A Tree Falls in the Forest and Nobody Hears It

On behalf of Google and its partners at the New America Foundation, Vint Cerf dumped an ocean of data on policy makers today that’s only going to confuse America’s broadband policy discourse even further. According to Cerf, the Google data is better than “advertised speed” data for assessing broadband deployment:

Google chief Internet evangelist Vint Cerf took the stage to show off new visualization tools that he said would help researchers make sense of the mountain of data. Cerf said the goal is to understand broadband experiences across the U.S. and internationally without relying on advertised speeds.

“There are lots of assertions about broadband rankings that are often not very quantitative,” Cerf said. “We’re interested in making those more crisp and understandable by getting solid data behind everything.”

Cerf is wrong and the data set is misleading. Google’s M-Lab system is one of several systems in the world that measures Internet performance. Similar systems are deployed by Ookla and SamKnows. All of these systems attempt to measure maximum TCP throughput between the consumers who choose to participate and some arbitrary point on the Internet (except SamKnows, which tests multiple arbitrary points.) They’re useful for getting a picture of consumer’s broadband plan, and by extension of the applications that the people who participate may run successfully. If you’re in the application business, as Google is, this is important data because it helps you understand how to tailor your apps to the currently available supply of bandwidth.

M-Lab is not at all useful for determining which countries are doing the best job of supplying bandwidth their people, however, but Cerf is pitching M-Lab as just this kind of tool with the rhetoric about “broadband rankings.” M-Lab doesn’t measure the network, it measures the plan the consumer has purchased. Rankings, if they have any meaning at all, should tell us something about the plans that are available, not just the ones in common use. This is because every criticism of a nation’s standing in the index of measured speed always leads to a plea for more network investment, higher speeds, and more high-speed switches and routers. All of this may be fine, but it’s at best a secondary or tertiary consideration for policy makers.

By any rational analysis, speed test rankings only measure one thing directly, and that would be the limits of the plans that people have chosen to purchase. If our concern is to make faster plans available, the M-Lab data has no value whatever, no matter how many visualizations Cerf and his colleagues have to offer.

America is in a situation with wireline broadband where the supply greatly exceeds the demand: 95% of Americans have access to some sort of wireline broadband, but only 65-70% choose to sign up. Wireline broadband adoption in 2010 was essentially flat, rising by one percent (85 million to 86 million connections) from the prior year, according to the FCC. And when people do sign up for broadband, they show a distinct preference for low-priced, low-performance service plans over high-priced, high performance plans.

You don’t need highly detailed and redundant data sets to understand this, and continuing to inject usage data into what is essentially an adoption problem doesn’t help anyone. Can’t we at least be honest about what we’re measuring?

The primary problem for American broadband policy – and broadband policy in every other country – is not performance, it’s adoption. It doesn’t matter how fast the network is if nobody chooses to use it. When a tree falls in the forest, and nobody hears it, its speed doesn’t matter.

To enjoy the benefits of broadband networks, people have to own computers, use computers, and sign up for broadband service. Broadband is cheaper than cellular service, but each of the top two cellular operators has more customers than all the wireline broadband network operators combined. It’s not a price problem, and it’s not a performance problem, it’s a “why do I want that?” problem, and having a bunch of network luminaries running around American stages talking down the alleged performance of American networks without any relevant data to back up the talk is not helpful.

Nine out of ten reporters covering this story are going to report “American networks are slow” when that’s not the story at all; “Many American consumers don’t see the value of broadband” is the real problem.