Broadband News

Why we don't think average UK broadband speed is 49 Mbps

Average broadband speeds and are you getting what was advertised has been all over the headlines for the last few weeks. We have had our extensive statistics on broadband speeds and coverage across the UK available for a couple of years now and believe that all the work that goes into monitoring the market in the UK and watching the patterns that emerge from our speed test results each month and have been following those trends since the days of 0.5 Mbps being the common consumer product.

So it is a bit of surprise to see that Ookla has launched a new global broadband speed index that places the UK at position 24 with an average download speed of 49.22 Mbps (upload 9.55 Mbps) - impressive!.

Alas we think that Ookla has got its analysis wrong (e.g. a few people testing a lot skewing the result) or is including an unusual number of Gigabit based academic and large corporate networks such as the major banks in its analysis, thus the figure is of little use to the consumer and SME market. We cannot comment on other countries but have included the following table of calculations showing why we believe 49 Mbps is far too high and suggest others with a sound knowledge of what people are buying and experiencing speed wise do similar.

The maths behind how you get to a 49 Mbps average speed for UK consumer/SME broadband
TechnologyMillions of premisesAverage Speed
FTTC/VDSL 8.1m 45 Mbps
ADSL/ADSL2+ 11.9m 10 Mbps
Cable 50 Mbps 1.45m 50 Mbps
Cable 100 Mbps 1.85m 100 Mbps
Cable 200 Mbps 1.45m 200 Mbps
Cable 300 Mbps 0.25m 300 Mbps
FTTH/FTTP 0.3m 500 Mbps

The split across the cable products is based on the rations we see each month from Virgin Media cable customers, so there may be some variation and given all the issues that existed until some point in June we would not expect the cable products to nail their speeds when talking about an average for each tier (an update on Virgin Media performance for July is expected in the next few days). The averages for VDSL2 and FTTH are probably also generous since the majority still buy either an up to 38 Mbps or up to 52 Mbps product and line distance issues will impact speeds, and on the FTTH/FTTP products people still very much buy a speed based on the size of the wallet usually.

Once you add up the 25.3 million connections and do the maths we arrive at an average of 49.6 Mbps and this should be the sort of sanity check that anyone publishing a nation wide statistic that is going to repeated as fact by many outlets should carry out.


Average is not the right metric, what's the median?

  • keith969
  • 12 days ago

Absolutely no idea what the median score is for the data Ookla hold, our stats site at does give mean and median plus the top and bottom 20% speed range figures.

  • andrew
  • thinkbroadband staff
  • 12 days ago

Worse than that, speed test results use sites that can DELIVER content at high speed. What matters to users is what their own browsing and downloading is like. Even if Ookla was right, the average real in - use speed will be pathetically below that. For example even Windows updates don't deliver at anything like that. Until the content providers all have sufficient bandwidth to deliver, my 76Mpbs is pointless. (I have it as it was the only way to get 9Mbps upload to make Dropbox and Flickr usable.)

  • Fellwalker
  • 12 days ago

Whoops! 19Mbps.(no edit facility for my original comment)

  • Fellwalker
  • 12 days ago

Fiber coming soon but won't be ordering it, turns out the speed I'd get would be pretty poor (<40Mbps) versus what I get on 4G at the moment (>70Mbps).

  • DrMikeHuntHurtz
  • 12 days ago

Paragraph one, sentence two currently reads as "We hav had . . ."

s/hav /have /

  • burakkucat
  • 12 days ago

@DrMikeHuntHurtz: I don't think you got fibre in your area. Fibre usually doesn't drop in speed for the access network, and is usually capable of 300Mbps, or 1Gbps, depending on provider. Some fibre providers do offer though lower-speed entry level packages, e.g. the 50Mbps 'Home 50 Unlimited' from Gigaclear.

