There’s an interesting debate going on between the FCC and the NCTA about measuring broadband speeds and ISP performance. The FCC is using comScore to calculate average data rates for broadband consumers. (See Ars Technica coverage) The NCTA is arguing that the FCC’s measurement approach is flawed.
While cable (and telecom) companies are naturally going to come out against any measurement that shows less-than-optimal delivery speeds, the NCTA has some very valid complaints. For one thing, in measuring user speeds, there is no accounting for errors outside ISP control. The NCTA references a few broad examples of this, but I can give a very specific one. When Motorola first started deploying DOCSIS 3.0 gear (two years ago!), there were some odd hiccups with Windows PCs. It turns out that a default configuration of Windows (and this may not be true with Windows 7) limits subscribers to 23 or 24 Mbps, far below some promised DOCSIS 3.0 speeds. Using the comScore measurement, consumers with a 50-Mbps service tier, who are also stuck in the Windows default configuration, would be recorded as getting about half their promised speed – through no fault of the service provider.
In short, if you’re going to measure real-world speeds, you also need some real-world caveats. ComScore measurements do not account for the hardware and software in use by consumers. Your mileage may vary.