Data sources & feasibility
How each US retailer's data can realistically be obtained. Program terms change, so verify with each program before relying on it.
| Retailer | Access path | Identifiers | Feasibility | Notes |
|---|---|---|---|---|
| Best Buy | Public Products API. Free key, self-serve. | SKU, UPC, model # | High | Adapter built: npm run ingest:bestbuy. Best starting point. |
| Amazon | Amazon Associates product API | ASIN, sometimes UPC and model # | Medium | API access is gated on referring qualifying sales, and the terms restrict caching and price display. Plan to launch without it, then add it. |
| Walmart | Affiliate program + affiliate product API (approval required) | Item ID, UPC | Medium | Good UPC coverage makes matching easy once you're approved. |
| Target | Affiliate program via an affiliate network. Product feeds for approved partners. | TCIN, UPC in feeds | Medium | No public product API, so you depend on feed access. |
| B&H Photo | Affiliate program. Feed availability varies. | B&H #, MFR # | Medium | Strong for monitors and cameras. |
| Costco | No known public product API or affiliate feed | Item #, retailer-exclusive model # | Low | Scraping carries ToS and anti-bot risk. Consider user-submitted prices or skip. Costco variants are still valuable to call out. |
Risk register
| SKU matching | Proven in this MVP: most listings match automatically through GTIN, model # and model family. The long tail (title-only listings, bundles) needs an LLM + review step. | Low–Med risk |
| Data access | The real bottleneck. Every retailer except Best Buy requires affiliate approval, and some require sales history first. | High risk |
| Price freshness & history | Needs a scheduled job snapshotting prices (daily, hourly for deals) into a time-series table. Cheap to build. | Low risk |
| Spec normalization | Hand-curated here. At scale an LLM extracts specs from retailer pages and manufacturer sheets into a per-category schema. | Medium risk |
| Distribution / SEO | Google has grown hostile to thin affiliate comparison pages. The differentiator has to be data others don't show (variants, cross-store history, spec equivalents). | High risk |
| Monetization | Affiliate commissions on electronics are low single digits. Needs volume, or B2B data sales to brands and retailers. | Medium risk |