Semantic Views: A New Era For Snowflake And Sigma

This feature is currently in development and will be available in a beta capacity soon! Please reach out to your Sigma Account team to request early access.

—

Snowflake just introduced Semantic Views, representing a significant shift in the data landscape, and Sigma is proud to be a launch partner. With Semantic Views, metrics and logic can now all live in the warehouse, avoiding the need for redundancy and duplication across consumption tools. 

The age-old debate, solved at the source

At Sigma, we’ve long believed that the cloud data warehouse should be the source of truth for data and logic across your business. Your data warehouse is the main repository for storing, processing, and querying your data. It’s the central technology in your organization that feeds all of your various downstream consumers, including your BI tools, Notebooks, custom applications, machine-learning algorithms, AI-powered interfaces, and more. All of these downstream consumers need to speak the same language and work off shared definitions of key business metrics (e.g. “How is ARR calculated?). The warehouse is the most logical place to keep all these definitions. 

Historically, though, certain types of logic haven’t been supported directly in the warehouse, so it was forced into the consumption layers. This is a very slippery slope. As soon as any logic is codified into a consumption layer (e.g. your BI tool, Notebook, app, etc.), that logic immediately becomes siloed and inaccessible by your other consumption tools. Your BI tool might calculate ARR correctly, but none of your other LLM-powered chat interfaces, notebooks, or applications share that definition. They’ll all show different values for ARR. This leads to distrust, misalignment, and a general lack of trust in your orgs data.

If you want each of your consumers using the same definitions of key metrics, you need to duplicate that logic across each consumption tool. Every time you update that logic, you need to update it in each downstream tool. This human-driven process is cumbersome, error-prone, and inevitably leads to metric drift.

This is a long-standing, deeply entrenched problem. It’s given rise to an entire industry of tools aimed at addressing this central challenge. Data catalogs, wikis, and independent semantic layers have all built entire businesses specifically to solve this challenge. But third-party logic stores add another hop between users and data. Sometimes they make sense, but more often they increase complexity, introduce latency, and make your stack harder to manage.

Now, with the introduction of Snowflake Semantic Views, all of this logic can finally live directly in your warehouse. All of your semantics can finally live right next to the data, and can be accessed by all your consumers! No duplication. No divergence. No extra layers.

Semantic Views unlock a new level of consistency and governance for the data stack. Sigma brings that power to life by making warehouse-native metrics accessible to anyone, not just analysts. Together, we’re enabling organizations to define logic once and trust it everywhere, from dashboards to apps to AI.
_{- Josh Klahr, Snowflake Product Director}

Let’s look at a few examples to illustrate why Semantic Views are such a big deal, and how Sigma is integrating with them.

Why Semantic Views matter: two key examples

1. Aggregate-sensitive metrics

Databases have never supported any notion of “aggregate-sensitive” metrics. Certain types of metrics only make sense when calculated at very specific levels of aggregation. For example, say we have three marketing campaigns, and we want to understand the conversion rate for each of those campaigns, as well as our overall conversion rate. 
‍

‍

To calculate the conversion rate of each campaign, it makes sense to add an extra column. 

‍

Now, let’s say we want to calculate an aggregate conversion rate. It’s perfectly reasonable that an end user might see a “conversion_rate” column and simply calculate an average of that conversion_rate. I won’t bore you with the math, but a strict average is wrong by about 25%. Metrics like these, where the math changes depending on aggregation, haven’t traditionally fit inside the warehouse. Semantic Views finally make that possible.

2. Infrequently used or expensive metrics

Other metrics, like customer lifetime value (LTV), are just too expensive to calculate on a periodic basis in a transformation layer. Customer lifetime value is highly sensitive to long-term revenue projections, churn assumptions, margin estimates, and other inputs that are infrequently updated. This often requires stitching together historical revenue from an ERP, retention curves from your web or product data, CAC from marketing systems, and data from several other sources as well. Joining all this data together can be expensive. Because of the effort and complexity, most companies only update LTV quarterly or biannually. Encoding LTV into a column on a warehouse table on a regular basis might be more expensive than it’s worth to you as a business.

The coming AI wave

As cloud data warehouses evolve to support native AI and LLM workloads, like Snowflake Cortex, the case for keeping your semantics in the warehouse becomes even stronger. These AI models rely on clean, consistent, and well-defined semantic definitions, to generate accurate, trustworthy outputs. These models live and operate on the data in your warehouse. When core business logic and metrics are encoded directly in the warehouse, LLMs can reason over governed, production-grade data without duplication or latency. Your LLMs shouldn’t need to jump through a middle layer or rely on external semantic definitions in the consumption layer to produce reliable results. This ensures that all your analytics tools and your LLMs, and the tools that invoke them, are working off the same trusted definitions. Even if you don’t have many applications leveraging LLMs today, it’s quite obvious that you will soon enough. 

Snowflake Semantic Views + Sigma 

Semantic Views eliminate the need for duplicative logic in downstream tools, and Sigma is proud to be the first BI platform with a native, first-class integration. 

In Sigma, users can create a Semantic View using DDL and access Semantic Views just like they would access physical tables in the warehouse—through an intuitive, spreadsheet-style interface. Your end users never need to know whether data is coming from a physical table or the Semantic View, Sigma abstracts away all the complexity for you, and acts as an interface to your warehouse. 

All dimensions from your Semantic View map one-to-one with columns in Sigma, and all of your metrics in your Semantic View map one-to-one with metrics in Sigma. Sigma always queries your Semantic View directly, so you never need to worry about definitions getting out of sync. As soon as you update your logic in your Semantic View, that change is automatically reflected in Sigma as well. 

Once users start with a metric or a table, they can do what Sigma does best: slice, pivot, filter, build dashboards, or even build full data apps. It’s everything you love about Sigma, now powered by semantic logic that lives natively in your warehouse.

‍

‍

A new standard for metrics

Semantic Views represent a major milestone in the evolution of the analytics stack. For the first time, core metric logic can live exactly where it belongs: inside the warehouse, accessible to every tool, automatically in sync, and governed by design. With Sigma and Snowflake, your organization can finally align on the same language—across every metric, every tool, and every user.

If you’re rethinking your semantic model strategy, or migrating from legacy tools that lock your logic downstream, we’d love to show you how Sigma and Snowflake can take you farther together.