Can I turn a PDF purchase order into a Shopify draft order?
Yes, when the PDF contains real text — the kind you can select and copy, which is what purchasing and ERP systems produce. NeatPO reads that text, rebuilds the table of lines, matches each to your catalog and shows every line next to the page it came from before any draft is created. A scan or a phone photo of a printed order is a picture of text, and the current deployment does not read pictures.
No install needed for the sample. Plans start at $149/month, billed by Shopify, with a 7-day trial.
Do it now
- Open the PDF and try to select a word. If you can, it has a text layer and NeatPO can read it.
- Upload it in the Inbox (up to 30 pages) or forward the email with the PDF attached to your store’s NeatPO address.
- Review the extracted lines against the page, resolve the issues, approve and create the draft.
Why text PDFs and scans are different problems
A PDF from a purchasing system stores characters and their positions, so the product codes and quantities are there to be read exactly. A scan stores pixels; reading it means recognising characters, and recognition makes mistakes in exactly the places that matter — a 1 read as a 7, a decimal point lost in a smudge.
Rather than guess at a scan, NeatPO reports it as unreadable on this deployment and keeps the original. That is a real gap, stated here so you do not find it on a busy Monday.
Tables without lines
PDF tables are often just text that happens to line up. NeatPO rebuilds columns from the positions of the text. When a value could belong to either of two columns, or a line has an extra number it cannot place, it raises an issue on that line instead of quietly shifting values into the wrong column.
Emailed purchase orders
Each store gets a private inbound address. Forward the buyer’s email and the attachment becomes a document in your Inbox, with the same checks and the same review as an upload. Nothing is sent back to the buyer.
The honest limits
- Scanned PDFs and images (PNG, JPG) are not read on the current deployment: no OCR or extraction model is configured.
- Up to 30 pages and 20 MB per PDF; password-protected PDFs are refused rather than broken.
- Very unusual layouts — several orders in one PDF, lines that wrap across pages mid-row — may need more manual review.
Checked against the product on October 7, 2026.