Overview
In this episode of The Dashboard Effect, Brick Thompson and Landon Oaks, Director of Solutions at Blue Margin, walk through the architecture that sits underneath effective business intelligence: the data lakehouse and the semantic model that connects it to the people who need to use it. The conversation moves from the technical layers of how data is stored and transformed all the way through to the business collaboration required to make those structures meaningful in practice.
Understanding the architecture is the first step. Building it well is another matter entirely. See how Blue Margin’s Managed Data Service handles the engineering, modeling, and collaboration work so your team can focus on using the data rather than building the infrastructure beneath it.
What This Episode Covers
The Bronze Layer (1:30)
The bronze layer stores raw data exactly as it arrives from the source. Transformation at this stage is minimal, limited primarily to format conversion, turning JSON into tabular structure, for example, to make the data usable for downstream processing. The goal is fidelity: preserve what came in so that every subsequent layer can be rebuilt from a reliable foundation.
The Silver Layer (2:12)
The silver layer handles pre-joins and light transformations. It is not always necessary, but it becomes valuable when working with large datasets that require combining tables before they reach the reporting layer. By doing that work upstream, the silver layer simplifies what the gold layer has to handle and improves performance where it matters most.
The Gold Layer (2:49)
The gold layer is where data arrives fully transformed and ready for reporting. Fact and dimension tables are structured and prepared at this stage, making the data accessible to BI tools without requiring additional transformation at query time. This is the layer that report writers and analysts interact with, directly or through a semantic model built on top of it.
The Role of Semantic Models (3:24 – 4:25)
Rather than connecting reports directly to the gold layer, organizations typically build a semantic model that sits between the data and the report writers. This model defines relationships between tables, implements star schemas, and encodes essential business logic like the specific calculation of total revenue or EBITDA. It is what allows different reports and different users to pull from the same definitions rather than each arriving at their own version of the same metric.
Discovery and Stakeholder Collaboration (7:37)
Building a semantic model that reflects how a business actually measures itself requires direct engagement with the people who own those definitions. Terms like revenue mean different things in different organizations, and the discovery process surfaces those distinctions before they become errors embedded in the model. Skipping that conversation produces technically correct logic built on the wrong assumptions.
Validation as an Iterative Process (8:41)
Getting the model right is rarely a single pass. Showing numbers to business users and asking whether they match expectations is one of the most effective ways to surface hidden nuances, catch incorrect prior calculations, and build the trust that makes a model useful rather than just technically complete.
Adaptability Over Time (9:52)
Business reporting needs evolve. An acquisition, a new PE firm, a change in how the organization is structured: any of these can require meaningful updates to how metrics are defined and calculated. Semantic models built with adaptability in mind can absorb those changes. Models built as one-time deliverables tend to become technical debt the moment the business changes around them.
Who It’s For
This episode is worth your time if you are a data engineer or architect building or evaluating a lakehouse implementation and want a clear explanation of how the medallion architecture is designed to function in practice, a BI developer trying to understand where semantic models fit and why they exist as a separate layer from the underlying data, a business stakeholder involved in defining metrics for a new reporting initiative who wants to understand what the technical team is building and why their input matters, or any organization preparing for a significant business change, like a PE acquisition, that will require its data infrastructure to adapt.
Why It’s Worth a Listen
The medallion architecture and semantic model concepts come up constantly in modern data conversations, but they are often explained in isolation from each other and from the business context that gives them purpose. This episode connects all three layers, technical architecture, semantic modeling, and business collaboration, into a coherent picture that is easier to act on than a purely technical description would be.
The emphasis on validation as an iterative process is particularly valuable. One of the most common failure modes in BI work is treating model building as a technical exercise and skipping the step where business users verify that the numbers match their understanding of reality. Landon and Brick make clear that this step is not optional and that the feedback it produces is often what separates a model that gets trusted from one that gets questioned every time it produces a number someone did not expect.
For teams building for the first time or rebuilding after a previous initiative fell short, this episode offers a grounded and practical framework for approaching the work in the right sequence.
