Finding and exporting your BigQuery data

What it holds

BigQuery holds analytical tables, event streams and transformed business datasets. Nested fields and repeated values can preserve richer event structure than a flat spreadsheet.

Where BigQuery shows the size

1. Run SELECT table_schema, SUM(total_rows) AS row_count, ROUND(SUM(total_logical_bytes)/POW(1024,3),2) AS logical_gb, ROUND(SUM(total_physical_bytes)/POW(1024,3),2) AS physical_gb FROM region-us.INFORMATION_SCHEMA.TABLE_STORAGE WHERE NOT deleted GROUP BY table_schema ORDER BY logical_gb DESC;
2. Substitute the correct region and execute the query in that location. Repeat for other regions.
3. For one table, open Explorer, select the table, and inspect Details > Storage info.

Plan and admin role. Storage metadata requires roles/bigquery.metadataViewer or bigquery.tables.get and bigquery.tables.list. Exporting needs bigquery.tables.export, bigquery.jobs.create and destination storage.objects.create/delete permissions.

Exporting the full history

1. Prepare a Cloud Storage bucket matching the dataset location.
2. Export a table with bq extract --destination_format PARQUET --compression SNAPPY mydataset.mytable gs://my-bucket/mytable/part-*.parquet.
3. Alternatively run EXPORT DATA OPTIONS (uri='gs://my-bucket/mytable/*.parquet', format='PARQUET', overwrite=true) AS SELECT * FROM mydataset.mytable;
4. Repeat per table, or use the table menu Export / sync > Cloud Storage.

For standard table extracts, use Cloud Storage rather than a direct local, Drive or Sheets destination. A single extract job handles one table; wildcard paths allow sharded output beyond the single-file 1 GB logical-data limit.

CSV cannot represent nested or repeated fields. JSON exports encode INT64 values as strings; Avro or Parquet may better preserve structure.

Shared-pool extract jobs have a 50 TiB daily project limit; PIPELINE reservations offer a paid capacity alternative. Choose destination retention settings carefully because they can interfere with overwrites.

You don't need to run an export to talk to us, and there's no need to clean it first — removing personal data is part of our work.