One worksheet per PDF page
MakeFileWork infers stable column anchors, recognizes numeric columns and now detects both large visual gaps and persistent column-layout changes that separate independent tables on the same PDF page. A page with distinct sections can therefore become multiple clearly named worksheets instead of being forced into one distorted grid. Within each detected region, missing values keep their correct empty column, reducing shifted cells in statements, invoices and reports.
Scanned PDFs use OCR automatically
If no useful text layer is found, the existing EN+ES OCR engine keeps each recognized word’s position and feeds that geometry into the same column-inference engine used for searchable PDFs. This preserves table columns far better than converting OCR into plain lines first. OCR fallback is limited to 20 MB and 20 pages per job.
Honest table extraction
PDF stores page appearance, not spreadsheet semantics. Merged cells, irregular borderless tables and complex layouts may require cleanup after conversion; MakeFileWork reports that limitation rather than claiming perfect reconstruction.