Data Products and the Dataset Marketplace

Learn how Zerocost data products work, how datasets are created from anonymized interactions, and how the marketplace benefits startups and partners.

Written By Harshil S

Last updated About 2 months ago

Zerocost supports optional data products alongside ads. The goal is to help startups monetize anonymized AI interaction data and help partners buy datasets for training and evaluation.

What is sold in the marketplace?

Partners may purchase datasets that include:

  • Anonymized prompt-response turns

  • Session-level grouping and useful metadata

  • Domain tags and categorization to support evals and training

How datasets are created (high level)

  1. Your app captures eligible interactions (if enabled and user opts-in).

  2. Events are stored with dataset metadata (domain, category, bundle key).

  3. Admin catalog entries surface bundles with tags, descriptions, and pricing.

Data quality rules of thumb

For higher-value datasets:

  • Keep schemas stable across exports (example: JSONL).

  • Maintain consistent domain taxonomy and labeling.

  • Scrub PII at ingestion and document what is excluded.

Recommended rollout

  • Start with ads only.

  • Enable data products only after you have privacy controls, retention policy, and a clear value proposition for dataset buyers.