Data cleaning

Clean data before you rely on it for reporting, decisions, or automation. Dawn can help you inspect a dataset, propose a sequence of changes, and turn agreed steps into a reusable pipeline.

Start with a data-quality review

Add the dataset to a conversation and ask Dawn to profile it before changing anything. Useful checks include:

• Missing values and empty columns
• Duplicate records or duplicate identifiers
• Inconsistent categories, labels, and units
• Dates, numbers, or currencies stored as text
• Values that look unusual for the business context

Try a request such as:

Profile this dataset for data-quality issues. Identify the fields that need attention,
explain the likely impact on analysis, and propose a safe order of operations.

Make changes deliberately

Agree on how each issue should be handled before applying it. Common decisions include whether to remove or retain duplicates, how to represent missing values, and which source should win when records conflict.

Note: Cleaning is not only a technical task. Ask a domain owner to confirm the meaning of ambiguous values, especially when changes affect financial, customer, or regulated data.

Turn repeatable work into a pipeline

When a cleaning step will be needed again, save it in a pipeline rather than repeating it manually. Give each step a clear name and keep the source dataset unchanged so the transformation remains easy to review.

A good pipeline makes it possible to answer three questions: what changed, why it changed, and whether the output is ready for its intended use.

Validate the result

Before sharing or scheduling the output, compare it with the source:

1. Check row counts and key totals.
2. Review a small sample of changed records.
3. Confirm important categories and date ranges are still present.
4. Record any business rules or assumptions in Context.

Next step

Once the dataset is trustworthy, use it in a dashboard, report, or further exploration with Dawn.