Warren Buyer Discovery

What it is

A pair of Python scripts (~300 lines total) that query the official UK Companies House API across five production-sector search terms, fetch per-company detail, and run each active result through a transparent 0-100 heuristic scorer weighting company-name keywords, UK-city location, and status. Results are deduped by company number, bucketed into Tier A/B/C, and written out as JSON plus outreach-ready CSVs. A bash wrapper loads the API key from a local .env (failing loud if missing) and an hourly scheduler runs the whole thing unattended overnight, so a fresh ranked prospect list is waiting each morning. It is deliberately a lean prototype: no test suite, no README, and scoring is keyword-heuristic rather than ML, but it is a genuinely working end-to-end pipeline that turns a public registry into a prioritized sales queue.

Highlights

• ◆Hits the official UK Companies House REST API (search/companies + company detail endpoints) with API-key auth, 10s timeouts, and 0.5s rate-limit sleeps to stay polite
• ◆Heuristic 0-100 relevance score: active-status base, name-keyword weighting (VFX/film/TV/post), UK-city location bonus, detail-fetch bonus, then bucketed into Tier A/B/C
• ◆Runs unattended: run_overnight.sh sources the key from .env (fails loud if unset) and a scheduler loops hourly through the night, verified across a full run producing 57 scored companies (10 Tier A)
• ◆Dedups by company number across five queries and exports both machine-readable JSON and outreach-ready CSVs, including a priority list and an outreach tracker
• ◆Concrete business intent: finds UK film/VFX/TV production firms as green-screen chromakey fabric buyers, translating a raw registry into a ranked sales pipeline