A procurement lead gets the same three questions every month: will we land the year on budget, which contracts need a decision before they auto-renew, and where can we save?
This cockpit answers them for a fictional Brazilian fintech with synthetic data (40 vendors, BRL 18.5M a year). It forecasts each contract by billing model and splits the budget variance into price, volume and unbudgeted spend, exact to the cent. It also gives every contract ending in the next 120 days a decision deadline and one recommendation: terminate, consolidate, right-size, renegotiate or renew.
In the synthetic run, FY2026 lands 5.5% over budget, almost all of it cloud volume. The 15 upcoming renewals are worth BRL 1.04M a year, more than the overrun.
Python builds and tests the data (110 tests). Postgres views are checked against the Python figures in CI. The React front end reads them through Supabase with read-only access for visitors. An AI step extracts renewal terms from contract PDFs for a person to review, and a weekly e-mail lists the decisions due.
Links
Live demo: https://stephanyalves-eng-renewal-cockpit.stephany-alves.workers.dev
Data and logic: https://github.com/stephanyalves-eng/vendor-renewal-cockpit
Front end: https://github.com/stephanyalves-eng/renewal-cockpit
Tools / tech stack Python, pandas, PostgreSQL, Supabase, React, TanStack Start, Tailwind CSS, Recharts, Cloudflare Workers, GitHub Actions, Activepieces, Lovable
Built with