Skip to main content

​
Format Overview

CSVJSONParquet
Extension.csv.gz.json.gz.parquet
CompressionGzipGzipSnappy (built-in)
Split Threshold50,000 records/file50,000 records/fileNo splitting needed
Best ForSpreadsheets, quick analysisApp integration, APIsData warehouses, analytics
Relative SizeLargest~1.1× CSV~0.3× CSV

​
CSV

Comma-separated values compatible with Excel, Google Sheets, and most data tools.
  • Extension: .csv.gz
  • Compression: Gzip (automatic)
  • Splitting: Automatically split at 50,000 records per file
  • Best for: Spreadsheets, quick analysis, data sharing
CSV files are always gzipped. If the export exceeds 50,000 records, multiple .csv.gz files are produced and bundled into a single .zip download.

​
JSON

Standard JSON array format for programmatic consumption.
  • Extension: .json.gz
  • Compression: Gzip (automatic)
  • Splitting: Automatically split at 50,000 records per file
  • Best for: Application integration, API-style workflows, programmatic parsing

​
Parquet

Columnar binary format optimized for analytics workloads.
  • Extension: .parquet
  • Compression: Built-in (Snappy)
  • Splitting: Not required — Parquet handles large datasets efficiently in a single file
  • Best for: Data warehouses, Spark/DuckDB/Pandas pipelines, BigQuery, Snowflake
Parquet files are significantly smaller than CSV/JSON for the same data (typically 3–5× smaller) and support predicate pushdown for efficient querying.

​
File Splitting & Naming

CSV and JSON exports are automatically split when the total record count exceeds 50,000 records per file. Each split file follows this naming convention: 
export_{export_id}_part001.csv.gz
export_{export_id}_part002.csv.gz
export_{export_id}_part003.csv.gz
When an export produces multiple files, they are bundled into a single .zip archive for download: 
export_{export_id}.zip
├── export_{export_id}_part001.csv.gz
├── export_{export_id}_part002.csv.gz
└── export_{export_id}_part003.csv.gz
Parquet exports produce a single file regardless of size: 
export_{export_id}.parquet

​
Size Comparison

Approximate file sizes for a 100,000-record export with all columns selected:
FormatCompressed SizeNotes
CSV (.csv.gz)~120 MB2 files × ~60 MB each
JSON (.json.gz)~135 MB2 files × ~67 MB each
Parquet (.parquet)~35 MBSingle file, columnar compression
Choose Parquet when working with analytics tools — it’s the most compact format and supports column-level reads, so tools like DuckDB or Pandas only decompress the columns you query.