Skip to main content
Use a file upload for a snapshot of data. Use a connection when you need to bring in data repeatedly from another system.

​
Upload a CSV or Excel file

  1. Open the intended folder in Workspace.
  2. Choose Create → Upload dataset.
  3. Select a .csv or .xlsx file and complete the name and import settings shown.
  4. Submit the upload and wait for processing to complete.
  5. Open the dataset and check its preview before using it in analysis.
For CSV files, a single header row, UTF-8 encoding, and comma-separated fields provide a reliable starting point. Keep identifiers such as postcodes and account numbers distinct from quantities: leading zeros can matter. For Excel, prepare a rectangular table with one header row. Put the intended data on the first worksheet for the standard import. Avoid merged headings, introductory text above the table, and totals mixed into the data rows. Export the intended worksheet as CSV if the workbook does not import as expected.

​
Validate the imported dataset

Check these items after every new import:
  • Rows: compare the count with the source, allowing for headers and deliberately excluded records.
  • Columns: check names and inferred types, especially dates, amounts, and identifiers.
  • Coverage: check the earliest and latest dates and whether the latest period is complete.
  • Values: spot-check records containing blanks, non-English text, decimals, and leading zeros.
Column names may be normalised during ingestion. Use the names shown in Dawn when writing a query or building a transformation.

​
Import from an external source

Create a connection with reusable source details and a credential. Then specify the data to import: a query, object path, spreadsheet range, or reporting period. These are separate choices. A connection does not itself create a dataset or establish a refresh schedule. Follow connecting and importing data, then choose the guide for your provider.

​
Refreshing data

A file upload is a snapshot. To refresh external data, run the relevant import or pipeline again, or configure the refresh workflow available for that asset. Verify a successful manual run before scheduling it. Decide whether the workflow should replace a snapshot, append new records, or update existing records. Do not assume a scheduled import automatically deduplicates data or captures only changes. Make the intended behaviour explicit in the pipeline and check the output after a second run.

​
Troubleshooting

For repeated cleanup, keep the source data and create a prepared dataset through a data engineering pipeline.