Skip to main content
Observe exports are Apache Parquet and CSV files, so any platform that reads these formats can load them. This page describes a loading pattern that stays correct when exports are regenerated and how common platforms load the same files.
Observe writes a nightly table export as Parquet files. An export is listed only when complete. Your pipeline downloads the files through the API or CLI and loads them into your data platform, such as a warehouse, lakehouse or BI tool. Each table and day is loaded once and can be reloaded safely.Observe writes a nightly table export as Parquet files. An export is listed only when complete. Your pipeline downloads the files through the API or CLI and loads them into your data platform, such as a warehouse, lakehouse or BI tool. Each table and day is loaded once and can be reloaded safely.

​
1. The loading pattern

1

List completed exports

Call GET /observe/tableExports once a day after 01:00 project time. Only complete exports are listed.
2

Skip what you already loaded

Keep a control table with table, export date and each file’s name, creation time and checksum. Load an export when its file inventory or metadata changes. Consume every listed part, including any sub-day parts.
3

Download the files

Request a download link per file and download the complete set into a fresh staging folder for that export version. Verify the files before replacing the previously loaded version, because a regenerated export can have fewer files.
4

Load a complete export version

Keep the source table, export date and version with the staged rows. Replace that export partition atomically after all parts load successfully. Reprocessing the same export then leaves the same result.
For journey analysis, account for the row grain and joins. The two time-series tables contain overlapping monthly snapshots: retain the export version and use the latest snapshot covering each reporting interval, rather than summing successive exports. For console-aligned reporting, start with time-series exports. Accept new columns during the load, because the schema grows as Observe adds fields.

​
2. Platform overview

​
3. Custom exports

Enterprise scoping. If your platform requires direct delivery into storage you own, agree the delivery design and availability with Corbado before implementation.
  • Destination and access: storage location, region, ownership, permissions and encryption requirements.
  • Contents and cadence: required tables or journey datasets, KPI definitions, delivery frequency and history.
  • Corrections: how late events and reclassified journeys reach your platform, including replacement or versioning rules.
  • Operations: batch completeness, retries, backfills, schema changes, retention and deletion handling.
The API and file downloads described above provide the existing integration path. Contact us to scope a custom delivery arrangement.

​
4. Current data and history

Table exports are written once per day, so they suit history, reporting and joins with your own data. For questions about the last hours, such as an alert that fired this morning, query the Observe API or the CLI directly. Both read the current state of your project.

Exports

Schedules, retention and download options.

Data model

Tables, keys and time columns.