CSV cells carry text but usually do not carry a schema that distinguishes identifier, quantity, code, boolean or missing value. Two fields can have the same characters and require different destination types. Guessing from spelling alone is convenient for demos but can alter data before the destination has a chance to apply its own rules.
Leading zeros often belong to an identifier
Parsing 00123 as a number produces 123 and destroys the original width. Postal codes, stock codes, fixed-width references and test identifiers may rely on that width even though their characters are digits. CSVBridge emits "00123" so the destination can decide deliberately whether it is an identifier or a numeric quantity.
Decimal spelling may carry chosen scale
The strings 12.5 and 12.50 can be mathematically equal while communicating different precision or formatting expectations. Binary floating-point conversion can also choose a different serialized representation for long decimals. Keeping "12.50" as text preserves the source token; a schema-aware consumer can later choose decimal arithmetic, validation and rounding intentionally.
Boolean-looking and empty text is still text
The token false might be a status label rather than JSON false; null might be four literal letters; an empty cell might mean blank, unknown or not applicable. CSV alone does not settle those distinctions. CSVBridge produces strings for all three and makes this policy visible beside the output instead of silently inventing nullability or boolean semantics.