Pricing Guidance System for Xianyu Secondhand Marketplace

Xianyu, China’s largest second‑hand trading platform, hosts billions of items. Accurate pricing is essential for quick sales, yet most sellers are individuals lacking pricing expertise.

The platform covers hundreds of millions of products across categories such as phones, cosmetics, clothing, pets, real‑estate, etc., each described by new‑product attributes and second‑hand attributes, with many non‑standard goods.

We first establish a pricing standard that combines product‑level dimensions (SKU/SPU/brand/category to obtain a new‑product market price) and second‑hand dimensions (usage time, condition, specific category attributes). A depreciation function f(x₁,…,xₙ) maps these attributes to a discount factor, typically fitted with interpretable models.

When a seller publishes an item, we extract attributes via real‑time text mining, image understanding, and optional user input, then generate a recommended price range. The final range is adjusted for market supply‑demand and seller willingness (fast sale vs higher price).

The system now covers over 60% of items, achieving >65% accuracy for new‑attribute recognition, >95% for second‑hand attributes, and >65% overall price recommendation accuracy.

Future work focuses on improving coverage and precision for non‑standard and niche categories.