What is the Classification module?

The Classification module in Supplier Data Manager (SDM) automatically assigns products to categories and families based on your predefined hierarchical structure. The AI Classification version uses AI to analyze product data and classify products accordingly, reducing manual effort.

How does the Classification module work?

After mapping your supplier's data fields to your internal structure (see Mapping module), the SDM Classification module uses AI to evaluate each product and assign it to the most appropriate category or family. Most products are classified automatically, but reviewing the AI predictions ensures accurate classification.

The Classification module presents products across three tabs:

• Rows to Check (red): Products that need classification or manual verification, including those the AI could not classify automatically.
• Rows Checked by the AI (blue): Products the AI classified automatically without requiring human review.
• Rows Checked by You: Products you have manually validated.

You configure whether classification runs in automated or to check mode:

• Automated mode: The AI classifies products automatically. Only products the AI could not classify appear in the Rows to Check tab for manual review. Products the AI classified appear in the Rows Checked by the AI tab.
• To check mode: All product rows require manual verification, even when the AI found a match. All products appear in the Rows to Check tab.

The screenshot above shows the SDM Classification module with the Rows to Check (red) and Rows Checked by the AI (blue) tabs, illustrating the two classification review states.

How to manually classify your products

In the SDM Classification module, you can manually classify products by assigning one or more categories or families.

1. Select the row(s) you want to classify.
2. Review the suggested category. If incorrect, use the search function to find a more accurate one.
3. Confirm the correct category.
4. Once all rows to check are validated, click Finalize to proceed.

The screenshot above shows the manual classification interface in SDM, with product rows listed, a category search field, and the Finalize button to confirm and proceed to the next step.

How does AI classify products?

AI Classification in Supplier Data Manager uses pre-trained large language models (LLMs) to classify products into categories or families based on their attributes. The models do not require additional training or feedback loops — they perform accurately out of the box.

Using LLM models for classification offers:

• Reduced time-to-value: Save up to 75% of the effort previously required for training-related tasks.
• Wider compatibility: Supports multi-language classification (for example, Japanese, German, and Arabic).
• Improved accuracy: Tested across diverse datasets to ensure robust performance in real-world scenarios.

Tailor the AI to your context with custom prompts

You can add custom instructions to the SDM Classification step to refine how the AI classifies your products. By default, the LLM has no context-specific knowledge about your company, category names, or product taxonomy — a custom prompt lets you supply that context.

To add a custom prompt to the Classification step in SDM:

1. Go to your workflow > Settings > Steps.
2. Select the relevant workflow and the classification step you want to customize.
3. Enter your instructions in the text box provided.

How to write an effective prompt

Custom prompts help refine classifications by improving accuracy rather than enforcing strict rules. By supplying relevant information and examples, you guide the AI toward more precise results.

To write an effective prompt:

• Ensure sources contain sufficient information, such as product names and descriptions, which often include key classification details.
• Provide examples to improve accuracy. The more examples you give, the more reliably the model follows your instructions.
• Use labels rather than codes. Codes must be explicitly defined in the prompt for the model to interpret them; labels are understood directly.
• Provide the full category path. For example, instead of "Wine fridge", use the complete path: "Kitchen > Large appliances > Refrigerators > Wine fridge".
• Avoid instructions that contradict the model's knowledge. For example, if a product description clearly states it is an oven, instructing the AI to classify it as a refrigerator will consistently fail.

Does the AI replace existing classification values?

By default, the SDM Classification module provides classification suggestions for all products, including those that already have family or category data in the source file.

If your source file already contains family or category information, you can configure the module to skip AI classification for products that already have values. This prevents unnecessary predictions on already-classified data, saving processing time and preserving data integrity. The AI only classifies products where this information is missing.

To set up this option:

1. Ensure the family and category fields are included in the Mapping step and correctly mapped to columns in the source file. The values in your file must be valid family or category codes.
2. Add additional configuration to the classification step by setting the replace_existing option to false. This change is made in the SDM admin panel.

Common questions

How does the AI Classification module understand product data? LLMs are trained on large amounts of text, enabling them to interpret and classify product attributes from the start. SDM uses AI-powered models to provide accurate classification without requiring upfront training on your catalog.

What if the AI makes a mistake? LLMs are designed for high accuracy but are not infallible. Misclassifications can be corrected during the classification step using the manual review interface.

Can the AI learn from corrections? No. The SDM Classification module does not use feedback loops. Custom prompts are the primary way to guide the AI toward better results for your specific context.

How long does setup take? AI Classification is the default setup in SDM and can be used immediately.

Is AI Classification available for all customers? Yes, with the exception of scenarios requiring highly customized taxonomies or where the technology is unsuitable.

Limitations

• Maximum number of products (rows) per job: 30,000
• Recommended maximum number of attributes as AI sources: 50 (using more than 50 attributes may reduce accuracy)