The FCC initiated its Measuring Broadband America (MBA) program in 2010 and has since then provided periodic reports on a sample set of consumers served by Internet Service Providers (ISPs) who serve approximately eighty-five percent of broadband consumers in the United States. In consultation with the program stakeholders, the FCC has agreed to base each MBA report upon a single reference month. The definition of a reference month allows all participants to be compared across the same time frame. Reference months are extensively audited for data integrity to help achieve a high level of accuracy in reports and the data set for the reference month associated with each report is released with the report. The FCC only uses data gathered in agreed-upon reference months and audited according to the established methodologies in the comparative analysis of ISPs included in its reports.
As part of its policy of openness and transparency for the MBA program, the FCC periodically releases all data collected through the program as part of a raw unaudited data set. Such data is provided “as is” with the FCC making no assertion as to the quality of such data. Specifically, the extensive integrity checking that is performed preparing data for each report during both the collection and in post-collection data analysis and processing has not been performed on this raw collected data. In addition, related customer data sets such as speed tiers for individual panelists and the identities of ISPs supporting each participant are verified only for indicated reference months and may be affected by such factors as customer churn following the verified months. Because such necessary auditing of data is not performed on raw data, the data is not suitable for FCC use in comparing ISP performance. However, the FCC recognizes that this data can have value to the research community and others interested in broadband network performance, and releases it for this purpose.
During the collection of data for the July 2012 MBA report, routine checks performed by the FCC’s contractor, SamKnows Ltd., detected data collection anomalies during March 2012 and for a more limited period in April 2012. As a result, the July 2012 report was based on a defined reference month of April 2012 excluding the period from April 8 through April 13, which is the period during which anomalies were detected. Several independent technical reports provide analysis of these anomalies and impacts on the data collected. At the request of the FCC, SamKnows described the isolation of the anomalies and inaccuracies in a report available at Samknows Data Analysis March 2012 Report. Detected anomalies associated with the month of March 2012 were also studied by Steve Bauer of MIT and Steven Soltesz and William Hawkins of the Open Technology Institute. Their report can be found at March Data Analysis.