T-Mobile's journey to turn crowdsourced big data into actionable insights

Description

Voice calling is one of the key services provided by cellular network operators, and the most important performance measure of this service is the audio quality. Alex Yoon walks you through how T-Mobile leveraged multiple teams to implement a crowdsourced solution to measure and respond to audio quality metrics and to put data into the hands of decision makers.

Turning this big data idea into an action requires multiple teams with different expertise. Mobile device engineers and radio frequency engineers are essential to invent the mechanism of measuring the audio quality. Technical account managers are a key link to persuade the cell phone manufacturers to implement the mechanism. Another major link comes from data scientists and data engineers, who enable easy interaction with the massive amount of data. Finally, data analysts interpret the data and find the correlations needed to turn the big data into insights. When combined with the field engineers who tune the configuration to run the test, the solution can be deployed and iterated upon rapidly. Making a real change out of the big data is a journey of achieving one goal with collaborations among many teams.

Crowdsourcing starts with identifying the data to collect. From the research in the lab, a mechanism was invented to detect the audio quality without having to listen to the actual audio. This mechanism was implemented in the cell phone software, then verified in the lab for data integrity and to ensure minimal impact on the battery performance. The software has been delivered to millions of cell phones to enable daily reporting of the audio quality.

In a collaboration with the field team in LA for a limited trial, the collected audio quality data was put to a practical use. A correlation analysis was performed to identify the network parameters that are most predictive of the audio quality so they can be tuned empirically in the field. A dashboard and an interactive heat map were created to track the changes in the audio quality with different parameters in multiple trials and to conclude the configuration that provides the best audio quality.

Scaling up to apply the change at the national level was a new challenge. Billions of records are collected each month, and crunching this data to generate the dashboard pages and heat maps required an efficient system to handle the huge volume of data. To tackle this problem, T-Mobile made a massive investment in computing power, capitalizing on state-of-the-science open source distributed computing technology. By combining the scalability of Hadoop services like the Hadoop distributed file system, Kudu, and Spark with powerful user-focused analytics engines like Alteryx, terabytes of data can be ingested on demand into visualization tools such as dashboards and maps in minutes.

You’ll discover the challenges faced in this large-scale crowdsourcing and data analytics project and how T-Mobile solved them to accomplish the original goal of improving the quality of experience for end users. In addition, you’ll find some unexpected benefits from big data analysis to highlight the unique advantages of big data.