Data cleaning suggestions for a new source table
Open a raw table in a data prep tool and get suggested cleaning and transformation steps that an analyst accepts one at a time.
The use case
Raw tables from new sources come with mixed date formats, stray whitespace and inconsistent codes. BigQuery data preparation uses Gemini to look at the data and suggest cleaning, transformation and enrichment steps, which the analyst can accept, edit or reject. Accepted steps become a reusable pipeline that can be scheduled. The analyst still checks each step on sample rows, makes sure nothing drops valid data, and documents choices like how missing values are filled. Writing down why each step exists makes the pipeline easier for the next person to trust.
Example workflow
- Open a raw table in data preparation
- Review suggested cleaning steps
- Accept, edit or reject each step
- Check results on sample rows
- Save and schedule the pipeline
Tools
- BigQuery data preparation
- Gemini in BigQuery
Risks and guardrails
- Suggested filters can drop valid rows
- Fill rules for missing values change results; document them
- Review before scheduling to production tables
Sources
Written by Agents Lookup. Credits and official pages for the tools this use case names:
- Google Cloud docs: BigQuery data preparation: docs.cloud.google.com/bigquery/docs/data-prep-introduction