Skip Navigation

Ofcom calculates UK average speed as 17.8 Mbps
Tuesday 15 April 2014 11:53:56 by Andrew Ferguson

The speed landscape of the UK is a complicated one with many parties doing their best to ensure their vision is the promoted one, and it this reason why Ofcom spends so much time analysing broadband speeds and has for some years produced a set of statistics to try and inform policy makers as to how the UK is doing. The latest round of testing by Ofcom using a panel of 985 hardware testers gives a UK average speed of 17.8 Mbps, a rise of 3.1 Mbps compared to May 2013.

The growth in the average speed is not because ADSL2+ has become faster, but that more people are buying fibre based connections via the wholesale Openreach service or Virgin Media. Our own estimate for the calculated UK average broadband speed is 23.7 Mbps and this is based on the financial figures to the end of 31st December 2013 and this takes into account the distance limitations of xDSL technologies, and is very close to a similar calculation by Ofcom which uses different statistical analysis to their main average to suggest a UK speed of 25.3 Mbps.

Ofcom comparing rural and urban broadband speeds. Caution urged on size of rural sample
Click image for larger version

We know the urban/rural digital divide will be upmost in many peoples thinking and Ofcom seems to confirm the existence of the divide with the headline that urban FTTx (Openreach and Virgin Media) has an average speed of 46.3 Mbps and rural 29 Mbps. Alas we must urge caution as only 17 rural fibre based connections were used to draw this conclusion which is not enough location samples to get a full picture of how VDSL2 and Virgin Media perform in rural locations.

While software based speed testing is not used by Ofcom, our own tester at thinkbroadband has the advantage of having a lot more geographic data points and our last set of data in January 2014 shows the postcode lottery that takes place. Places like Crawley, Luton, Nottingham and Stevenage all show more than 40% of tests on our broadband speed checker are at superfast (30 Mbps or more) speeds, while at the other end of the scale Conwy, Purbeck, Derbyshire and Forest of Dean all have less than 1 in 40 tests at superfast speeds. As with all our speed test analysis we average out individual results to avoid one person skewing an area with thousands of tests.

If you want to play with various scenarios for where the UK may be in the future, this calculator will let you play with take-up and coverage figures.


Posted by johnpopham over 3 years ago
How can a sample of 17 be statistically valid?
Posted by WWWombat over 3 years ago
My biggest reaction was to the size of the panel: 985 testers is less than half that were used 6 months ago. Had the increase in bandwidth usage killed off so many participants?

The summary page at Ofcom is bad - as it uses very different panel sizes every time it chooses to mention it. The 985 seems to refer to the upload tests alone.

However, in the main PDF, Ofcom state that they are using 2,391 panellists.

As for the 17 rural users, the report is careful to mention that those results need to be treated with caution. They know it isn't yet statistically valid.
Posted by herdwick over 3 years ago
It also shows 20 - 30 Mbits/s for the rural fibre so the range isn't hiding behind an average.
Posted by AndrueC over 3 years ago
"Had the increase in bandwidth usage killed off so many participants"

It certainly stopped me contributing. I unplugged the box and stuffed it on the top of a cupboard to gather dust.
Posted by TheEulerID over 3 years ago
Why is it that almost everybody insists on reporting arithmetic means when it's wholly inappropriate for characterising skewed distributions like download speeds? It's well know that a few high results gives a wholly misleading impression, which is why income is normally reported using medians, quartiles, percentiles and so on.

Please Ofcom (and others), read some books on statistics.
Posted by New_Londoner over 3 years ago
See page 58 onwards in the Ofcom report for details of the methodology, panel size, statistical analysis etc.
Posted by gerarda over 3 years ago
it tells you it uses mean so as TheEulerID says - its wholly inappropriate.
Posted by themanstan over 3 years ago
not really inappropriate in this case, as it is the equivalent of per capita income... which is a mean calculation.
Posted by TheEulerID over 3 years ago

It is the equivalent of per capita income, which is exactly the point. The mean is a dreadful way of summarising the statistical character of skewed curves, which is why economists tend towards the median when they want to represent the central tendency of the curve. Avg per capita income misrepresents the typical experience for the same reason avg download speed misrepresents.
Posted by themanstan over 3 years ago
but they still report pci... because it is simple and representative of the "gross" market and not of any particular segment.

as soon as you look at more complicated analyses then get a real statistician because, unless you ask the right question and use the right method for analysing the data the results can be worse than useless. books on stats dont help those who cant apply stats.
Posted by gerarda over 3 years ago
Ofcom say they are required to produce a report that reflects user experience. Per capita income does not reflect the actual experience of the population which is why average earnings based on the median is always the stat chosen by the press.

It is also in some ways a simpler measure than the mean.
Posted by R0NSKI over 3 years ago
So does "As with all our speed test analysis we average out individual results to avoid one person skewing an area with thousands of tests." mean that if I conduct 10 tests, 8 of which are low because I have connection issues, then the last two are fine, my average speed will be skewed because of the problems, not sure how you would overcome this, but I'm sure there's a way.
Posted by andrew (Favicon staff member) over 3 years ago
@R0NSKI There is a way, we employ a dozen people to spend all day every day checking your speeds and determining what looks to be the 'real' one.
Posted by themanstan over 3 years ago
Outliers will have an undue influence on the report if insufficient data points are available. Is this the case here?
Posted by andrew (Favicon staff member) over 3 years ago

We plotted all 2,300 testers from Ofcom

NOTE: Not all are used for the various results, subsets are created.
Posted by themanstan over 3 years ago
Yuk! Based on that i'd suggest using a Cauchy-Lorentz distribution.
Yes , for this mean is nonsense, luckily Cauchy-Lorentz doesn't allow means. But you can have subfamily distributions, best left to professional statisticians.
Posted by wrecker over 3 years ago
I do not know where OFCOM Dig up these figures from. They may be fine for the city living people but try living in a third world environment all of 50 miles from Central London.
Where we are lucky to get 1Mbit/s and at busy times it drops to dial up speed.
Posted by rkimber over 3 years ago
"Only 17 rural fibre based connections were used"

Rural fibre based connections - what are they? :-)

Where I live in small town West Wales we have copper and 5 - 6 Mbits.

It would be interesting to know the speed that most people have - i.e. the modal averages in urban and in rural areas.
Posted by galacticz00 over 3 years ago
Well that's really great to know but my average speed is 2Mbps with no promise that it will improve in the near future. I still pay the same as the folks that get 17Mbps though.
Posted by maxwild over 3 years ago
I agree that 2Mbps is barely fast enough but an urban ADSL2+ of 6.5Mbps is fast enough for HD TV so that suits me fine. No need to pay the extra £10pm for fibre!! Very few need more than 20 so I hope that any subsidy is exclusively targeted at the 2Mbps areas.
Posted by otester over 3 years ago
Availability is what matters, not average speed measurements.
You must be logged in to post comments. Click here to login.