Supply Chain Analytics

Why I built it

Inventory questions are rarely about inventory. They are about cash sitting still. This project follows that thread from raw movement data all the way to a briefing a supply chain lead can act on.

Questions it answers

• How much working capital is trapped in stock that is not turning?
• Which items and locations are driving that, and how far off target are they?
• What should the operations team change first, and what is it worth?

How I built it

1. Lakehouse on DuckDB. Raw files land as Parquet and move through bronze, silver and gold entirely in DuckDB — no cluster, no cloud bill, same architecture.
2. Kimball star schema. Conformed dimensions and a declared fact grain, so measures aggregate correctly at every level instead of only at the one they were written for.
3. 101 DAX measures. A full measure layer covering stock coverage, turns, exceptions and variance, organised so the report is filter-safe.
4. Executive briefing. The findings written up as a client-facing case study PDF and a briefing deck, not a screenshot dump.

Method and validation

• Estimated trapped capital of roughly $95K — calculated as on-hand quantity × unit cost for stock exceeding the aging threshold, then attributed to the specific items and locations driving it rather than reported as a single total.
• Measures are written against the star schema, so slicing by any dimension gives a defensible number.
• The full build — pipeline, model, dashboard and briefing — is documented in the repo.