Skip to content

Classification Setup

Last updated: 03.09.2025 • Reading time: ~6 minutes

In order for Smart Processing to execute different processes for each document type, incoming documents must first be classified. Classification is the first step in the entire processing workflow and is configured on the Classification Setup page.

Note

A clean pre-sorting of incoming documents, for example, only invoices per incoming list, reduces complexity and costs. In such scenarios, an automatic classifier can be omitted.

Open Classification Setup

To start the configuration, proceed as follows:

  1. Start in the Smart Processing Setup area.
  2. Open the Classification Setup section.
  3. Select the desired Classification Method and configure the method-specific settings.

Available Methods

None (no Classifier)

  • No automatic classification is performed.
  • Suitable for pre-sorted incoming channels, where, for example, only invoices are present.
  • Advantage: No additional setup or service effort.

Document Intelligence

  • Classification is performed using models in Microsoft Document Intelligence.
  • Supports custom models that can be trained on specific layouts.
  • The training of the models takes place on the Document Intelligence platform, not directly in Business Central.
  • Recommended when document layouts are stable or a custom model is needed.

Content Understanding

  • The simplest way to classify documents directly in Business Central.
  • Classifiers can be managed and created without external tools.

Setup Steps:

  1. Select Content Understanding Classification under Classification Method.
  2. Open the action menu Classification → Manage Classifier.
  3. Click on Create Classifier Model to start the wizard.
  4. Define the desired classifications, such as invoice or credit note, and provide descriptions to improve recognition.

Rule-Based Classification (Regex)

  • Uses predefined rules based on Regular Expressions (Regex) applied to the OCR text.
  • Requires an activated Document Intelligence Service for OCR.

Setup:

  1. Select the method Rule-Based Classification.
  2. Now open a desired workflow.
  3. Enter your rules there.

Rule Definition:

  • Regex Pattern – Regular expression searched in the document content.
  • Rule Description – Description of the rule and its purpose.
  • Confidence Factor – Weighting of the rule for the result calculation.

Confidence Calculation: Sum of the factors of the applicable rules ÷ Sum of all factors. Example: If Regex1 (1) and Regex2 (2) are configured and only Regex2 applies → 2 ÷ (1+2) = 66%.

Note

The Rule-Based method is particularly suitable for standardized, layout-based documents such as PDF files or scanned images. An overview of supported formats can be found in the article Smart Processing Setup.

XML Classification Mapping (ZUGFeRD / XRechnung)

For XML documents, classification is based on the TypeCode.

Standard Mapping (Excerpt):

Invoice 380,384,386,388,389,393,395,575,623,780,870,875,876 (e.g., 380 – Commercial Invoice (Default), 384 – Corrected Invoice, 388 – Tax Invoice, 389 – Self-billed Invoice, 393 – Factored Invoice, 395 – Consignment Invoice, 575 – Insurer’s Invoice, 623 – Forwarder’s Invoice, 780 – Freight Invoice, 870 – Consular Invoice, 875/876 – Construction Invoice)

Credit Note 381,81,83

Debit Note 383,80,84

Adjust Mapping:

  1. Go to Smart Processing Setup.
  2. In the Classifications area, select the action XML Classification Mapping.
  3. Enter the desired Classification and the corresponding Type Codes (separate multiple codes with a comma).

Important

A Type Code may be assigned to only one Classification. If assigned multiple times, an error message will be displayed.

Split Handling

Split handling is available for Document Intelligence and Content Understanding.

Options:

  • None – No split of documents.
  • Per page – Each page is classified individually.
  • Auto – Multi-page documents are automatically divided into sections, e.g., pages 1–3 invoice, pages 4–5 another document.

Configuration:

  • In Document Intelligence, freely adjustable in Smart Processing Setup → Split Handling.
  • In Content Understanding, the option is set in the Classifier Wizard. Changes require an update of the classifier.

Handling Multiple Classifications

When multiple classifications are detected, the setting Handle Multiple Classifications controls the behavior:

Option Behavior
Use Highest Confidence The classification with the highest probability is selected.
Static Classification A fixed classification is used. If none fits, manual selection is required.
Manual Classification The user manually selects the classification.

Next Steps