If you have sat through a data platform pitch recently, you will have heard the words "Bronze, Silver and Gold." It is the architecture pattern behind almost every modern cloud data platform - and it is usually explained in engineering terms that leave decision-makers none the wiser about why it matters to the business.

It matters because data quality has become a board-level problem. IBM reported in 2026 that data quality was the single biggest data priority for a large share of operations leaders, and that concerns about data accuracy are now one of the leading barriers to scaling AI. Medallion architecture is, at heart, a disciplined way to fix that,  to turn messy source data into trustworthy, decision-ready information. Here is what it is, in plain English, and what you should take from it as a leader.

What medallion architecture actually is

Think of it as a quality assembly line for data, with three stages. Raw materials come in at one end; finished, reliable products come out the other. The three stages are named Bronze, Silver and Gold, and data moves through them gaining quality and business meaning at each step.

The pattern originated at Databricks around 2019–2020 and has since become a widely adopted standard across modern platforms, including Microsoft Fabric. A useful clarification: medallion is an organising pattern that runs on top of a "lakehouse" (the underlying storage-and-compute foundation). The lakehouse is the factory; medallion is how you organise the production line inside it.

The three layers, in business terms

Bronze - the raw, untouched record. This is data exactly as it arrived from each source system, kept immutable with a full audit trail. Nothing is changed or cleaned here. Its job is to be the trustworthy "system of record" you can always go back to - essential for audit, replay and regulatory defensibility.

Silver - cleaned, conformed and integrated. Here data is validated, de-duplicated, standardised and joined across sources, so that "customer" or "student" or "asset" means the same thing regardless of which system it came from. Silver is where data quality rules are applied and history is preserved. It is the trustworthy, general-purpose middle of the platform.

Gold - business-ready and certified. Gold tables are purpose-built for specific uses: a finance dashboard, a regulatory return, a forecasting model. They are pre-shaped and aggregated so that reports run fast and answers are consistent. Because the heavy lifting is done in advance, end users get near-instant answers to questions that would otherwise take minutes to compute. Gold is what most people in the organisation actually touch.

The flow is always the same: raw ingestion → Bronze → Silver → Gold → consumption.

Why this should matter to you as a leader

You do not need to implement medallion architecture - but you should understand why it earns its place, because it directly affects five things you care about.

• One version of the truth. Conforming data in Silver and certifying it in Gold is what stops five teams arriving at five different numbers in the same meeting. That single benefit often justifies the whole approach.

• Auditability and compliance. Because data is layered and its journey is tracked, you can trace any figure on a dashboard back to the exact source record that produced it. For UK GDPR, statutory reporting and external audit, that lineage is not a nice-to-have.

• AI-readiness. AI is only as trustworthy as the data beneath it, and data-quality concerns are a top barrier to scaling AI. Medallion produces the clean, conformed, well-described data that AI and natural-language querying need to give answers you can rely on.

• Cost control by layer. Separating raw, cleaned and curated data lets you store and process each appropriately - keeping cheap raw history available without paying premium compute to query it.

• Faster, safer self-service. With certified Gold data and a semantic layer on top, business teams can answer their own questions without going off-piste - governed self-service rather than a free-for-all.

Is medallion architecture still relevant in 2026?

It is a fair question, and one the data community is actively debating. The honest answer: yes, it remains the dominant standard - but it is a pattern, not a magic wand, and treating it as one is where organisations get burned.

Two cautions worth carrying into any conversation with your delivery team. First, medallion does not automatically make a platform fast or cheap; badly designed Bronze, Silver and Gold tables can be just as slow and expensive as the legacy estate you are leaving. Second, persisting every dataset at every layer carries real storage and processing cost - it needs to be a deliberate decision, not a default. The pattern delivers when it is designed with intent, not adopted as a label.

What to ask your team

You do not need to design the layers, but a few pointed questions will tell you whether they are being designed well:

• How is the Silver layer organised? The most common and costly mistake is mirroring source systems in Silver rather than conforming around business entities. You want the latter.

• Are Gold tables certified and owned? Each should have a named owner, a clear definition and a quality status - ideally linked to a business glossary.

• Is there a semantic layer on top of Gold? This is what adds business meaning and makes the data usable by people, tools and AI. Gold without a semantic layer is half a solution.

• How is lineage captured? You should be able to trace any Gold metric back through Silver to its Bronze source.

• How is the cost of keeping every layer controlled? There should be a clear answer, not a shrug.

Beyond Gold: where the thinking is heading

The pattern is also evolving. The leading edge of the conversation is about what sits beyond Gold - richer semantic layers, "data-as-product" thinking where each dataset is treated as a product with consumers and service levels, and emerging ideas around ontologies and knowledge graphs that capture the relationships between entities, not just the entities themselves. You do not need to act on these yet, but it is worth knowing your platform foundation should be able to grow into them.

For most organisations, though, the message is simpler: a well-designed Bronze, Silver and Gold structure, with a semantic layer and proper governance, is the difference between a data platform people trust and one they quietly work around.

Designing your platform foundation? Talk to us about getting the Bronze/Silver/Gold structure and semantic layer right from the start.