April 13, 2011 - 12:18 pm
Walter Johnston & James Miller | Office of Engineering and Technology

Consumers may need better information to determine which broadband service to purchase to meet their needs. Yesterday, the FCC’s Consumer and Governmental Affairs Bureau (CGB) released a Public Notice requesting comment on this topic. As CGB Bureau Chief Joel Gurin noted, the metrics for broadband speed and performance are not as meaningful to consumers as, say, measures of fuel economy:  “Most people don’t understand megabits-per-second in the way they understand miles-per-gallon.”
But in another way, fuel economy and broadband performance are similar—they’re complex measurements influenced by many factors. How much gas you use depends on everything from how full your tires are to how cold it is outside to how aggressively you drive. Similarly, a single user checking email at 10 AM might find a lower speed broadband service to be sufficient, but the same connection might not support a household with a heavy gaming user, a VoIP- and VPN-using telecommuter, and movie fans watching HD-quality streaming video in the living room at 7 PM. 
It’s no simple task to determine how “fast” your Internet connection is; just developing a measurement methodology that accurately captures and represents performance is a serious technical challenge.
The FCC took on this challenge when we set out to test broadband performance in the homes of 10,000 volunteers across the nation. To do this test, we contracted with SamKnows, a consulting company that conducted a similar effort in the United Kingdom with the regulatory authority for the UK communications sector. Our nationwide test is the first precise U.S. measurement of internet services as delivered to the home by the service provider.
This program places a special hardware device in each volunteer’s home to measure broadband performance. These devices, called “white boxes,” continually measure internet service as it is delivered to the home, independently of other network activities and the performance of hardware in a subscriber’s home, like the home computer or WiFi router. Results from each “white box” are periodically aggregated at a central server, where approximately 100,000,000 data points are collected each day.
To design the processes and methodologies for this trial, the FCC collaborated with a number of interested parties. We’re grateful for the support we’ve received from industry, public interest groups and academia. Over a period of months, proposals were discussed, critiqued and modified by all parties. The methodologies developed represent the first broadly-accepted measurement program to characterize broadband Internet service across service providers. Independent network test points were provided during the trial by M-Lab, and many internet service providers installed, at their expense, additional testing servers at various points in their network. During these months of discussions, the FCC and SamKnows provided white boxes to several academics and others whose observations improved the program.
The information collected will allow us to analyze, in detail, upload and download speeds as well as other performance parameters, including latency, availability, domain name resolution, simulated VoIP, video streaming and website access times. In addition, working with the service providers, we can verify the actual speed or service tier each customer subscribes to, giving us a clearer picture of how actual performance matches with advertised. This effort, combined with responses to yesterday’s Notice, will help consumers make more informed buying decisions about what kind of service and what service tier they should purchase to support the applications they need.
These and other technical details will be described in greater detail in one or more forthcoming FCC reports on the broadband testing and measurement program. We’ve held several public meetings and will continue to reach out in other public meetings soon. In the meantime, staff will continue to respond to questions via FCC.gov.