PDF to Excel data extraction

A step-by-step walkthrough of the third option is on how to extract data from a PDF to Excel.

Rather than reading text and hoping, you tell AxioExtract what kind of document it is — engineering drawing, exploded view with parts list, P&ID, electrical schematic, piping isometric, SPIR workbook, invoice, or an office table — and it returns the fields that document type is supposed to contain. The output columns are identical on every file of the same type, so sheets from one set append cleanly into one workbook.

Rates follow document complexity: invoices and plain tables are 1 credit per page, drawings 4 credits per page, and P&IDs 8 credits per page. Every rate is listed on supported formats.

PDF, PNG and JPG are all read as pages, so a scanned drawing, a photographed nameplate or a screenshot of a table are all valid inputs. Up to 12 image pages are read per file, rising to 40 pages for text-based PDFs and spreadsheets, and files can be uploaded singly or in a batch.

Extracted rows sit beside the page. Selecting a row highlights where the value was read, low-confidence readings are flagged, and any region can be boxed and re-read. Exports — Excel (.xlsx), Google Sheets, JSON, XML — reflect the reviewed values.

Choosing the document profile is what turns a text dump into columns. An invoice batch goes through the invoice profile and returns supplier, invoice number, dates, line items and net/VAT/gross totals. A drawing sheet goes through the drawing profile and returns the title block, revisions, notes and dimensional callouts. A report with tables goes through the office table profile. A spare parts record goes through the SPIR profile, which is read deterministically rather than visually.

If you are not sure, the type can be detected on upload and changed afterwards, and a workspace can define its own document type with bespoke fields and columns when none of the built-in profiles fits the paperwork you receive.

Files can be uploaded singly or as a batch, and each page inside a file is read in its own right, so a ten-page invoice run comes back as one set of rows rather than ten separate jobs. Credits are charged per page at the rate for the chosen profile, so the cost of a run is known before it starts.

Long documents are the usual reason a conversion disappoints elsewhere. Here the page count is read up front, the quote reflects it, and anything above the ceiling is reported rather than silently truncated.