Reliable invoice extraction starts with five choices: which PDFs and images you must handle, which fields to capture, how to accommodate layout differences, how to validate results, and how to test performance on unfamiliar invoices. Treat the output as proposed data—not a verified accounting record—until each important value is checked against its source and your workflow rules.
1. Choose the input and OCR route
First determine whether your invoices contain machine-readable text or are scanned pages and photographs. A digital PDF may provide text and layout information directly; a scan or phone-captured image generally needs optical character recognition (OCR) before fields can be identified. Image quality, skew, orientation and document layout can all affect the result.
As an Amazon Associate I earn from qualifying purchases.
Microsoft documents invoice processing for PDFs and images, including digitally generated PDFs, scans and phone-captured images. Its version 4.0 invoice model documentation also specifies page and file-size constraints. Check that live, version-specific documentation against your actual files before implementation; do not assume limits apply identically across versions or services.
2. Define the fields before comparing systems
Write down the fields you need, along with each field’s expected type, before judging extraction results. A schema turns a vague goal—“read the invoice”—into a testable task. Google Research describes a schema as fields paired with types, with examples including dates, integers, alphanumeric codes and currency amounts.
#1 Best Overall
- PORTABLE SCANNER FOR USE ON-THE-GO — The fastest and lightest mobile single-sheet-fed compact document scanner in its class¹
- QUICK DOCUMENT SCANNING ― This Epson ultra-fast scanner scans a single page as quickly as 5.5 seconds²; Windows and Mac compatible
- VERSATILE PAPER HANDLING ― Portable scanner scans documents up to 8.5 x 72 in; Also easily digitizes receipts and ID cards to make accounting, bookkeeping, and organizing simpler
- INTUITIVE, HIGH-SPEED SOFTWARE — Epson ScanSmart Software³ is a smart tool allowing you to easily scan, review, and save; Stay organized easily with the help of this Epson scanner
- EASY SETUP — USB-powered connect to your computer for quick and simple scanning; No batteries or external power supply required to operate portable document scanner; Standard Connectivity: USB 2.0
Typical candidates include invoice ID, invoice date, due date, purchase order, supplier or customer details, bill-to and ship-to addresses, total amount, tax, amount due, and line items. This is a starting list, not a universal standard: providers expose different fields, and your accounting process may require a narrower or broader schema. Microsoft’s invoice output, for example, includes invoice-specific fields and line items such as invoice ID, bill-to, ship-to, customer and total.
For every field, specify its format and what counts as correct. Decide, for example, whether dates should be normalized to a single representation, whether currency codes must be captured separately from amounts, and how missing or ambiguous values should be represented. These rules make field-level evaluation meaningful.
Rank #2
- FAST SPEEDS - Scans color and black and white documents a blazing speed up to 16ppm (1). Color scanning won’t slow you down as the color scan speed is the same as the black and white scan speed.
- ULTRA COMPACT – At less than 1 foot in length and only about 1. 5lbs in weight you can fit this device virtually anywhere (a bag, a purse, even a pocket).
- READY WHENEVER YOU ARE – The DS-640 mobile scanner is powered via an included micro USB 3. 0 cable allowing you to use it even where there is no outlet available. Plug it into you PC or laptop and you are ready to scan.
- WORKS YOUR WAY – Use the Brother free iPrint&Scan desktop app for scanning to multiple “Scan-to” destinations like PC, Network, cloud services, Email and OCR. (2) Supports Windows, Mac and Linux and TWAIN/WIA for PC/ICA for Mac/SANE drivers. (3)
- OPTIMIZE IMAGES AND TEXT – Automatic color detection/adjustment, image rotation (PC only), bleed through prevention/background removal, text enhancement, color drop to enhance scans. Software suite includes document management and OCR software. (4)
3. Plan for layout variation
Invoice extraction is not just word recognition. A system must interpret what text means and where it sits: a total near the bottom of one invoice may appear elsewhere on another, while tables and multi-page documents add relationships that plain text alone can lose. Google Research describes the task as spanning natural-language processing and computer vision, with a pipeline that uses OCR text and layout, proposes candidate spans for schema fields, scores them and selects likely values.
Invoices can vary between suppliers and even between departments within one organization. Consider two broad approaches:
Rank #3
- FAST DOCUMENT SCANNING — Document scanner with feeder allows you to speed through stacks with a 50-sheet Auto Document Feeder (ADF); Efficient office scanner to help you scan more productively
- INTUITIVE, HIGH-SPEED SOFTWARE — Quickly scan with this desktop document scanner; Epson ScanSmart Software lets you easily preview scans, email files, upload to the cloud, and more; Plus, automatic file naming saves even more time
- SEAMLESS INTEGRATION — Easily incorporate your data into most document management software with the included TWAIN driver; Office document scanner integrates seamlessly with business workflows
- EASY SHARING — Duplex scanner allows you to scan straight to email or popular cloud storage2 services like Dropbox, Evernote, Google Drive, and OneDrive for simple storage and sharing
- SIMPLE FILE MANAGEMENT — Scanner allows the creation of searchable PDFs with Optical Character Recognition (OCR) and convert scans to editable Word or Excel files effortlessly; Designed for home and office document scanning
- Template-driven extraction: useful when formats are stable and known. Layout changes may require maintaining or revising templates.
