Case Study | legacy-modernization new-business customer-engagement microservices fast-data akka scala

Real-Time Predictive Routing for Reducing Customer Churn

Use Case

  • Reducing customer churn with real-time insights from CDRs and EDRs

Challenge

  • Leverage perishable insights from Call Data Record (CDR) and Enhanced Data Record (EDR) metadata to improve call quality, find outages, and improve web browsing speed.
  • Customers generating 10 million records per second, representing 500 billion records per day, producing 600 terabytes of data per day.
  • Existing application architecture running on a traditional thousand-node Hadoop infrastructure delivered insights in days and hours; too late for the business to make adjustments to improve the customer experience.

Solution

  • Goal was reducing the wait time until data was useful to less than one second. Lightbend was brought in to help T-Mobile understand Reactive and implement a streaming data pipeline POC.
  • Hypothesis was to prove the application architecture could deliver insights on 32,000 records per second on 150 nodes. The outcome was a blazing 5 million records per second on 50 nodes.
  • Streaming data pipeline now in production and generating insights on 10 million records per second on reduced 5:1 hardware footprint. Current Hortonworks HDP data lake investments still highly utilized for data storage.

Benefits

  • Real-time call and browsing quality insights inform Predictive Dropping and automatic rerouting to towers with more capacity to improve customer experience, and ultimately, loyalty.
  • Real-time insights open up new monetization streams for Major Events Marketing informed by social sentiment around Superbowl, Oscars, and more.
  • Meeting real-time SLAs for First Responders and Nationwide 911 does good for the community and delivers millions of dollars in incremental revenue from federal government.

REAL-TIME FAST FACTS

 

BENEFITS

Real-time insights improve customer service delivery and loyalty

DATA VOLUME

600 terabytes/day
500 billions records/day
10 million records/second

BATCH PROCESSING INSIGHTS

Delivered in days

STREAM PROCESSING INSIGHTS

Delivered in real-time
10 million records/second
50 nodes

HARDWARE UTILIZATION

Reduced server requirements by 5:1

FIT FOR OTHER USE CASES

Call Centers, Field Service, Ad Targeting

 

Powered by Lightbend Platform: A brief look inside T-Mobile’s real-time streaming platform

Lightbend Platform provides a complete streaming architecture, including Reactive microservices frameworks, multiple streaming engines, and an end-to-end console for cluster management, monitoring, and security. To ensure developer happiness and reduce business risk, T-Mobile benefits from on-site and direct portal access to Lightbend experts and 24/7 production support.

Streaming Engines
Akka Streams
Apache Spark

Machine Learning
Spark MLlib

Data Backplane
Apache Kafka

Data Persistence
Apache Cassandra

Microservices Architecture
Lightbend Platform
(Akka, Lagom, Play, Scala)

Infrastructure Orchestration
Kubernetes
Mesosphere DC/OS

Lightbend Subscription
Developer Assist
Embedded Lightbend Engineer (ELE)
Production Support 24/7
Legal Indemnification & Warranties