Home » Internet

Measured broadband versus advertised broadband ranking

By 9 October 2009 11 Comments

It’s one thing to have pretty graphs and fancy charts, but substance should always trump superficial looks.  Unfortunately, too many people are citing the latest estimated broadband rankings without analyzing what the numbers mean or ever question if the numbers are right to begin with.  When we compare real-world Akamai results versus synthetic speedtest.net tests versus 2008 estimates from ITIF, we see a huge disparity in the numbers.

Note that all costs are based on ITIF estimates.  ITIF costs appear to be underpriced for countries like Japan so the actual cost per Mbps delta between Japan and the US is far smaller than this chart suggests.

Akamai is the world’s largest CDN which alone handles 20% of the world’s Internet traffic.  It is a real-world test because it’s measuring how fast people are getting the files they are actually looking for and use.  Speedtest.net uses short synthetic sample files and the results may be inflated since ISP’s often leave speedtest.net traffic alone because there is no need to deprioritize short burst traffic.  Some even theorize that this is done by ISPs to make their service look better than they actually are though I haven’t seen anything that confirms this theory.  As for the ITIF estimated bandwidth, they seem extremely inflated compared to any measured results.

Another interesting statistical artifact is that when we compare the top 10 individual states in the U.S. to Europe or compare U.S to Europe, Europe comes out behind in both cases.  Does that mean it’s time to celebrate in the U.S.?  No, because it is just another statistical artifact.  But the fact of the matter is, comparing individual European Union states to the United states as a whole is really no different than comparing the United States as a whole to individual EU states.  That’s why it makes little sense to put too much emphasis on any of these statistics.

Another factor that isn’t considered is how the pricing data is gathered.  Most of the time, they don’t factor in usage caps and the U.S. has some of the most generous usage caps in the world.
U.S. versus global usage caps on Internet
Source: OECD, Comcast, AT&T, TimeWarner (no Japan data from OECD)

So if we factored in price caps by using Mbps/dollar/GB, the results would push the U.S. way up over most other nations (except Japan and Korea) because U.S. broadband providers typically offer 4 or more times usage allowance.  But this metric would be just as inflated as the Mbps/dollar metric because larger usage cap plans will always score much higher but at the expense of everyone paying a little more.  Countries like Korea and Japan which have accelerated broadband deployment will always have lower Mbps/dollar because broadband service tends to stay at the same price when speeds go up.  The U.S. and Europe will be following similar trends in the next few years in adopting faster broadband connections and this will affect the Mbps/dollar metric.  But again, let’s not get too hung up on statistics.  What’s important is that we continue to grow at a reasonable rate.


  • Paul William Tenny said:

    It’s one thing to have pretty graphs and fancy charts, but substance should always trump superficial looks.

    Absolutely, which is why I’m dropping by to point out (not to you, George, because you already know this) that your graph doesn’t tell the whole story either.

    We had a nice discussion about broadband a few weeks ago and a big sticking point was keeping satellite service in the picture for defining what is and is not “broadband”. If it’s that important because that’s all that millions of people can get (even though they can’t, because the only two real players in the U.S. are both maxed out on subscribers) then tossing a satellite provider into that graph really changes things. Wildblue’s 1.5mbit tier has a 17 GiB/30 days (rolling) cap, and Verizon’s EVDO plan — so-called “mobile broadband” — I hear has a 5 GiB/month cap.

    Verizon in particular shows the obscene extremes. FiOS would require a graph of infinite height because they have no caps, but their EVDO service looks like it’s competing with the Slovak Republic for worst on the chart. (And I’m not saying that’s unfair or unreasonable, obviously they are two very different communication mediums.)

    I wonder what our bar would look like if you did some averages instead of just casually tossing a few sample providers on there that happen to be the providers with the highest caps that aren’t unmetered.

    That’s important because those other country’s must be averaged already. My friend with one of Canada’s bigger cable outfits has a 100 GiB cap, double what that graph shows.

  • George Ou (author) said:

    Paul, I even said that my graphs and statistics shouldn’t be taken too seriously. The point I was trying to make was that my graphs and statistics make as much sense as the other graphs and statistics. And while these numbers are useful in some ways, we should not put too much weight into them.

  • George Ou (author) said:

    Don’t ever compare wireless services to wired broadband service. Yes it costs way more for wireless services and it’s much slower, but how fast is your broadband service when you’re away from your house? What you’re paying for is mobility and mobility never comes cheap.

  • Digital Society » Blog Archive » Why municipal fiber hasn’t succeeded said:

    […] many other countries in broadband deployment for a variety of reasons[i] [UPDATE 3/11/2010 – the rankings are extremely dubious], some communities have lost faith in the private sector and have begun to look to community-based […]