A seed is a small CSV file versioned with the project.
dbt seed loads it into a relation that models can reference. Seeds work well for maintained lookup lists, such as a currency label or an approved event category. They're a poor substitute for a loader copying a changing operational system.
CSV has rules of its own. A field containing a comma needs quotation marks, and a blank field may arrive as a missing value. Check the parsed rows before using a seed in a join.
A lookup also may not contain every source code. If the business wants an exception list, keep those unmatched records visible instead of losing them through an inner join. Build loads seeds along with the selected models, while seed on its own only loads the CSV resources.