as a way of giving people a metric they can understand rather than “Megabits per second” which is often mis-quoted as “MegaBytes per second” (8 bits form a Byte when talking broadband speeds) it’s understandable. Our thinkbroadband Q2 2017 speed test results actually recorded a single thread mean download speed of 25.1 Mbps and a median of 17.8 Mbps. What is interesting is that our median result comes out so close to what is describing as the mean, which suggests they may either be seeing a higher proportion of low speeds or lacking some of the faster users.

On the sample sizes the number of unique IP addresses in the M-Labs data was 354,329, but countries such as Germany which are larger was recording a much lower figure of 141,699, and what we don’t know is the profile of users i.e. are the ratios of users at different providers what we would expect based on customer numbers something we keep a close eye on and is one of our alarms for when a provider has issues e.g. higher or much lower volumes of tests than we’d expect normally. The Ofcom model is based on many tests from just some 2,500 locations. So as a UK mean we believe the M-Labs analysis from is low, but this is down to the single-threaded nature and those running the test may not reflect the full gamut of the UK consumer broadband market, or put another way if we were in Germany we would not be publishing a national average based on 114,572 IP addresses without significant modelling work and similar with some other countries.

Reliable international comparisons on broadband speeds are difficult to do and with roll-outs meaning the picture is constantly changing by the time rigorous analysis has been done the reality on the ground has moved on. The countries that are above the UK generally have much faster packages from cable broadband providers such as Liberty Global (current owners of Virgin Media) or also have a high proportion of full fibre (FTTP or FTTB). What is interesting though is even that the average is not massively higher in countries where ultrafast services are much more widespread, and that is where the debate needs to be i.e. in countries with full fibre services are people buying the bare minimum package that means their needs and budget or are full Gigabit packages selling like hotcakes?

If the latter, then the UK is very quickly going to lose its digital economy gloss. We specialise in the UK broadband market and as such our Q2 2017 analysis as the postcode area level highlights the variations across the UK and the fastest areas are unsurprisingly those that have had good superfast and ultrafast coverage for some time, i.e. long enough to allow take-up to make an impact on the average speeds. The slowest postcode areas out of the several thousand there are across the UK don’t make for pretty viewing and the DG3 area is particularly slow; with a superfast coverage level of around 57% this should be no surprise and as the rural roll-outs in Scotland are still well underway.

The lag between better speeds being made available and people upgrading are a big factor. As with the BIG report we have seen some coverage use the low average speeds to talk about problems with the broadband USO and it appears that the only way the broadband USO will make some happy is if upgrades are automatic rather than on-demand and that people will pay no more than they do now. The unfortunate truth though is no matter who delivers the USO, it involves work and new hardware both costing money and someone somewhere has to pay for it, but it will help to safeguard better affordable services and provide a safety net for social inclusion.

As a society we need to realise that free lunches are never free. Update 12:40pm Just to add a note as to why we are using the median speed in our postcode area comparison, as we are dealing with clearly smaller areas and thus to avoid the pressence of a small number of ultrafast users skewing the results we feel that median is more representative of the picture in the postcode areas. The top/bottom 25% 20% 10% figures would help to make it clearer what the speed profile is for those areas, but that would swamp people with figures and similar reasons as to why we’ve not featured upload speeds.

Any media outlets wanting more in depth information such as the upload speeds or ranges can use the contact details on our press page[3].


Post a comment

Login Register[4][5]


  1. ^ (
  2. ^ our speed test (
  3. ^ press page (
  4. ^ Login (
  5. ^ Register (

Leave a Reply

Your email address will not be published. Required fields are marked *

This site uses Akismet to reduce spam. Learn how your comment data is processed.