Blog

Why We Built Count Metrics: A New Approach to the Semantic Layer

No items found.
Oliver Pike
March 19, 2025
March 19, 2025
min read
Share this post
Contributors
No items found.
Subscribe to newsletter

Subscribe to receive the latest blog posts to your inbox every week.

By subscribing you agree to our Privacy Policy.
Thank you! Your submission has been received!
Oops! Something went wrong while submitting the form.
No items found.

At Count, we're known for empowering users with intuitive analytics and lightning-fast exploratory workflows. Yet we kept encountering a persistent challenge: how to maintain consistency, reusability, and governance around core metric definitions without sacrificing the flexibility our users love. While semantic layers aren't a novel concept, we envisioned something different—a solution that would seamlessly integrate with Count's flexible exploratory style while delivering uncompromised performance and accessibility.

Enter Count Metrics—our vision of what a semantic layer could do as well as should do.

We could have integrated with established players like Looker, Cube, or dbt Semantic Layer. Instead, we chose to craft our own solution. Here's why:

DuckDB: Performance Without the Price Tag

Users consistently tell us about the same pain point with existing semantic layers: skyrocketing compute costs from data warehouse queries. Count Metrics tackles this head-on by embracing DuckDB, an OLAP engine purpose-built for analytics:

For quite some time we've leveraged DuckDB instances in your browser to keep canvases responsive, eliminate workflow disruptions from query delays, and thoughtfully manage compute resources and costs.

Our recent introduction of DuckDB on the server takes this further, automatically distributing queries across browser, server, and data warehouse based on result complexity. Whenever possible, we run queries in-browser using cached data within Count, dramatically cutting both costs and latency, but we are now able to utilise server-side instances at scale to widen the reach of this performance advance.

By building upon this model with Count Metrics, we are able to cache data in more complete ways than existing semantic layers. Rather than accelerating individual or common queries, we can cache the entire data model, allowing for all downstream queries to run outside of your data warehouse.

Seamless Experience. No Compromises

By developing Count Metrics in-house rather than integrating with third-party solutions, we've created an experience without seams or friction. We believe this unified experience is essential for making data the natural language of organizational improvement—people simply won't embrace tools that fight against them.

While other semantic layers often sacrifice flexibility for governance, we refused that trade-off. Metrics and dimensions defined within Count Metrics can be used seamlessly within Count’s visualization and low-code features, and then later referenced by any cell, including SQL and Python cells.

Our approach enables users to transition effortlessly from standardized reports into free-form exploration—focusing on answering "why" rather than merely describing "what."

Trust Through Direct Connection

If a semantic layer's ultimate purpose is to make metrics trustworthy and governable, then fewer layers of abstraction between it and your BI tool is better. Count Metrics brings first-class semantic analysis capabilities directly into Count, automatically detecting references, flagging errors, and alerting users to inconsistencies across reports.

To widen the user base that can benefit from this going forward, our future plan is to directly ingest dbt model definitions into Count Metrics, giving teams end-to-end control of their data pipelines from raw data to final visualizations—all in one place.

Forward-thinking: Metric Relationships & Rich Visual Models

We believe that metrics never exist in isolation. They describe an interconnected web representing complex business logic and processes within organizations. That is why the canvas is so effective at giving organisations clarity over not just their data, but what it means in terms of what they do.

We're designing Count Metrics to embrace this reality. Upcoming features will allow metrics relationships ("metric trees") to be explicitly encoded directly in the semantic layer, enabling business stakeholders and analysts to create sophisticated visualizations like process flow maps and workflow diagrams effortlessly and clearly.

Count Metrics delivers the best of both worlds—standardized, reusable metrics combined with powerful ad-hoc exploration capabilities to seamlessly integrate consistent governance with agile analytics workflows. We’re thrilled to continue empowering analysts, teams, and entire organizations to uncover deeper insights faster, smarter, and more cost-effectively than ever before.