A Telecom Player Boosted ARPU by 20% and Reduced Churn with an AI-Powered Insights Dashboard

Case Study | Data Analytics | Telecom

Problem Statement

A regional mobile service provider in Singapore was struggling to increase its average revenue per user (ARPU), which had stagnated at $35 per month, compared to an industry average of $45. The marketing and product teams relied on traditional metrics like call volume and data usage but overlooked other behavioral and network data, missing opportunities to upsell services and improve customer experience.

Solution

A mobile service provider implemented a data analytics solution with a hero dashboard that integrated previously unconsidered data sources, enabling the discovery of hidden revenue opportunities and better decision-making.

Technical Approach

1. Data Collection and ETL Process:
   • Extracted data from previously siloed sources, including:
     • Network performance logs (e.g., signal strength, dropped calls).
     • Customer support tickets
     • App usage data (e.g., time spent on streaming apps) collected via SDKs in the company’s mobile app.
     • Billing and subscription data from a database.
   • Built an ETL pipeline to ingest, clean, and transform data, including parsing unstructured log data using regular expressions and natural language processing (NLP) to analyze customer support ticket text.
2. Data Integration:
   • Stored processed data in a centralized data lake using Amazon S3, with metadata managed by AWS Glue for efficient querying.
   • Created a unified schema to join network, customer, and behavioral data, enabling cross-functional analysis.
   • Identified previously unconsidered data points, such as:
     • Network Quality Impact: Customers in areas with frequent dropped calls were 30% more likely to churn but also 25% more likely to upgrade to premium plans if offered better coverage.
     • App Behavior: Customers spending >5 hours/month on streaming apps were not subscribed to high-data plans, indicating an upsell opportunity.
     • Support Ticket Sentiment: Negative sentiment in support tickets correlated with a 40% higher churn risk but also a 20% higher likelihood of upgrading if offered loyalty discounts.
   • Used correlation analysis and feature importance techniques to validate the significance of these new data points.
3. Dashboard Development:
   • Built an interactive dashboard using Power BI, hosted on Azure, to visualize insights and enable decision-making. Key features included:
     • Network Quality Map: A geospatial visualization (using Power BI’s ArcGIS integration) showing areas with poor signal strength, overlaid with customer churn and ARPU data.
     • Behavioral Segmentation: A clustered bar chart showing customer segments based on app usage, data consumption, and plan type, highlighting upsell opportunities.
     • Sentiment Analysis: A word cloud and sentiment score trend line for customer support tickets, linked to churn risk scores.
     • Actionable Recommendations: Embedded rules (using DAX in Power BI) to recommend actions, such as offering premium plans to high-data users or loyalty discounts to at-risk customers.
   • Enabled drill-down capabilities to allow users to explore data at the individual customer level.
4. Deployment and Automation:
   • Deployed the dashboard as a web application on Azure App Service, integrated with Single Sign-On (SSO) for secure access.
   • Automated data refreshes using Azure Data Factory to ensure real-time insights with a latency of <5 minutes.

Tech Stack:

• Data Storage: Amazon S3, SQL Server
• Data Processing: Apache NiFi, AWS Glue
• Programming: Python
• Libraries: pandas, numpy, scikit-learn, nltk, seaborn
• Visualization: Power BI (dashboard), ArcGIS (geospatial)
• Deployment: Azure App Service, Azure Data Factory
• Integration: Zendesk API

Implementation:

• Conducted discovery sessions with marketing, product, and network operations teams on interpreting dashboard insights and acting on recommendations.
• Rolled out the solution across three regions over 5 months, starting with a pilot in the highest-revenue region.

Impact:

1. Efficiency: Reduced time spent on manual data analysis by 70%, as teams could now access integrated insights directly from the dashboard.
2. Productivity: Marketing campaign planning productivity increased by 40%, as the dashboard provided pre-segmented customer lists and actionable recommendations.
3. Decision-Making:
   • Uncovered Insights: The dashboard revealed that poor network quality was a key driver of churn, prompting investments in network infrastructure in high-value areas.
   • Data-Driven Actions: The leadership team used app usage data to launch targeted upsell campaigns, increasing data plan upgrades by 25%.
   • Proactive Retention: Sentiment analysis enabled proactive outreach to at-risk customers, reducing churn by 15%.
4. Overall Growth: ARPU increased by 20% to $42 per month, contributing to a 10% revenue growth within a year, driven by improved customer retention and upsell conversions.