United States – – October 27, 2026 -- Brilliant Earth has moved more than 100 showroom operations users off a legacy BI platform and onto Posit dashboards running natively inside Snowflake, cutting software license costs while widening self-serve access to governed data.
The luxury jewelry retailer already reprices its diamond inventory daily through an automated model running in Snowflake. That capability prompted the company to adopt the platform. Weekly data science reports captured only moments in time, so Brilliant Earth set out to give decision-makers direct access to the same data and models.
Posit Team runs as a Snowflake Native App on Snowpark Container Services
Brilliant Earth installed Posit Team inside Snowflake instead of maintaining a separate server with its own security protocols. Snowflake roles, user permissions and table access now govern Posit as well.
Provisioning is automated. An employee requests access through a tile in the IT ticketing system. After the requester's manager and the analytics team approve, the system assigns the user to the correct group at the single sign-on provider. That assignment triggers the matching Snowflake roles and Posit Connect account settings.
Matthias Mueller, Senior Director of Analytics and Data Science, previously deployed Posit on stand-alone servers at other companies. He said he has not regretted the native-app architecture and would use it again.
Growth team replaces static decks with an interactive marketing mix model
Brilliant Earth's proprietary Bayesian marketing mix model was previously used mostly by data scientists. Business leaders received summary statistics and PowerPoint decks. The team now serves the model through a Streamlit application hosted on Posit Connect and grounded in Snowflake data.
Growth team members can explore response curves, evaluate channel performance and set budgets. They can also apply constraints such as channel saturation points and run scenario simulations with projected revenue and ROI, all without writing code.
Showroom managers query dashboards in natural language
The analytics team wrapped the new Posit dashboards in querychat, a governed conversational interface. A showroom manager can request the last 14 days of bookings performance for a single location and receive a customized view.
LLM writes weekly reports only after code validates the numbers
Python and R code compute metrics directly against Snowflake data. Validated figures are then passed to an LLM API, which summarizes trends in prose. Mueller said mathematical functions must own the computation because code-based output can be verified, repeated and maintained.
Domain experts prototype reports with Claude and Snowflake CoWork agents
Merchandisers connect frontier models such as Claude to Snowflake CoWork agents built on governed semantic layers. They prototype reports on metrics such as carat weights and sell-through rates. If a prototype proves broadly valuable, the analytics team rebuilds it on validated Snowflake data and publishes it to Posit Connect.
Merchandising, digital and finance teams are next to onboard. Mueller said he wants to build custom solutions with stakeholders rather than keep adding dashboards.
Sem Sergunin, Retail and Consumer Goods Industry Lead at Snowflake, warned that deploying AI agents on unprepared data will "scale your mess."