Retail Analytics Warehouse

Problem

Raw retail data needs to be reliable before it can be trusted for decisions, but most projects at this scale skip data-quality testing entirely.

Solution

A dbt + DuckDB ELT pipeline with 35 automated data-quality tests and a Streamlit dashboard, so every number shown is backed by a passing test.

Overview

A dashboard is only as trustworthy as the data underneath it — and most personal-scale data projects skip testing that data entirely, so a broken pipeline can silently ship wrong numbers to a dashboard.

How it works

Retail data flows through a dbt + DuckDB ELT pipeline, with 35 automated data-quality tests (uniqueness, referential integrity, not-null, accepted values) running on every build. A Streamlit dashboard sits on top for exploration, so every chart is backed by a passing test suite, not just a hopeful assumption.

Result

A small-scale analytics stack that follows the same data-quality discipline as a production data engineering team, not just a one-off notebook.

Further reading

See Why Most Data Portfolios Get Rejected and dbt + DuckDB vs a Cloud Warehouse for more on why untested pipelines fail and why this stack was chosen.