Case Study

How a telecom giant fast-tracked their data lake implementation

Modern machine learning implementation on the data lake led to 6% increase in customer acquisition


Challenge

One of the top 3 US-based telecom providers wanted to migrate from Oracle DWH to a Cloudera Hadoop-based data lake to take advantage of advanced analytics. A scalable data storage solution was needed to reduce the cost of data management and analysis, effectively utilize their marketing budget, and increase business revenue. The telecom provider also wanted better insights to grow their B2B lead generation and improve the effectiveness of marketing campaigns and channels.

 

Discovery

The existing analysis mechanism was not able to increase net new and upgraded conversions. They were looking for a partner to analyze their customer journey and set up a modern CDH Hadoop-based data lake.

 

Impact

Impetus Technologies Inc. analyzed the existing data from RDBMS and external systems like Salesforce and Marketo and created a Hadoop data lake on the Cloudera (CDH) cluster for lead generation. Data from all these datasets were mined and mapped to enhance lead generation and conversion.

The solution involved implementing cross-sell/up-sell, continuum, lead-generation, and multi touch-point attribution models to enhance marketing campaign effectiveness. We also applied audit-trail policies to monitor and handle failovers. The solution had the following impact:

  • Increased revenue by lead generation and conversion
  • Effectively utilized campaign budgets
  • Reduced cost of data storage
  • Increased capability to store and analyze data

 

Functional aspects:

  • Helped in coming up with a Customer 360 journey
  • Migration of data and ingestion pipeline
  • Modernized the existing lead generation models

 

Operational aspects:

  • Security: Used Kerberos on CDH cluster for authentication and authorization
  • Monitoring: Added email notifications to all the ML models to monitor data ingestion workflows for failovers
  • Reliability: Tested all ML models and ingestion workflows
  • Ease of maintenance: Ability to optimize the solution by tuning the hyperparameters in the property files, without any change in code