Recently, I had the opportunity to join the EDM Council webinar, Semantic Layers in Data Strategy: Enabling Governance and Democratization, alongside industry experts, including Morgan Templar, Chief Executive Officer at First CDO Partners, and Leigh Pence, Data Governance Senior Tech Lead at Freddie Mac. Moderated by Jim Halcomb, Global Head of Research & Development at EDM Council, the discussion explored how semantic layers serve as the cornerstone of modern enterprise data strategies. We examined how organizations can balance governance with flexibility, enabling businesses to make data-driven decisions with consistency and accuracy.
My Journey to the Semantic Layer
My journey into the analytics space began in 2012 when I worked at Yahoo. Yahoo, one of the largest websites at the time, collected an enormous amount of data—far too large for traditional relational databases. To solve this, we pioneered the concept of a data lake, leveraging Hadoop to store and manage data at scale. While we did an excellent job capturing and storing data, making it actionable was an entirely different challenge.
Yahoo, like many enterprises today, had a fragmented analytics landscape. Different teams used different tools—Tableau, MicroStrategy, QlikView—and each business unit developed its own definitions of key metrics like impressions and clicks. This led to inconsistencies, making it impossible to compare performance across teams. We had to dedicate a team of seven people solely to data reconciliation—an inefficiency that I knew could be solved with a better approach.
The realization was clear: We needed a way to define business metrics once while still allowing different teams to use their preferred analytics tools. This was the foundation of the universal semantic layer—a structured approach to defining and governing business metrics at the data level while maintaining flexibility in consumption.
Semantic Layers as the Key to Data Democratization
As Morgan Templar noted during the webinar,
Data products are a way for the business to be able to access information. When we think about the data itself, we’ve got to have an underlying understanding of how it works, which is where semantic layering and knowledge graphs really provide an opportunity for the business to interact with data without needing to be technical.
One of the biggest challenges in data strategy is ensuring that business users have the freedom to analyze data without creating inconsistencies. The key is to extract semantics from analytics tools and centralize them in a governed layer that can be consumed across the organization. This approach ensures that:
- Business definitions are standardized and trusted.
- Data engineering is done once, eliminating redundant pipelines.
- Analysts can use the tools they prefer without breaking data governance rules.
This concept is particularly relevant today as enterprises shift from an application-centric to a data-centric approach. Instead of forcing all teams onto a single platform, organizations should focus on enabling a consistent data foundation that can be accessed universally.
The Role of Data Products and Marketplaces
Another key topic in our discussion was the evolution of data products and data marketplaces. Previously, data catalogs functioned more like glossaries, storing metadata without offering real operational value. Today, data marketplaces aim to make data accessible, allowing users to browse, access, and consume trusted datasets.
The combination of semantic layers, data products, and data marketplaces creates a powerful ecosystem where:
- Data is modeled for business context rather than raw storage.
- Users can shop for data in a governed way, similar to an Amazon-like experience.
- Organizations can scale analytics without losing control over definitions and access.
The Intersection of Semantic Layers & AI
As enterprises explore generative AI and natural language queries (NLQ), the importance of semantic layers becomes even more pronounced. Large Language Models (LLMs) are powerful but unreliable when querying enterprise data—our research shows that they can be wrong 80% of the time when applied to structured analytics.
To make AI work for enterprise analytics, businesses need:
- A strong semantic foundation to ensure AI understands business concepts accurately.
- Governed data definitions to prevent hallucinations and misinformation.
- Context-aware models that connect AI outputs to trusted business data.
By applying a retrieval-augmented generation (RAG) approach, organizations can significantly improve AI accuracy, ensuring that responses are syntactically correct and contextually valid.
Moving Forward: A Balanced Approach to Data Strategy
One of the key takeaways from this discussion was that governance and flexibility are not mutually exclusive. By investing in a semantic layer strategy, organizations can:
- Reduce data inconsistencies and improve trust in analytics.
- Scale self-service analytics without overwhelming IT teams.
- Enable AI-driven insights without compromising accuracy.
At AtScale, we are committed to enabling this vision. Our semantic layer platform provides the foundation for businesses to define, govern, and share trusted data across their organizations while maintaining flexibility in consumption. Whether you’re looking to improve analytics consistency, build a data marketplace, or leverage AI for decision-making, the semantic layer is the key to bridging the gap between governance and innovation.
To learn more about how AtScale can help your business scale analytics with a semantic layer, see it in action.
SHARE
WHITE PAPER