# BGraph synthetic clickstream events feed

This package mirrors the supplied raw field structure and expands the synthetic examples across 12 category and market combinations. It includes hashed randomized user ID, UTC date and millisecond timestamp, country, optional city and postal code, referrer URL, target URL, user agent, and optional HTTP code.

## Delivery alignment

- Public evaluator file: CSV and JSON
- Production-ready raw option: TSV compressed with gzip
- Partition pattern: Year/Month/Day
- Line endings: Linux LF
- File sizing: scoped during delivery planning, with the supplied reference allowing files up to 1000 MB
- Parquet option: schema included in this package

## Synthetic category examples

consumer electronics, travel, productivity software, smart home, financial services, apparel, beauty and personal care, online education, automotive, media streaming, cybersecurity, and food delivery.

## Limitations

- Every row and identifier is fabricated.
- All URLs use reserved example domains.
- Markets, cities, postal values, categories, and paths are placeholders.
- The file does not demonstrate current users, traffic, coverage, geography, volume, history, or availability.
- Pseudonymous event-level delivery requires source-rights, privacy, retention, security, jurisdiction, and intended-use review.
- Fields may be removed, generalized, aggregated, or delayed for an approved product.

## Files

- sample.csv
- sample.json
- parquet-schema.json
- data-dictionary.json
- example-query.sql