Enterprise Data Modeling for AI-powered telecom

Mobile finally has a sexy side again as automation and AI grab headlines for the potential to supercharge operations and drive revenues.?

But these buzzworthy developments are completely dependent on unified quality data, a rare commodity within most tier-one telco organizations.?

This is driving the need for Enterprise Data Modeling (EDM) strategies built for the realities of today's mobile operators.

Company mergers, system consolidations, decades of upgrades, siloed technical teams and varying regional approaches are just some of the contributors to fragmented data handling. Network rollout data, network resource performance and customer service data all represent varying dataset types that must be holistically approached for successful telco evolutions.?

Just because data is a product of the current state of technology doesn't mean it is prepped for data structures required to support ML/AI and automation use cases. Even the latest sophisticated systems, databases and equipment powered by new technology are typically challenged when it comes to dataset compatibility and overall data normalization. This data is sometimes vendor-specific, difficult to integrate and stored in varying formats.?

These realities are driving brownfield operators to step back and consider enterprise data modeling as an urgent strategic initiative.?

One transformation driver that has become common to every telecom operator is the evolution of site management–the process of planning, building and maintaining networks, and managing the field force executing them. Getting these deployments right means the difference between network rollouts that take months versus years.?

According to Igal Elbaz, SVP, Network CTO, AT&T, the company was able to achieve progress in 18 months that would normally take three to five years, with Rakuten Symphony’s Site Manager solution. The deployment and construction tool is being used in all AT&T’s markets with upwards of 7,000 users across its internal and contractor teams.

Let’s explore why telecom organizations must pursue new data strategies, considerations for various approaches and what to expect along the way.?

Business drivers for Enterprise Data Modeling

Harnessing the full potential of AI and fostering a roadmap to network autonomy requires a solid data strategy.?

From a business perspective, a strategic approach to data can significantly reduce operational costs and lower TCO by enhancing operational efficiency through smart system integration and umbrella tooling systems. The result is a foundation for data convergence, not just in format but also periodicity, with data constantly fine-tuned for the applicable use case.?

This approach manages telecom data effectively while contributing to growth by accelerating time-to-market for new services. Importantly, it paves the way for informed decision-making by consolidating all data in one place for comprehensive visibility, supporting the future readiness of network technologies and optimizing customer experience through insightful business intelligence, bringing things closer to near-real time, proactive, automated and AI-supported decision making.

On the technical front, a robust data strategy enables the existence of a diverse application ecosystem, including traditional and next-generation applications (e.g., r-apps, x-apps in the RAN) built on industry-standard interface principles like 3GPP NEF, TMF, ETSI-MANO and SMO, enabling a common strategic EDM wrapper. It provides a foundation for tackling future challenges and rolling out novel network solutions, improving automation for critical, time-sensitive operations.?

The implementation of AI is another cornerstone, requiring a framework that not only empowers discovery and exploration of data but also ensures its accessibility for application developers and data scientists, enabling a broad spectrum of AI use cases. Finally, robust data governance and security measures are imperative for managing the abundance of data while continuing to add value.

Securing the correct foundation for automation, AI and autonomy

Telcos may identify varying business drivers for kicking off data modeling and consolidation efforts but the destination is always the same.?

An operator with 75 million subscribers in a low-ARPU country can’t simply introduce new services to increase profitability. Instead, they must lower costs.?

Meanwhile, an operator in a small country that enjoys high ARPU and aims at perfect service quality with great signal coverage across the territory, might be bleeding money on expensive subcontractor fees during network rollouts.?

In these scenarios, data and inventory consolidation, and improved asset management can deliver end-to-end service assurance and decreased TCO on umbrella OSS systems.?

While the entry points on the path to data consolidation are different, digital transformation strategies will typically address the EDM across business and technology strategy, phased migration to the new systems and AI that runs atop of the data.?

1. Business and technology strategy. From the outset, it's essential to establish a C-level collaboration to align the data transformation with the company's broader business goals and enterprise architecture strategy. Establishing strategic priorities for the organization's data and AI platforms involves assessing new technologies and system integrators, ensuring there's a clear backlog of valid data to be integrated, and applying a prioritization framework, such as the WSJ (Weighted Shortest Job First) method, to manage the requirements effectively.
2. Phased migration to the new systems that have to address the data model. This step requires the orchestration of a centralized transformation project team, which includes data engineers tasked with constructing data pipelines essential for the migration. To ensure the process is seamless, data governance procedures must be established, potentially through a dedicated AI/automation office. Keeping track of the data through catalogs, supporting data and AI operations, and creating a data marketplace are also crucial. This marketplace will facilitate third-party rApp and NWDAF (Network Data Analytics Function) developers by providing them with the necessary tools and data accessibility.
3. Enable AI on top of the data. This final step involves vectorizing datasets for advanced AI search and Large Language Model (LLM)-based inferences, establishing rigorous data governance protocols for data access, usage, and transfer, and identifying use cases that add business value. It's also about providing APIs to meet various AI requirements and outlining a user experience (UX) and processes that allow for the correct interpretation of AI insights.

The focus of the data transformation strategy will depend on the strategic business goals of the implementing operator, be it rapid rollout acceleration, zero-touch provisioning, end-to-end performance management and customer experience management.?

An effective data transformation strategy isn't just about adopting new technologies, but migrating the existing data infrastructure to be able to support company automation, AI and autonomy goals, ensuring every step—from planning to execution—aligns with the company's innovation trajectory and operational efficiency goals.

Diving into data lakes

Rakuten Mobile understood the future business needs of a telecom operator when it built a single data lake to power IT and Network operations in Japan. Starting from scratch, this was a luxury and stands as a proven testament to the efficiency, performance and rapid evolution outcomes that accompany a consistent enterprise data model.?

Brownfields have it harder but the opportunities are identical. A phased business approach to data normalization, consistency and quality lays the foundation upon which all future automation, AI and autonomy initiatives will be built. It is the most important transformation strategy on which to focus because it is fundamental to all others.?

Want to learn more about Rakuten Symphony's enterprise data model for a fully cloud-native Open RAN network? Mention the authors Anshul Bhatt , Perivoye Stoyanovski and Geoff Hollingworth in the comments to start a conversation.