Interactive report created in Power BI
US energy, mining and banking data
One method, three different tools.
Interactive dashboards built from raw US federal files. Each one is authored in a different tool, rebuilt from source on every run, and its method is written up alongside it.
Power BI
Excel
Python
DAX
SQL
Data Modeling
Public sources
USGS, FDIC and EIA. US federal data, public domain, no private extracts.
Reproducible
Every pipeline rebuilds from the raw files. Nothing is cleaned by hand.
Checked
Each build validates its own numbers against the source before it ships.
Dashboards made with different tools
Power BI
STATES50
TOTAL VALUE$105B
Python
LARGEST BANKS20
CORE KPIS6
Excel
TEXAS SHARE42%
SINCE2000
Selected work
Three dashboards, one reading room.
Interactive report created in Python
Bank Health Scorecard
Interactive report created in Excel
Oil & Gas Production by State
The pipeline is part of the work.
Each dashboard rebuilds itself from the raw public files, and refuses to build if its own checks fail. The full method for each one is written up on its own page.
01
Data as code
Scripts turn the published files into tidy tables - mixed encodings, three-row headers, withheld and range codes and all. No step is manual, so a refresh cannot silently skip one.
02
Checks that stop the build
Every pipeline recomputes known figures and compares them to the source. If an anchor fails, the build exits without writing anything, so a wrong number cannot reach a chart.
03
Dashboard as code
The Power BI report is generated, not dragged together: Python writes the semantic model and the report definition, a validator checks them, and a drift guard refuses to overwrite hand edits it did not expect.
04
Designed before it is built
Palette and layout are decided first, then every colour pair is tested for contrast and for colour-vision deficiency before a line of it is transcribed. Failures get corrected, and the deviations are recorded.