Ofcom, the U.K. communications regulator, reports that average actual fixed-line residential broadband speed has increased by over 25 percent over the past year from 4.1 Mbps to 5.2 Mbps as Internet service providers sell higher-speed broadband packages, and as customers buy them.
In November 2008 just five percent of U.K. consumers had a service offering speeds up to 10 Mbps. In May 2010 fully 25 percent of U.K. consumers were buying a service running at 10 Mbps or higher.
Conversely, in November 2008 29 percent of U.K. consumers bought services running at less than 8 Mbps. In May 2010 just eight percent bought services running at less than 8 Mbps.
But Ofcom also notes a greater disparity of advertised speeds and actual speeds on digital subscriber lines. In practice, users got 3.3 Mbps on packages described as supporting 8 Mbps to 10 Mbps.
What to make of the results? They actually show a more-complex situation than might first appear.
First of all, the increase in average "experienced" speed has at least a few causes. Providers are offering higher speeds, users are buying packages offering higher speeds and there might also be a bit of a shift to cable-provided services, which will raise "average" speeds simply because cable modem services appear clearly to be more capable of delivering an "advertised" speed than digital subscriber line services, which are distance dependent.
The marketing gap appears to be the bigger "problem." Why do carriers advertise "up to" speeds? Because there is no way, on an unmanaged connection, to know precisely what the end user experience is going to be, or how fast the connection will be. Under those conditions, marketers can tout only their lower prices or better value.
There is a logic to citing only "typical" or "real world" speeds, but that presents some problems. Even an "average" speed is going to change over time, vary between neighborhoods, time of day, day of week and types of applications used.
And to the extent that ISPs need to determine "typical" speeds, the method of measuring and who measures is important. Researchers at the Massachusetts Institute of Technology, for example, have found that the way speed is measured can result in figures that vary by 50 percent to 100 percent.
Average "speed" also is affected by all the other parts of the Internet ecosystem, including the user's hardware, software and server performance on the far end of the connection.
The reason there is "marketing creep" is that broadband access tends to be seen by end users as a commodity with just two fundamental variable attributes: speed and price. To the extent that prices can be hard to compare or offer little room for differentiation, "speed" is going to get the attention, and higher numbers are better than lower numbers.
Also, there is no contradiction between "highest" speed and "average" speed. They simply are different measures of performance.
Ofcom’s research, conducted in partnership with broadband monitoring specialist "SamKnows", found that 24 percent of U.K. fixed-line residential broadband connections had a headline or advertised speed greater than "up to 10Mbps" in May 2010, compared to just eight per cent in April 2009.
On telco-offered 20 Mbps or 24 Mbps packages users got 6.5 Mbps. On telco-offered 10 Mbps packages users got about 8.7 Mbps.
On cable-offered 10 Mbps packages, users got 8.7 Mbps. On cable-offered 20 Mbps cable services, users got 15.7Mbps.
In April 2009, average actual download speeds were 4.1 Mbps, 58 per cent of average advertised "up to" speeds. In May 2010, average download speeds were 5.2 Mbps, 45 per cent of average advertised "up to’ speeds."
Perhaps the more-important finding is that cable broadband services delivered significantly faster actual (or download) speeds than comparable DSL services. In part, that is because cable modem services are not distance dependent, as are digital subscriber line services.
Virgin Media’s services delivered average download speeds around twice as fast as DSL packages with the same or similar headline speed. But it is fair to say no ISP can deliver anywhere close to its headline speed at hours when most users are awake.
Virgin Media’s 50 Mbps service reached 36 Mbps, but only between 4 am and 6 am. That's the other problem with measuring "average" speeds: they vary greatly, and always will.
The cable broadband service on average delivered higher download speeds at all times of the day than comparable DSL services, however it showed a greater slowdown during peak periods than some DSL providers.
The other observation one might glean from the Ofcom results is that consumers get the "best value" from services running between 8 Mbps and 10 Mbps, as "experienced" speeds more closely approximate "advertised" speeds than for packages advertising 20 Mbps to 24 Mbps speeds.
One can surmise that this is the case because more fo the latency begins to appear when the local access pipe effectively does not impose constraints. In other words, the contribution of far-end servers starts to dominate application latency.