Mitratech has cut the time needed to onboard a new product to its customer data-sharing platform from three to six weeks down to one to three days - roughly 90 percent faster - by rebuilding the service with dbt as its orchestration layer. Brian Kasen, director of data engineering and analytics, and data engineer Abhra Gupta presented the work at dbt Summit 2026.
The compliance software company has grown largely through acquisition, with three divisions spanning 20 to 30 products. Kasen's team joined via the 2023 Trakstar acquisition, and Mitratech has made eight or nine more since, each bringing its own tech stack and data sources. Requests to pull data from yet another SaaS product, usually via custom Python scripts hitting vendor APIs, kept piling up - each product with its own schema, tenancy model and often sensitive legal or HR data.
The first version of DataShare, built with partner phData for the Align governance and risk product, used Snowflake reader accounts for customers without their own Snowflake instance - what Mitratech calls "bring your own BI" - or direct Snowflake-to-Snowflake secure shares. It worked and turned the data team into an upsell revenue source, but it was built specifically for Align, so every new product meant starting over.
The rebuild made dbt the orchestrator rather than just the transformation layer. The DataShare package has no models, seeds or tests - just macros for provisioning, deprovisioning, metadata and logging that install and call Snowflake stored procedures through dbt run-operation. A five-phase pipeline builds schemas, control and log tables, secure views and share objects, applying tenant and customer filters, PII exclusions and encryption from configuration. Separating configuration from code let CloudOps staff, who aren't developers, onboard customers as a data-entry task.
CloudOps time per request fell 97 percent, from an estimated 26 to 40 hours to a couple of platform commands, and delivery dropped from three to seven business days to about one day, with a full audit trail captured automatically in Snowflake. The team's takeaway: design for generalization from the start, since the second customer request forces the refactor anyway, and treat dbt macros as infrastructure for the version control and idempotency they bring to Snowflake objects. They are also experimenting with AI tooling including Cognition's Devin and Claude Skills, with dbt Wizard and Snowflake CoCo next.
Source: dbt Labs Blog