The problem
Cost per Mile (CPM) became the most influential metric in Construction, and it was informing merger-related questions and strategic decisions. But the analysis lived in a fragile chain of three Oracle views and a large Excel-formula workbook, and every new question meant a new spreadsheet.
What I built
- One data model. Re-engineered the three-view chain and the workbook into a single materialized view of about 166 columns with a daily refresh. Categorical logic moved into maintainable lookup tables.
- Correctness first. Found and fixed defects that had inflated blended cost-per-mile, then reconciled the dashboard to the authoritative monthly financials.
- An API. A Python/FastAPI backend with server-side whitelisting, fully parameterized SQL, and connection pooling against Oracle.
- A front end. React, TypeScript and Recharts with reusable charts and a shared measure-and-dimension engine.
- Persona-based dashboards (Executive, Manager, Front-line) designed to answer the roughly 300 most common business questions in three clicks or fewer.
- Delivery. Deployed as unattended, reboot-proven services on a Windows server, with runbooks, Git/Bitbucket adoption for the team, and onboarding documentation to reduce key-person risk.
Why it matters
The platform keeps all enterprise data in-house, removes the need for commercial BI seats, and gives the team a repeatable “build the next dashboard” playbook that new hires can follow.
Figures are drawn from my resume and shared in general terms; no proprietary datasets or screenshots are published here.