New Dataset Aids Data Pricing in Marketplaces.

The increasing importance of high-quality data in machine learning has led to a proliferation of data marketplaces like AWS Marketplace and Databricks. However, accurately pricing data products remains a significant challenge due to data's unique properties, such as near-zero marginal replication costs and unpredictable revenue generation. Traditional economic pricing models often fall short; cost-based approaches fail due to replication ease, and income-based approaches struggle with revenue unpredictability. While sales comparison is viable, it lacks standardized benchmarks across diverse marketplaces. To address this, the DaDaDa dataset has been created, comprising metadata for 16,147 data products from nine global data marketplaces. This dataset aims to enable the training of pricing models, establish benchmarks for new data products, and support tasks like data product classification and retrieval, as demonstrated by initial experiments.