Supplier-spend analytics

The real suppliers (HD Supply, Amazon Business, a few others) get paid by credit card, so they never hit accounts-payable aging. The usual "who do we owe" report is blind to exactly the vendors we spend the most consumable money with, and I could not see spend by category, by property, or over time.

How it works

I export order history from each supplier, HD Supply's invoice CSVs and Amazon Business's Order History Report, and parse them in Python. The files are messy: duplicate column headers, denormalized print templates, properties identified only by shipping address. The parser dedupes by a stable line identity, reconciles line nets back to invoice subtotals, and harmonizes each supplier's coarse categories into common buckets. Then it normalizes everything to available and occupied room-nights using occupancy from the archive, so a big property and a small one are actually comparable.

• Cost per occupied room-night per property, which surfaced one property running several times another's rate.
• A category and price-watch pivot that caught real vendor inflation on specific SKUs.
• A supplier crossover, one vendor declining as another accelerates, that only shows up once the data is unified.

It is live as a dashboard, refreshed from manual exports for now. The automated path (fetching invoice PDFs over IMAP, plus a persistent-session downloader for the other vendor) is built and being hardened on the always-on host.

This is analyst work, pull, clean, normalize, pivot, that no one was doing because it never fit in anyone's week. The agent does it on demand, so procurement decisions get made against evidence instead of instinct.