Guavus Blog

All the latest from the world of Guavus

Large North America Operator Uses Adaptive Learning, AI-based Analytics to Gain New Subscriber Insights

by Kanthi Vedaraman, Principal Technologist – Product Marketing at Guavus, a Thales company

Choosing a subscriber analytics solution is just a starting point for communications service providers (CSPs) looking to gain new insights into subscriber behavior and customer experience.

How do you measure its success and make it sticky?

  • It can’t be based on just insights, those insights should be addressing some or most of your business goals.
  • Almost all deployed analytics products need customization based on real-time customer data and business goals. In the telecom analytics domain, a deployed product giving results out of the box rarely meets CSPs’ business goals due to the variety and uniqueness of CSP goals. The products require customizations most of the time.

Real-time customizations based on continuous validations of results and feedback from customers is a crucial aspect in the success of any telecom subscriber-based analytics solution.

One of the largest North American telecom operators found this to be true. They have deployed Guavus’ Service-IQ analytics solution to provide marketing and general subscriber and service usage information.  Here’s their experience and journey with adaptive, continuous learning-based subscriber analytics with Service-IQ.

The Challenge

With close to 10 million subscribers, the North American telecom giant has been using the Service-IQ analytics solution to get insights into their subscribers’ usage and behavior patterns.

With the explosion of mobile applications, OTT services and Internet usage, the volume of subscriber data has grown exponentially. And, like other CSPs, this operator has been faced with opportunities as well as challenges in areas such as:

  • Correlating all disparate kinds of data sources with cardinality to improve Quality of Service (QoS).
  • Capitalizing on new opportunities like monetizing the data via various marketing campaigns.
  • Sharing/selling the non-Personally Identifiable information (PII) data with/to third-party advertising partners and OTT service providers.

To address these, the operator is using Service-IQ’s big data analytics to showcase insights on subscriber behavior by correlating all disparate data available and applying enrichment cardinalities on the correlated data.

Their foremost demand was to enable mobile application usage trend analysis within a micro-segmentation of subscribers, done in a low-latency manner without increasing  resource needs. This had to be enabled by smart tools and techniques as well as Artificial Intelligence/Machine Learning (AI/ML) models to mine the mobile application data usage across 10 million subscribers’ records flowing in as event data records (EDRs), deep packet inspection (DPI), and network equipment-specific telemetry data.

This requires that Service-IQ be readily configurable to focus on highly variable business needs. This was a big challenge to reliably identify various mobile applications and OTT services used by their customers, which is a crucial factor for customer retention and delivering best-in-class service.

Adapt on the Go

Service-IQ meets this challenge through two key classification features — Operator Data Services (ODS) Global Data Services (GDS). The microservices model enables complex and disparate data to be ingested, enriched, correlated and aggregated via common service-oriented architecture.

Nimble Operational Model

  • ODS and GDS are powerful Service-IQ features to quickly introduce new focused measurements in identifying customer behavior. Extensible enrichments specific to mobile application usage were made possible by applying autonomous customizations to Service-IQ’s ODS classification algorithms. Complex problems are indeed solved by the most intuitive and simple solutions.
  • These classification algorithms are dynamically plugged into ODS/GDS categorization services, which subsequently enabled data slicing and segmentation at macro and micro-level details.
  • The results got validated in real time. The categorization granularity improved by a significant percent – more than 50%.
  • Subsequently, there is a closed feedback loop system where these customizations are fed back into the AI/ML-based training module, which is integrated into Service-IQ’s categorization service resulting in a generic product update. This updated Service-IQ categorization service is serving not only this operator but also other telecom operators worldwide with these adapted enrichments.

Guavus Service-IQ – Adaptive Learning and New Focused Subscriber Insights:

 

Mobile Application Enrichment – BEFORE

Mobile Application Enrichment – AFTER

Operator Data Service (ODS) – BEFORE

Operator Data Service (ODS) – AFTER

Global Data Service (GDS) – BEFORE

Global Data Service (GDS) – AFTER

After Mobile App enrichment is applied, data slicing and categorization at this level (as depicted above) helped the operator to convert the data insights into monetization and gain a competitive advantage in their market.

What is at Stake for the Future?

With 5G playing out in the near future, CSPs are faced with the challenge of dealing with real-time understanding of data with speeds and volume significantly higher than 4G. For example, the 5G standard is designed to support up to 10 TB/s/km2 compared to 4G which only handles 1/100 of TB/s/Km2.

Adding to the speed and volume, in the world of 5G, IoT and now with a global pandemic, we’re seeing an even greater need for operators to take advantage of AI and advanced analytics to deal with increased network complexity, operational costs and subscriber demands for improved experience. To address these challenges, operators need to better understand network and subscriber behavior and be able to do so in real time.

Service-IQ has given this leading operator the confidence to embrace 5G with an open architecture and scalability as it’s based on an “ingest once, analyze many” platform powered by a streaming analytics engine.

Additional Resources:

As previously published in TelecomDrive.

Image attribution: iStock