LARGE SCALE DATA PROCESSING AND WAREHOUSING
The MassiveData Matrix™ aggregates and processes second-by-second television viewing data collected across a large number of markets and a wide range of devices. The Matrix scales as more markets and viewing platforms are added to the TVPanel™, and the Matrix ingests and distributes data reliably every day. For these reasons, the Matrix is built upon an Apache Hadoop architecture that supports large-scale dataset processing within a fault-tolerant, redundant hardware environment that is both highly scalable and massively parallel making it the leading choice for Big Data solutions.
As data comes into the MassiveData Matrix, it is cleansed and normalized so the resulting viewing events are consistent and complete regardless of their source (data collection technology or MVPD). In this process, a series of edit rules are applied to remove devices or events not suitable for reporting so that only high-quality, reliable data gets into the TVPanel. Next, the appropriate channel and program information is integrated based on the channel line-up for the source market. Lastly, the data is connected to anonymous household identifiers that allow the viewing data to be connected to first and third-party datasets to dynamically describe audiences while also strictly preserving privacy.
REVEAL: ANALYTICS ENGINE AND REPORTING
The Reveal™ Portal delivers real-time and scheduled report generation on the TVPanel stored within the MassiveData Matrix. It contains a set of standard audience measurement reports that rank and compare channels and programs. The Rank and Compare reports can be run for a specified date range and filtered by DMA, Zip code, and/or household demographics. Rank and Compare report on ratings, shares, reach, and total minutes viewed for live viewing as well as same-day, live+3, and live+7 based on DVR playback data. The Rank and Compare output is grouped by the top channels or programs matching the report criteria and is calculated down to the second, or rolled up to the minute or quarter hour.
In addition to the standard audience measurement reports, users can generate their own ad hoc reports within the Reveal Portal and save the reports for future use. All reports can be exported to PDF, Excel, or CSV formats.
MassiveData® uses data mining techniques to produce profiles on TVPanel households, individuals, and set-top boxes. The profiles are inferred using a proprietary statistical method from aggregate households without reference to any Personally Identifiable Information (PII). Complex analytics based on household composition, TV viewing patterns, and attitudes and interests are enabled via these models.
Every household contains thousands of descriptive characteristics including demographic, geographic, and behavioral features. In addition to the basic demographics commonly available, MassiveData incorporates attitudes, beliefs, and interests determined from behavioral information that is statistically propagated throughout the aggregated population of panel viewers.
MassiveData allows household and individual profiles to be segmented based on any combination of features as needed. Unlike conventional static audience segmentation that pigeonholes households into pre-defined groups, MassiveData segments can be created dynamically for any analysis.
“Partnering with FourthWall Media to produce the Mavs Team Widget is great for us.
We believe in the future of interactive television.”