Cox Communications is a top-10 US provider of TV and internet services. They have over 5MM residential and business subscribers in 18 states. Their network contains 25,000 node boxes distributing gigabit fiber signals to the node boxes each serving 500 customers with megabit speeds via the familiar copper RF coaxial connector.
The digital nodes are in all kinds of weather and experience about a 12% failure rate. Each node failure impacts 500 customers with degraded or missing TV picture and internet speeds.
Switchboard complaint calls have been the only way to know a node failure was underway. That could take as long as 2-3 weeks from first call to determine via a clustering of “no problem found” truck service rolls.
Cox operations monitors 27 digital signals (variables) for each node box. Cox engineering requested DecisionIQ perform a diagnostic analysis to determine if node failures could be predicted in advance of customer complaint calls, using existing data.
Diagnostics Analysis & Report
Node box operational history, customer switchboard history and corresponding zip codes were provided. The DecisionIQ engineering and data science teams curated the data and ran it through the machine learning platform.
The DecisionIQ platform was able to find a correlation of 17 variables that successfully predicted node failures before the first complaint calls. A monitoring solution was simulated. Projected savings were significant.