AT&T doesn’t want you to see its slow Internet speed-test results

AT&T doesn't want its home Internet speeds to be measured by the Federal Communications Commission anymore, and it already convinced the FCC to exclude its worst speed-test results from an annual government report. "AT&T this year told the commission it will no longer cooperate with the FCC's SamKnows speed test," The Wall Street Journal wrote in an investigative report titled "Your Internet provider likely juiced its official speed scores." AT&T already convinced the FCC to exclude certain DSL test results from last year's Measuring Broadband America report. The reports are based on the SamKnows testing equipment installed in thousands of homes across the US. "AT&T was dismayed at its report card from a government test measuring Internet speeds" and thus "pushed the Federal Communications Commission to omit unflattering data on its DSL Internet service from the report," the Journal wrote. "In the end, the DSL data was left out of the report released late last year, to the chagrin of some agency officials," the Journal wrote. "AT&T's remaining speed tiers notched high marks."

Pai’s FCC gives less attention to speed tests

The Obama-era FCC began the Measuring Broadband America program in 2011 to compare the actual speeds customers receive to the advertised speeds customers are promised. The FCC released reports annually through 2016, but the testing program has gotten less attention since Ajit Pai became chairman in January 2017. As we wrote in November 2018, the FCC hadn't yet released any new Measuring Broadband America reports since Pai became chair. Pai's FCC in December 2018 finally released both the 2017 and 2018 reports, tucking them into the final appendices of a larger "Communications Marketplace Report." You can see all the Measuring Broadband America results from over the years at this page. The 2017 report includes two categories for AT&T, one for its oldest DSL technology and another for its DSL-based IP broadband with speeds of up to 45Mbps. While AT&T's oldest DSL service only provided 82 percent of advertised download speeds, AT&T IP broadband was over 100 percent. The 2018 report only includes AT&T's IP broadband category, leaving out the company's worst results. Satellite Internet provider ViaSat also "left the FCC's program" last year, the Journal wrote. ViaSat results were included in the 2018 report, which covers tests from September 2017. We asked the FCC yesterday if it will include any AT&T and ViaSat test results in future reports, since SamKnows testing equipment could still be in AT&T and ViaSat customer homes, and we asked when the next Measuring Broadband America report will come out. We'll update this article if we get any answers.

AT&T says its own speed test is better

AT&T defended its decision to drop out of FCC testing when contacted by Ars. "AT&T developed a best-in-class tool to measure its consumer broadband services," the company said in a statement provided to Ars. "This tool measures performance on all AT&T IP broadband technologies and is more accurate, versatile, and transparent. For these and other reasons, our tool provides better and more useful information to our customers." But consumers have less reason to trust a speed-test tool created by AT&T than one created by the FCC. Even with the FCC's speed tests, AT&T was able to exclude unflattering results. It would be even easier to dump slow speed-test results when AT&T is the one determining which numbers to show the public. AT&T and the mobile industry's top lobby group have also argued that carriers shouldn't have to submit detailed 5G maps to the FCC. Separately, the FCC said this month that Verizon, T-Mobile, and US Cellular exaggerated their 4G coverage in official government filings. Back in 2011, AT&T touted the FCC's in-home speed tests as being far more accurate than previous testing methodologies. But the company's opinion then seems to have been influenced by early test results that AT&T said showed "consumers are getting high-quality broadband services from their ISPs."

AT&T defends exclusion of “obsolete” DSL

AT&T today also defended its successful attempt to exclude certain DSL data from the 2018 FCC report. "First-generation DSL is an obsolete technology that AT&T does not market and, based on the FCC's practices with other ISPs, should have been excluded from its reporting," AT&T told Ars. "The FCC provided guidance to ISPs that a speed tier would no longer be included in the MBA [Measuring Broadband America] program if it wasn't actively marketed and is less than 5 percent of the subscriber base. Also, the FCC omits speed tiers that are below its threshold of a minimum number of statistical targets. DSL speed tiers were always below statistical targets." But not everyone at the FCC agreed with the decision to omit AT&T's slowest results from the report. The Journal wrote:

AT&T was repeatedly at loggerheads with the FCC over its results and worked closely with the agency to have favorable items included in the report, a former AT&T employee familiar with the testing program said. FCC officials decided reluctantly not to publish AT&T's DSL data in last year's report, according to Walter Johnston, a former FCC official who oversaw the program until he retired last year. "Personally, I was not happy about that," he said. Overall, Mr. Johnston defended the FCC test, saying the agency never meant it to be a "consumer guarantee."

The Journal report also included details about other providers trying to artificially boost their FCC speed-test scores. "Sometimes providers boost speeds for households during the actual FCC speed-testing period," the Journal wrote. "Comcast a few years ago upgraded speeds in some regions without notifying the FCC, making test results look stellar, people close to the FCC program said. The FCC discovered the changes after spotting anomalous data and adjusted the numbers." A 2017 lawsuit filed by the New York state government accused Time Warner Cable (now owned by Charter) of manipulating speed-test results and of deceiving the FCC in order to get bad speed-test results excluded from an annual report. Charter eventually agreed to pay $62.5 million in refunds to 700,000 customers in New York.

ISPs upgrade homes in testing pool

For each yearly test, the FCC "inform[s] the Internet providers the dates of the testing period and send[s] them lists of individual testers' names so they can confirm which speed packages the customers have signed up for," the Journal wrote. "That makes it possible for companies to keep tabs on many testers from year to year, engineers said," the Journal report continued. "Providers can use that information, as well as real-time data from the dashboard SamKnows sells them, which anonymizes the users but shows their states and regions, to make targeted upgrades that improve service for those households, engineers said." Cox was able to get some poor results excluded from the 2018 report because it blamed a weak score on "a congested link to an Internet middleman [that] had lowered scores for some of its Arizona customers," the Journal wrote. "Some FCC officials were uncomfortable with the omission" because "network experts say broadband providers share responsibility with Internet middlemen to keep their connections upgraded." Disclosure: The Advance/Newhouse Partnership, which owns 13 percent of Charter, is part of Advance Publications. Advance Publications owns Condé Nast, which owns Ars Technica.