Work with your data
Data is useful before you have an app. You can upload a table, inspect a connected dataset, query several tables together, and investigate their keys and relationships from the CLI or platform.
Use this section to understand what you are looking at and how to investigate it. Use the CLI data reference for command options.
Choose the right kind of data
| Data | Where it comes from | Typical use |
|---|---|---|
| Uploaded table | CSV, TSV, or XLSX ingestion | Explore a file or start with an existing dataset |
| Synced table | A connector's dataset sync | Analyze data refreshed from another system |
| Transform output | A saved background calculation | Reuse prepared data, aggregates, or model results |
| Managed database table | App-owned PostgreSQL storage | Read and update operational records through backend code |
A table's ID identifies it within the project. For file-backed data, its publication identifies the output of a particular completed run. A stable table ID can therefore have changing rows and schemas over time.
Investigate before building
Start with tables and publications: find the dataset, inspect its columns, and choose the version/run you mean. Use SQL analysis to check filters, aggregates, and joins without creating a transform. Review keys and relationships when you need to understand the grain of a table or how datasets connect.
For example, before building an order dashboard, check whether orders contains one row per order or one row per order line. Then validate how customer IDs join and whether an aggregate counts orders or lines. These choices affect the app's behavior more than the chart component you choose.
Turn an investigation into part of the app
A one-off analysis produces a saved query result. When the calculation should refresh and supply other resources, make it a transform. When a user needs a request-specific response, expose an API. When users need to save decisions or create records, use a managed database.
See Bring in data for uploading and implementing source integrations.