Parquet files are much smaller than the CSV they came from and carry the type of every column, so analytics tools such as DuckDB, Spark and pandas load them quickly. Converting a CSV is usually the first step of getting a dataset into a data warehouse.
One CSV can produce several PARQUET files; when it does, they are offered together as a ZIP download.
Each column gets one type chosen from all of its values: a column is numeric only if every non-empty cell is a number, and anything mixed, or a value such as a ZIP code with a leading zero, stays text. A workbook with several sheets gives one file per sheet.