MerRec: 1-10 Billion C2C Recommendation Records from Mercari US

MerRec is a large-scale consumer-to-consumer (C2C) recommendation dataset containing between 1 billion and 10 billion records derived from the Mercari marketplace. Created by Mercari US researchers for KDD 2025, it provides anonymized interaction and item data for sequential recommendation and click-through rate prediction. The data is formatted in Parquet and supports both tabular and text modalities.

Use Cases

• Sequential recommendation modeling using user interaction sequences
• Click-through rate (CTR) prediction based on item features
• Text-based item representation learning using listing descriptions

Strengths

• Scale of 1 billion to 10 billion records
• Multimodal support including both tabular features and text descriptions
• Specific focus on C2C (consumer-to-consumer) marketplace dynamics

Limitations

• Restricted to non-commercial use under CC BY-NC 4.0 license
• Anonymization may obscure specific geographic or user demographic details

Provenance

Source

Mercari US (KDD 2025 publication)

Collection Method

Derived from Mercari's internal marketplace data

Freshness

Last updated July 2025; reflects data prepared for the KDD 2025 conference.

Geography

United States