Extraction problems.
THE COMMON CAUSES
Poor image or scan quality Complex table layouts Fields not present, filled by inference Ambiguous formatting
THE INSTRUCTION THAT PREVENTS MOST OF IT
"If a field is not present, write 'not stated'. Do not infer or guess."
Without it, gaps are filled by inference and presented as extracted data.
FOR VERIFICATION
"Quote the text each value came from."
Slower, and it lets you check.
FOR POOR IMAGES
Rephotograph. Clear, well-lit, straight, whole document in frame.
Image quality affects accuracy more than anything else.
FOR TABLES
Complex layouts with merged cells are handled imperfectly.
Split them, or check carefully.
WHAT TO ALWAYS CHECK
Every figure, against the source Totals especially
FOR BATCHES
Check a sample thoroughly, plus every total.
