How We Verify Every Deal

Most luxury-deals sites are rankings of whatever's on sale at whatever retailer they have an affiliate relationship with. The deal gets listed; you decide if it's worth it.

Archive Luxury works differently. Every item we surface passes three verification gates before it's published, and our scraping code is built to reject items that don't have defensible source data — even when the sale price looks great.

Here's the full methodology.

Gate 1: Source Authorization

The retailer has to be in our authorized-retailer registry. Not every store that sells luxury qualifies.

What we require:

• A direct brand relationship with the houses whose product the retailer carries, OR a documented wholesale distribution channel traceable to the brand.
• A US-facing storefront, either US-domiciled (Saks, Neiman, Nordstrom) or explicitly DDP to the US (Mytheresa, THE OUTNET, YOOX, Cettire).
• Clean returns and warranty posture — no end-user surprises.

What we reject:

• Resale platforms (The RealReal, Vestiaire, Rebag). Their product is real; we just don't cover secondary market.
• Grey-market channels, even well-known ones. Warranty and service exposure is different from authorized; that's material.
• Retailers whose landed-price posture is unclear. If a European retailer doesn't explicitly state duties are included, we classify them as manual-review and their deals don't publish until a human verifies DDP.

Gate 2: Price History Validation

This is the gate that rejects the most items, and the one that defines Archive Luxury's editorial discipline.

Our scrapers are forbidden from fabricating prices. Every deal we publish must have:

• A real sale_price_cents from the retailer's structured data — JSON-LD price, Shopify variant price, Magento special_price. Never a number we derived or estimated.
• A real original_price_cents from a distinct source field — JSON-LD highPrice or priceSpecification[ListPrice], Shopify compare_at_price, Magento price. Never estimated from a "typical markdown percentage."
• Confirmed USD currency. Non-USD retailers are skipped entirely; we don't do in-app FX conversion.

This is a binding rule. The scraping codebase will skip a product rather than emit a payload with a fabricated price.

Why: past incidents. Multiple retailer scrapers used ESTIMATED_DISCOUNT_PCT constants to back-calculate "original" prices from sale prices — a $4,400 Gucci item was displayed as $7,563 → $10,805 using a 40% assumed markdown. A small fabricated dataset is worse for users than a large honest one. We rewrote the ingestion layer around this principle.

Beyond the source-field rule, Gate 2 also enforces:

• Not stale. If an item has been "on sale" for more than 3 weeks, the markdown is treated as the real price and the "original" comparison is rejected. Evergreen markdowns aren't sales.
• Minimum 30% discount. Shallower discounts rarely represent genuine markdowns in luxury; most are opportunistic re-pricing.
• Consistent across sources. If two retailers list the same SKU with wildly different "original" prices, we flag for manual review rather than publish both.

Gate 3: Editorial Review

After the data gates, each deal is tagged:

• Category — bags, shoes, ready-to-wear, accessories, jewelry, leather goods.
• Urgency classification — quiet flex (evergreen), pricing anomaly (unusually deep discount), final unit (limited stock).
• Historical low flag — if our price tracker shows this is the lowest the item has ever been, it's flagged.
• Era tag — archive, runway, current season, pre-collection.
• Editorial note — a one-line human-readable explanation of why the deal is notable.

Items above $2,000 receive additional manual review before publishing, because high-ticket deals have a higher standard for verification and editorial framing.

What gets filtered out