Compare Empire

Price comparison for Australia, starting with mobile phones. Live at compareempire.com.

How it works

1. npm run scrape — reads the public phone listings of 17 stores (scripts/scrapers/) into data/phones/raw/<retailer>.json: JB Hi-Fi, Harvey Norman, The Good Guys, Mobileciti, Apple, Samsung, Telstra, Optus, Vodafone, amaysim, TeleChoice, Costco, Umart, MSY, digiDirect, Betta and Mwave. A store that fails keeps yesterday's file. Carrier listings are kind: "plan" (device repayments) and never mix into the outright ranking. docs/stores/survey.md lists the 80+ stores checked and why the others (Amazon, eBay, Kogan, Big W, Officeworks…) aren't scraped.
2. npm run build:catalog — matches listings to phones by brand + model + storage (src/lib/normalize.ts), drops refurbished/carrier-locked/outlier listings, writes data/catalog.json, today's snapshot to data/phones/history/<date>.json and the chart series to data/series.json.
3. next build prerenders the home, category, brand, deal, browse, compare and guide pages; product pages render on their first request and Vercel's CDN keeps them until the next deploy. No database.

.github/workflows/refresh-prices.yml runs steps 1–2 every morning and commits the data, which triggers one Vercel deploy a day.

Develop

npm install
npm run dev          # http://localhost:3000
npm run refresh      # re-scrape + rebuild the catalogue (≈3 min)
npm test && npm run typecheck && npm run lint

Adding a retailer

Write scripts/scrapers/<id>.ts returning RawOffer[], register it in scripts/scrape.ts and src/lib/retailers.ts, then run npm run refresh and check the new titles parse (tests/normalize.test.ts).