  • JNeuhoff
  • 11 days ago

DrMikeHuntHurtz JNeuhoff

i think he means near fibre (VDSL FTTC) not actual fibre to the house (FTTP)

FTTC should never been called fibre just because a small % can get 70mb (as its going to get really confusing when openreach actualy providers Fibre) i guess calling it fast upto 30 to 70mb VDSL does not sound catchy

its like the usa networks calling 3g DC-HSPA+ 4g when its not 4G

  • leexgx
  • 11 days ago

should've been called "near fiber upto 38mb/78mb" ?

  • leexgx
  • 11 days ago

That is a new, exciting and original point Mr Neuhoff. My life is enriched seeing your name appear here knowing that you are likely to state that. Again. Ad infinitum.

  • CarlThomas
  • 9 days ago

What's the median?

In terms of line speed, rather than download speed:

Note the number of packages sold in @Andrew's table. 53% are now "superfast" packages, so the median (50th percentile) is probably amongst the "up to 38Mbps" FTTC packages. In fact, it looks like the median will be at approx 0.8m of the 8m FTTC connections, or above the slowest 10% of FTTC.

Some graphs from Ofcom (2014-15) suggested that around 10% of all FTTC connections were sub-30Mbps. Some 2016 graphs suggest similar numbers.

That probably puts the UK median line speed at around 30Mbps.

  • WWWombat
  • 9 days ago

I find it hard to believe such analysis. I live 4 miles from Liverpool & have a fibre cabinet 100m down the road. I upgraded to Fibre without doing any research. More fool me, my telephone lines travels 1.4km to my actual fibre cabinet. My line contains aluminium. Result is my speeds are 18Mbps & 1.1Mbps. Hardly superfast.
I suspect there are many properties like ours built in the late 50's that were on the edge of urban areas in those times that have ali lines that can never achieve >20meg. There is absolutely no prospect of a new Cu line or connection to the nearby FC. Slow lane for us!

  • dar2211
  • 8 days ago

@dar2211 1/2
The mass market for telecoms works on the statistical effect of providing service to a lot of lines. In statistics, some people are outliers - and in this case, you are one of them.

While it is hard for such individuals to accept the analysis, the true answer comes from looking at hundreds and thousands of lines.

Though it doesn't help you at all to know this, around 4% of premises are 1.4km+ from their cabinet.

Any solution for these people involves adding extra infill cabinets, or rearranging copper to closer cabinets. It does happen, but only late in the BDUK projects.

  • WWWombat
  • 7 days ago

@dar2211 2/2
Aluminium was a feature more from the late 60's to mid 70s.

Not that it matters much: Homes with phones were few and far between in 1965 - just 2% of homes. Almost every domestic phone line in the country was created after 1965.

Mind you, there have been more than 10m homes built since 1965 too.

The haphazard nature of building houses also has an impact - and explains why more cabinets have been added as infill, and why the nearest cabinet isn't necessarily the one that serves your home.

And, until now, there has never been a reason to retrofit everyone to the closest cab.

  • WWWombat
  • 7 days ago

Only 2% of homes connected in 1965? Almost every domestic line created after 1965? If correct, then given the the comment that aluminum was more a feature from the late 1960s (i.e. after 1965), we should not be surprised that it is such a problem in the estates and tower blocks that were built and/or connected in the housing booms of the 1960s and early 1970s.

Hence the pace at which old tower blocks are being retrofitted with full fibre by players like Hyperoptic.

  • PhilipVirgo
  • 7 days ago

We need to use current, not historic tests. Performance falls (for a variety of reasons) as usage (and contention) build up

Thus my FTTC line which ran at over 70 down and 18 up, after the last repair, is currently (just tested it)running at just over 43 down and 23 up. I assume part of change is rebalancing but rather more is contention.

We really do need to find a way of measuring performance based on what consumer actually experience - not the speed in the middle of the night when (it is alleged) that most line tests are (for obvious reasons) run.

  • PhilipVirgo
  • 7 days ago

Post a comment

Login Register