A Simpler Way to Build a Data Vault Using Templates and Code

If you’ve ever worked on a Data Vault project, you know how quickly things can spiral into complexity. In his talk, Viktor Hrtánek shared a refreshing approach that keeps things clean and manageable by treating Data Vault modeling more like software development than drag and drop configuration.

Here’s the short version of what he covered.

Start With the Target Model Not the Sources

Most projects begin by loading source systems and shaping the model around whatever those tables look like. Viktor flips that around. He suggests starting with the business concepts first and defining your hubs links and satellites based on the target model. Only then do you attach sources. This avoids drift unnecessary harmonization and messy logic sneaking into your model.

Why YAML Helps Keep Things Clean

Instead of GUI tools or proprietary metadata layers Viktor’s team uses YAML files to define Data Vault objects.

YAML is easy to read works beautifully with Git and makes it simple to version review and collaborate on models. Templates remain consistent so you are never guessing how something should be structured. You fill in the object type key settings and structure and you are done.

A Fast Workflow Validate Generate Deploy

Once the YAML templates are written the workflow moves quickly.

Validation checks for mistakes missing columns or Data Vault rule violations. Generation creates Snowflake Dynamic Tables from the templates. Deployment is either automated or handled via a simple command.

Changes like adding a new column or a new satellite often take minutes not hours.

Why Snowflake Dynamic Tables Make It Even Better

Because Dynamic Tables refresh themselves you do not need pipelines or orchestrators. You define how fresh the data should be and Snowflake handles the rest. In one real example a model of about two thousand objects went from a three hour refresh window down to minutes.

The Big Idea

Viktor’s approach boils down to a few principles.

Model your business first. Use simple readable templates. Version everything like code. Let automation generate the heavy lifting. Let Snowflake handle refresh logic.

The result is a cleaner faster and more maintainable Data Vault.