Medallion Architecture in Microsoft Fabric: A Contractor's Guide
Updated: 3 hours ago
If you are moving construction data into Microsoft Fabric, one of the first design questions is how to organise it. Load everything into one place and you quickly end up with a data swamp: tables nobody trusts, logic hidden in reports and no clear path from source to dashboard.
Medallion architecture is Microsoft's answer. This guide explains what medallion architecture in Microsoft Fabric is, how the bronze, silver and gold layers work using real construction examples, and how to implement it without overcomplicating things.
What is medallion architecture?
Medallion architecture is a design pattern for organising data in a lakehouse in three stages, with each stage making the data more reliable and easier to use. Microsoft describes it as the recommended design approach for Fabric, implemented as lakehouses in OneLake, Fabric's single data lake.
The idea is simple: data flows from raw, to cleaned, to business-ready. Each layer has a clear job, so when a number looks wrong you can trace it back step by step.
The three layers, with a construction example
Imagine a contractor running Viewpoint Vista for finance, a project management system for site data and spreadsheets for forecasts. Here is how each layer would handle that data.
Bronze: raw data, exactly as it arrives
The bronze layer stores source data as it arrives, with no changes. Microsoft recommends keeping it in its original format where possible and using shortcuts rather than copies when the source already sits in OneLake or supported cloud storage.
Vista job cost, AP, AR and payroll tables, landed on a schedule.
Project management exports such as RFIs, change events and daily logs.
Forecast spreadsheets and budget files, kept as received.
Bronze is your safety net. If logic changes later, you can rebuild everything downstream without going back to the source systems.
Silver: cleaned and standardised
The silver layer is where data is fixed and conformed: errors corrected, formats standardised and duplicates removed. For construction, that typically means:
One consistent job number across Vista, project management and spreadsheets.
Cost codes and cost types mapped to a single agreed structure.
Dates, companies and currencies standardised.
Basic data quality checks, such as jobs without a project manager or costs posted to closed jobs.
Gold: business-ready for reporting and AI
The gold layer is organised for reports, dashboards and analysis. This is where business definitions live, usually as a star schema of facts and dimensions:
Job cost, billing and WIP snapshot fact tables.
Dimensions for job, customer, project manager, cost type and period.
Agreed calculations for committed cost, projected margin, over/under billings and margin fade.
Power BI semantic models are then built on gold, so every report uses the same numbers.
Why medallion architecture suits construction data
Traceability. Every gold figure can be traced back through silver to the raw bronze record, which helps when finance asks why a number changed.
Multiple systems, one view. Silver is where Vista, project management and spreadsheet data are aligned on the same jobs and codes.
One definition of WIP and margin. Business logic sits once in gold, not in dozens of report files.
History. Monthly snapshots in gold show how each job's forecast has moved over time.
AI readiness. Clean, well-described gold tables are the foundation AI tools need to give reliable answers. Read more about AI-ready data.
How to implement medallion architecture in Microsoft Fabric
Microsoft's guidance gives a clear starting framework:
Separate lakehouse per layer. Keep bronze, silver and gold apart. Microsoft also recommends a separate workspace for each layer for more control and better governance.
Delta tables for silver and gold. Fabric standardises on the Delta Lake format, which brings performance benefits across Power BI, SQL and Spark.
Lakehouse or warehouse for gold. Gold can be a lakehouse or a data warehouse; Microsoft says the choice should depend on your team's preferences and skills.
Choose your transformation tools. Pipelines, Dataflows Gen2, Spark notebooks and materialized lake views (declarative SQL that manages transformations between layers) can all move data from one layer to the next.
Build Power BI on gold. Semantic models and reports should read only from the curated layer.
For more on the platform itself, see our Microsoft Fabric consulting page.
Common mistakes to avoid
Skipping bronze. Without raw history, you cannot reprocess data when definitions change.
Transforming in Power BI. Logic hidden in report files undermines the whole point of a governed gold layer.
Building gold before agreeing definitions. Sit down with finance and operations to agree what margin, WIP and committed cost mean first.
Trying to land everything at once. Start with the data behind the reports that matter most, then extend.
No quality checks between layers. Simple checks in silver catch problems before they reach a board pack.
No clear ownership. Each layer, and especially each gold definition, needs a named owner.
Where to start
A practical first phase for a contractor might look like this:
Pick one high-value report, such as monthly WIP or job cost by project manager.
Land the relevant Vista tables in bronze (see our guide on connecting Viewpoint Vista to Power BI for the connection options).
Clean and conform them in silver.
Model a small gold star schema with agreed definitions.
Build one Power BI report on it, validate the numbers with finance, then extend to the next report.
Key takeaways
Medallion architecture in Microsoft Fabric organises data into bronze (raw), silver (cleaned) and gold (business-ready) layers.
It gives construction firms traceability, one definition of WIP and margin, and a foundation for AI.
Keep layers separate, use Delta tables, and build Power BI only on gold.
Start small with one valuable report and grow from there.
Frequently asked questions
Is medallion architecture required in Microsoft Fabric?
No, but Microsoft describes it as the recommended design approach for Fabric. It is a pattern, not a product feature, so you can adapt it to your needs.
Should the gold layer be a lakehouse or a warehouse?
Either works. Microsoft presents both patterns and suggests choosing based on your team's preferences and skills. SQL-focused teams often prefer a warehouse for gold.
Do we need three full copies of our data?
Each layer holds a different shape of the data rather than a straight copy. In bronze, Microsoft recommends shortcuts instead of copies when the source already sits in OneLake or supported cloud storage.
How does this relate to Power BI?
Power BI semantic models sit on top of the gold layer. That keeps report logic simple and makes sure every report uses the same governed numbers.
The gold layer is also what makes AI practical. See AI data readiness for construction, or explore our Vista ERP analytics service.
Next step: check how ready your data is
Want to know where your business stands? Take our free 3-minute Construction Data Readiness Check. You will get a score out of 100 across reporting, data foundation, platform and AI readiness, plus practical next steps. Prefer to talk it through? Book a discovery call and we will look at your setup together.
Comments