PREPROD
ReaderAuthor

Tally · Compliance

OCR confidence

How confident the automatic scan was when it read the document's fields.

1 min read · 2 depth levelsstratum-tally · vpilotMatches production
Everyday

When a health paper is scanned, Tally reads its text automatically and fills in the fields. This score says how sure it was — closer to 1 means a clean, reliable read; lower means the scan was blurry or hard to parse and the fields are more likely to need fixing. It's only set when the document was auto-extracted.

A low confidence score is a prompt to eyeball the document and correct any fields, then mark it manually verified. Don't trust a low-confidence auto-read for a legal compliance record.

Technical section
Data model, API, and internals — visible to developers.
View as developer
Was this helpful?