Updated Nov 18, 20241 min read
datasets
data quality
csv cleanup
Start with a tidy header row
Use short, descriptive column names like company_domain or contact_url. Avoid empty columns or merged cells, especially when exporting from spreadsheets. PromptLoop matches these headers to the YAML defined inputs in your task.
Normalize the key values
For the column you plan to use as the task input, make sure each row follows the same pattern:
- Domains should exclude protocols (promptloop.com, not https://promptloop.com).
- Locations should follow City, State or ISO country codes.
- If you need multiple values, place them in separate columns rather than comma separated strings.
Consistent formatting increases the first-pass accuracy of the browsing service.
Tag optional helper columns
Many teams include helper columns such as keywords, product_notes, or exclude_terms. Add them even if they are blank for now. It reminds collaborators what context is available and lets you populate the data later without editing the task definition.