Mapping is very important, you’re absolutely right. However even CWA’s method isn’t so hot because the sample size is pathetically small.

My idea: if the cable/DSL companies don’t readily give boundary markers for their systems, grab a list of valid addresses (the post office has those, right?) and plug ‘n’ chug with each respective company’s qualification tool. Either you’ll get results that are as accurate as those companies have (not 100%, but 99% at least) or you’ll get the companies annoyed enough to provide data voluntarily.

Of course, for the smaller companies quo don’t have prequal pages the situation will be a bit different, however between Verizon, AT&T, Qwest, Comcast, TWC, Charter, Insight, CableOne, etc. (who do have such a tool) you’re going to cover a lot of ground.

They’ve been tailor-made to win state mapping contracts, in part thanks to Sen. Dick Durbin (D-Illinois), who creatively wrote language in the authorizing legislation that effectively funnels mapping grant money only to well-connected non profits like CN.

Instead of the kind of fine grade map information, including pricing, speed, competition, and access (right down the street level), we end up with maps like the one CN produced for Kentucky which showed the state had 100% broadband coverage – a ludicrous conclusion.

I’d actually trust CWA’s findings over that of Connected Nation, because it at least depicts coverage and speed.

For example my brand-new Comcast Ultra DOCSIS 3 connection, boasting a whopping 22 Mbps of download speed (roughly 31.5 Mbps with PowerBoost on) registers only around 18.5 Mbps on the SpeedMatters test. It’s not my end that’s the problem here; speed tests elsewhere (including ones that aren’t run by Comcast) report 22+ Mbps.

On uploads SpeedMatters similarly falls short; I get 11.5 Mbps or so with PowerBoost to other speed test sites (sometimes closer to 8 Mbps) however SpeedMatters puts me below 7.

Additionally the folks at SpeedMatters are using a TINY sample size for their connection tests; 400,000 connections is less than 1% of the number of connections in the US, assuming no user tested twice. More to this, SpeedMatters includes in their results tests in areas where only one person in a given zip code took the test. I’d say that five different IPs per zip should be required to make any sort of educated guess about that location’s real internet speeds.

I would trust Speedtest.net more with speed rankings, and there we compare a litle more favorably. The US average in that location is 6.79 Mbps down, 1.51 Mbps up. Granted, Speedtest.net likely has an upward bias on speed rankings, but I’d trust their dozens of millions of results versus SpeedMatters’ 400k. Plus, Speedtest.net still shows that the connectivity situation in the US isn’t too rosy; we’re 27th on the list by download speed, 28th by upload speed.

To get to the top of the list the average connection speed would have to rise to 21 Mbps down, 9 Mbps up, something that’s very doable with even DOCSIS 2.0 systems if managed correctly, though DOCSIS 3 and fiber leave a whole lot more headroom.

Of course, to *really* get some accurate stats, check with Akamai. They serve enough content to have a few fingers on the pulse of the internet, so I’d trust their information more than others’. No upward bias there since they serve pretty much every type of content online. I’d put my trust in those results, since to my knowledge they’re completely unbiased.

Not that I don’t want 25 Mbps down, 15 Mbps up (or better) fiber available everywhere. I just want any speed results representing the nation to be accurate.