PDF form extraction guide
Map fillable PDF fields without OCR or silent guesses
A clean table is useful only when its transformations are explainable. Good mapping combines known aliases, applies narrow normalization rules and leaves uncertain values blank with a reason.
Start from the field inventory
List field names and types from every supported form version. Build output columns from this evidence instead of from labels that merely appear near the field on a page.
Map aliases deliberately
If an older form says Customer Name and a newer version says Full Name, map both source fields to one output column. If both appear with different values, record a conflict rather than picking one.
Normalize bounded types only
Whitespace and case changes are predictable. Boolean, integer, decimal and date parsing should accept documented formats and create an unparseable issue when a value falls outside them.
Preserve review states in the handoff
Keep missing, duplicate, conflicting and unparseable issues alongside each source row. A blank cell with a reason is more trustworthy than a guessed value with no provenance.