Skip Navigation

Ofcom publishes annual round-up of 19 broadband package speeds
Wednesday 12 April 2017 11:31:45 by Andrew Ferguson

Ofcom monitors broadband speeds in the UK via two main methods, sync speed data supplied by operators which is usually summarised in the Connected Nations report and a deep analysis from actual testing via some 4,824 linux boxes deployed to panellists across the UK. Today sees Ofcom release the results from analysis carried out in November 2016.

Click image for larger version

Obviously comparisons will be drawn with our own monthly publication of speed test results, but there are lots of differences in the way the data is collected, at the simplest level you can call the Ofcom data the modelled approach and ours is a massed observation exercise. Various plus points and minus points exist on both, e.g. ours includes Wi-Fi which can influence results, but conversely with so many using Wi-Fi all the time its important in the full consumer experience, and modelling from a small sample can miss variations and it is also difficult to look at areas like FTTP (Ofcom has recruited more FTTP based testers, but not enough to feature in this report). One emerging trend is that ECI cabinets for VDSL2 are not increasing in speed the same way as Huawei cabinets due to the roll-out of various technical changes.

"Cable broadband is typically faster than copper-based services, but a significant minority of users experience severe slowdowns in peak times

Contention occurs closer to the customer in superfast cable networks, making it more difficult to add new capacity to reduce the effects of network congestion. But despite suffering from higher levels of slowdown than ADSL and FTTC, superfast cable services still had the highest average download speeds throughout the day. There was also much greater variation in the levels of contention experienced by superfast cable customers, with a significant minority experiencing severe slowdowns. For example, while 47% of ‘up to’ 50Mbit/s cable panellists had a peak-time average speed of 50Mbit/s or higher, 9% received less than 10Mbit/s. There was less variation in the performance of FTTC connections; less than 1% of our FTTC panellists received a peak-time average speed of less than 10Mbit/s, while less than 5% had a maximum speed equal to or above the advertised speed, compared to 90% of superfast cable panellists."

Ofcom on cable broadband speed variations

Each month we cover our speed test results there are invariably questions about the Virgin Media speed variations, but Ofcom is clear that their data is also recording wide variations, and given the fact that this can be very localised with cable broadband the reality for some people is likely to be very different to the Ofcom results. As with previous SamKnows results Virgin Media cable also fares badly for jitter which is crucial for gamers and its worse at peak times with cable broadband performing worse than ADSL2+ in this respect.

As an incentive to upgrade to a better service if it is available (and millions have the option but have chosen not to, which is especially galling for those crying out for better services and do not have the option) Ofcom has looked at disconnections, with ADSL performing worst at an average of 1.1 disconnections per day, ADSL2+ was better a 1 and cable and FTTC had a combined average of 0.3. So much better, but oddly the data suggests the cable up to 200 Mbps service is worse than the cable up to 50 Mbps service for disconnections and up to 52 Mbps is better than up to 38 Mbps VDSL2.

One take-away for anyone looking at the difference in performance in rural and urban areas, is that Ofcom has used the BlueWave Geographic Locale dataset to split into 7 groups, this set is apparently widely used in market research, our data analysis relies on the ONS classification system, thus the two may not fully align and this is important when comparing speeds, coverage and other metrics.


Posted by bartman007 14 days ago
Give me a few grand and I will write a report and make horse crap look like roses.
Posted by mike41 14 days ago
Re ADSL2, PN appears to be 0.5Mbit/s faster than BT although both are on same network one presumes.
Posted by 961a 13 days ago
Why does it take them 4 months to publish?
Posted by rtho782 13 days ago
PN appears 0.5mbit faster than BT but this could be simply that there is a slight difference in the sync speeds of the sample houses.
Posted by comnut 13 days ago
.... and every forgets, just like cars, YMMV...

they say '50 Miles per gallon' but of course it depends on how many big hills, long straights, long winding roads, and of course how much other traffic is on the road, reducing your speed, and wasting petrol!!! :)
Posted by Kr1s69 13 days ago
I'm not sure I see much point in this kind of analysis. I don't know how you isolate the network performance from that of the underlying openreach network.

Speed tests are only relevant if you know the connection speed.
Posted by Mikejc 8 days ago
What does this mean in plain English "Contention occurs closer to the customer in superfast cable networks, making it more difficult to add new capacity to reduce the effects of network congestion." Presumably if you add more capacity you reduce contention. Period.
Posted by aaaashy 8 days ago
the minimum given for VirginMedia (up to 200Mb) does not go anywhere near the minimum i receive which can be as low as 5Mb/sec, not sure where this report gets its number from, but it sure isn't London
Posted by Mikejc 8 days ago
aaaashy, be Happy :-).
The lowest I have recorded has been 0.04 Mbps on a theoretical 77 Mbps service i.e. 0.5% of what I am paying for. Fortunately it does not happen too often, but that is not the point, delivering one 200 hundredths of what one is paying for is totally unacceptable.
Posted by ian72 8 days ago
@Mikejc - the further out in the network contention occurs the more places you have to upgrade to deal with it. If you contention essentially all happens on one central pipe then a single upgrade will fix it. If it happens on 10,000 different nodes (random number plucked out of air) then that is a lot of network links to have to upgrade.
Posted by Mikejc 1 day ago
@ian72 - thank you for that. I understand that with a tree typology distribution system (which is I assume what we have) then closer to the root one will be dealing with a fatter telecoms channel, and thus less of them, but it it is probably a) more expensive, and b) more difficult to fix in so far as you may have to take more people out of service whilst you fix it.
Posted by Mikejc 1 day ago
Further away from root I imagine that you are dealing with smaller, less expensive equipment, and they will not all get overloaded at the same time, so VM should have a rolling pre-planned program to upgrade or replace them, which in my book is less difficult. The problem is, I suspect, that they do not. It is all done only when the customers are screaming at them.
You must be logged in to post comments. Click here to login.