Today, we’re excited to launch the public beta of Semantic Model Sync for a faster, easier self-serve experience in Hex! You can connect semantic models from dbt MetricFlow, Cube, and LookML to use pre-defined measures, dimensions, joins, and more.
Data teams are dealing with tough tradeoffs. They want to keep up with the infinite demand for insights in their organization, while making sure results are generated in an accurate and consistent way. They want flexible tools that don’t lock them into a vendor-specific spec, but also want a way to create governed metrics without having to redefine them every time.
This is a tricky situation — and it’s why we’ve taken a novel approach when it comes to semantic modeling in Hex, with a focus on flexibility, interoperability, and openness.
As a next step, we’re introducing Semantic Model Sync, which lets you leverage trusted metrics and business logic from dbt MetricFlow, Cube, and LookML. Measures, dimensions, and joins from your imported semantic models show up in Hex’s Explore experience, allowing stakeholders to self-serve governed concepts in an intuitive way.
Semantic Model Sync is available on Team and Enterprise plans.
The fragmented state of semantic modeling
If you’re not familiar, a semantic model is basically “instructions for use” for a given set of tables. How should you calculate measures based on the columns? What are valid joins and aggregations? Semantic models encode this up front, so everyone knows how to use the data correctly downstream.
This is an immensely useful concept especially for “self-serve” analytics, where you want a broader set of users in the organization to leverage data in a consistent, standardized way.
The current state of semantic modeling in the data stack, however, is a little weird.
On one hand, you have fully-laminated, locked-in modeling solutions integrated into BI tools like Looker. These are popular and effective, but leave you stuck defining vendor-specific logic that’s hard to reuse or repurpose. We knew early on that the wrong approach would be forcing everyone to re-define their logic in a closed, proprietary “HexML” format.
On the other hand, there are standalone modeling solutions like dbt MetricFlow or Cube. These are both powerful and expressive specs that are still developing compatibility ecosystems — and not every team wants to pay for another cloud service to host them.
What we’ve seen and heard is that data teams are stuck between a rock and a hard place, trying to figure out how to bring semantic modeling into their organizations without getting locked into a path they’ll regret.
Enter Semantic Model Sync!
Our interoperable approach to semantic models in Hex
Today, we’re allowing data teams to have the best of both worlds: an integrated, first-class experience for leveraging semantic models, without compromising the flexibility of ad hoc analysis or getting locked into a proprietary spec.
Semantic models in Hex persistently sync datasets, measures, dimensions, and joins from popular third-party semantic models. This allows users to do analyses based on these governed concepts in Explore.
Check out our guide on data curation for a better self-serve experience in Explore.
Now, data teams can model what’s needed without getting boxed in. When you want to go off-road, the Hex notebook is built for agile exploratory data analysis. No need to submit a pull request to answer a one-off question or crank out LookML code for single-use logic.
Getting started with Semantic Model Sync
In three steps, Admins can set up a sync via GitHub, and Hex will directly parse the underlying semantic code without needing to go through a third-party API.
Add a new model on the Settings page by specifying the data connection and generating a new API token.
In your GitHub repository where your semantic model files live, add the API token, create a new folder, and paste in the script provided by Hex.
Every time you merge new code, it will automatically trigger a GitHub action that sends the contents of your repository to Hex. (Behind the scenes, Hex reads your files and translates the specs into measures, dimensions, and joins.)
Once the integration is complete, you can browse the datasets in your model and quickly jump into Explore to visually analyze your data using measures and dimensions. Hex discovers joinable datasets automatically, and we also built error-free, fan-out tolerant aggregations, so you do less modeling work and get consistently accurate results.
Traditionally, measures, dimensions, and joins have only been accessible to BI users, but these pre-defined units of logic can also cut repetitive work for data teams. With semantic models in Hex, you can use semantically-enriched Explore cells in notebooks, and chain them into SQL and Python cells.
Want to dive in today? Get started with our Semantic Model Sync documentation.
The Looker integration is in private beta. Contact your rep or request a demo for access.
Towards a more unified world for data teams
At Hex, we’re on a mission to bring together fragmented data workflows. To us, this means helping organizations consolidate tooling, but it also means interoperability and allowing you to leverage your semantic models no matter where they’re defined.
We believe the biggest challenges facing data teams today can’t be fixed by superficial changes in UI or UX. Our platform’s unique atomic building blocks — cells, dataframes, and DAGs — imbue flexibility and modularity into everything we build, including our semantic models.
Whether you’re looking for an all-in-one platform to combat tool sprawl or a powerful data science notebook that integrates with your semantic layer, Hex is built differently, so you can work differently.