- Learned or managed extraction: may handle a wider range of layouts, but its actual performance depends on the fields and documents you use. Do not assume it will outperform templates for every supplier mix.
Microsoft also notes that invoices can arrive in varied formats and quality. Compare options using your own document mix rather than a single clean sample.
4. Validate values against evidence and workflow rules
A structured response is not automatically a verified accounting record. Where available, preserve the connection between an extracted value and its evidence on the page. Microsoft’s documented output separates recognized text, page-level tables and cells—including bounding boxes and confidence—from invoice-specific fields and line items. That separation can help an operator inspect the source region when a result looks wrong or uncertain.
Rank #4
- Scanner type: Document
- Connectivity technology: USB
- With Auto Scan Mode, the scanner automatically detects what you're scanning
- Digitize documents and images
Apply checks at the field and document level. Google Research gives a domain-specific example: an invoice date should precede a payment date. Depending on your workflow, useful checks can include whether totals reconcile with line items and tax, whether dates are plausible, and whether a purchase-order reference matches the corresponding record. These checks are workflow guidance, not rules that a particular vendor requires.
Route uncertain or high-impact values to review rather than silently accepting them. A confidence score can help prioritize attention, but it is not proof of correctness. Keep the extracted value, source location and any correction together where your system permits; this makes errors easier to diagnose and corrections easier to audit.
Best Value
- OUR MOST ADVANCED SCANSNAP. Large touchscreen, fast 45ppm double-sided scanning, 100-sheet document feeder, Wi-Fi and USB connectivity, automatic optimizations, and support for cloud services. Upgraded replacement for the discontinued iX1600
- CUSTOMIZABLE. SHARABLE. Select personalized profiles from the touchscreen. Send to PC, Mac, mobile devices, and clouds. QUICK MENU lets you quickly scan-drag-drop to your favorite computer apps
- STABLE WIRELESS OR USB CONNECTION. Built-in Wi-Fi 6 for the fastest and most secure scanning. Connect to smart devices or cloud services without a computer. USB-C connection also available
- PHOTO AND DOCUMENT ORGANIZATION MADE EFFORTLESS. Easily manage, edit, and use scanned data from documents, receipts, photos, and business cards. Automatically optimize, name, and sort files
- AVOIDS PAPER JAMS AND DAMAGE. Features a brake roller system to feed paper smoothly, a multi-feed sensor that detects pages stuck together, and skew detection to prevent paper damage and data loss
5. Evaluate each field on unseen layouts
Measure performance field by field, using a representative sample of suppliers, image quality and document types. Include layouts that were not used to build or tune the system: performance on familiar templates can conceal failures when a new supplier or format appears. Track errors that matter to your process, such as a wrong invoice ID, a missed tax amount or an incorrect due date, rather than relying on one overall accuracy figure.
Historical studies illustrate why results need context. Google Research’s 2020 article reports F1 scores on its internal test set of invoice layouts disjoint from its training and validation sets:
| Field | F1 score |
|---|---|
| amount_due | 0.801 |
| delivery_date | 0.667 |
| due_date | 0.861 |
| invoice_date | 0.940 |
| invoice_id | 0.949 |
| purchase_order | 0.896 |
| total_amount | 0.858 |
| total_tax_amount | 0.839 |
The authors noted that delivery_date appeared in only a small subset of training examples. The scores show that results can differ by field even within one study; they are not current product benchmarks or predictions for your invoices. See Google Research’s 2020 account for the study context.
A 2018 paper by Holt and Chisholm reports an average accuracy of 92% across field types on unseen documents and a median prediction latency of 3.8 seconds for its evaluated system. It also reports an absolute accuracy gain of 20% across compared fields and a 25%–94% reduction in extraction latency in its comparison. These are results from that paper’s system and comparison, not expected performance for current services. Its evaluation on unseen document formats reinforces the value of testing layouts beyond those used in development. See Holt and Chisholm, “Extracting structured data from invoices” (2018).
For a practical evaluation, keep a labeled set of real invoices separate from any data used to configure or train the system. Compare field-level precision and recall or F1 where appropriate, record the share of values that require human correction, and review results by supplier or layout. There is no apples-to-apples provider ranking established by these studies; your own representative test is the useful comparison.
Quick Recap
Putting the decisions together
- Inventory the PDFs and images you receive, distinguishing text-based PDFs from scans and photographs.
- Define the required fields, types, formats and handling for missing or ambiguous values.
- Choose an approach suited to the amount of layout variation, then test it on multiple suppliers and multi-page cases where relevant.
- Preserve source evidence and add field, arithmetic and workflow checks before values enter accounting records.
- Measure field-level results on held-out invoices and unfamiliar layouts, then direct review toward weak or high-impact fields.
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
