Purchase invoice workflow
Extract invoice line items into Excel without losing context
Keep invoice headers, line items, adjustments, and source references separate when preparing a purchasing workbook.
Separate the invoice from its lines
Before converting invoices, decide which values describe the entire document and which belong to each line. Supplier, invoice number, date, and currency usually provide document context. SKU, description, quantity, unit price, and line amount describe an individual purchase. Keeping those roles distinct makes it easier to check and combine repeated documents.
Use a stable internal document identifier to connect the header and line tables. Invoice numbers alone may not be unique across suppliers. Keep the original filename and page reference available for review. These are workflow choices for traceability, not instructions to replace your accounting system's own record rules.
Choose extraction by the required output
PDF.co's AI Invoice Parser documentation describes structured invoice fields and line items. A general table converter instead needs to be evaluated for how it reconstructs the rows you require. Choose based on the output contract, rather than assuming that every PDF-to-Excel operation understands an invoice's business structure.
Run a small sample before processing a folder. Include a single-page invoice, a document with several pages, and one containing a discount or credit line. Compare every required field in that sample with the source. A successful extraction response does not establish that all line items or document adjustments have been captured.
Preserve adjustments and missing values
Keep freight, discounts, and other document-level adjustments distinct from purchased items when the source presents them separately. Do not repeat a full invoice total on each line and then sum that column as if it were line revenue or cost. Give each field a clear name that reflects its level.
Leave a missing unit price unresolved if it is not present in the source. A mathematically implied value can be useful for analysis, but label it as calculated and keep it separate from the extracted value. The same principle applies to dates, currencies, and identifiers: inference should never quietly become a claimed transcription.
Check numerical interpretation and completeness
Use the source's decimal and date conventions. Microsoft's Power Query guidance explains how locale affects conversion into typed values. Preserve the original text when a value is ambiguous, and record the chosen interpretation in the working process. Formatting a cell as currency is not a substitute for verifying its underlying value.
Count the invoice's line items and inspect page boundaries. Reconcile line amounts and document adjustments against the printed summary using the supplier's displayed arithmetic. Treat an unexplained difference as an exception. A matching total is helpful, but it does not prove that quantities, descriptions, or SKU assignments are correct.
Approve a complete document
Mark a document ready only when the required fields, lines, and checks are complete. If one page is missing, retain the partial output but keep the whole invoice out of the approved batch. This avoids presenting an incomplete purchase record as a fully processed document.
Save the raw extraction, the corrected workbook, and a brief correction record. For recurring suppliers, use observed problems to refine the next small trial. The goal is a repeatable process that exposes uncertainty and reduces rework, not a spreadsheet that merely resembles the invoice on screen.
Sources
Evidence status: methodology. No unverified accuracy, savings or traffic claim is made.