What are Primary vs Supporting Pages?
Primary pages
Used for data extraction
Drive extracted fields, totals, and line items
Contain core transactional data
Supporting pages
Attached for reference only
Excluded from extraction
Examples: cover pages, summaries, vouchers, receipts, statements
Let tofie label pages for you
If your documents consistently follow the same pattern, for example, the first page is always the bill and the rest are supporting, you can teach tofie to classify pages automatically instead of labeling them by hand.
Go to Knowledge Base and add a card describing the pattern (for example, "the first page is the bill, the remaining pages are supporting documents").
Go to Entity Settings → Features → set Page Classification to:
No Page Classification: tofie will not classify pages; every page is extracted. This is the default.
Classify and Review: tofie classifies pages as primary or supporting and holds extraction until you review and confirm the labels.
Classify and Extract: tofie classifies pages and extracts immediately, using only the primary pages.
Pages tofie has classified show a small AI icon, so you can tell them apart from pages labeled manually.
💡 Tip: tofie defaults to Primary whenever it isn't confident, so a document is never accidentally excluded from extraction.
How to label pages as Primary or Supporting
1. Upload your document
Upload your document as usual
All pages will appear in the Document Review panel
2. Label pages
Select a page in Document Review or Extraction Review
Choose Primary or Supporting
Tip: You can multi-select pages (Ctrl/Cmd + click) to update labels in bulk from both Document Review and Extraction Review. A confirmation dialog will appear.
3. Extract using only Primary pages
Only Primary pages are used during extraction
Supporting pages are excluded from extraction
All pages remain attached to the extraction
Note: An extraction must contain at least one Primary page. If all pages are marked as Supporting, extraction will be disabled and the Extraction Review will show "Select a page to view extracted data."
If you update page labels after extraction:
Re-extract to apply the changes
Extracted results will reflect only the current Primary pages
Best practices 💡
Label pages before extracting whenever possible
Keep only pages with transactional data as Primary
Use Supporting for reference-only pages
Re-extract after changing labels to refresh results