The Mapping module in Supplier Data Manager (SDM) aligns columns from supplier files with your internal taxonomy, making it straightforward to integrate supplier data into downstream systems such as your PIM or ERP. You can complete mapping yourself or delegate it to suppliers through their supplier portals.

What is the Mapping module?

The Mapping module in SDM is the step where you connect the columns in a supplier's file to the fields in your taxonomy. Once fields are mapped, SDM remembers the configuration for future imports from the same source — so you only need to map once per supplier.

• Required fields must be mapped before you can proceed to the next step.
• Optional fields can be left unmapped if the data is not needed.

How the Mapping module works

When a job reaches the Mapping step in SDM, the module reads the column headers from the supplier's file and compares them against your taxonomy. Fields that match a previous mapping are pre-filled automatically. Any remaining fields appear as unmapped and must be handled manually or with AI assistance.

Key features

• Automatic mapping — SDM pre-fills fields based on mappings learned from previous imports.
• Manual mapping — For unmatched fields, you can select the correct source column from a dropdown.
• Visual indicators:
  • Red asterisk — Required field. Must be mapped before proceeding.
  • Green checkmark — Field successfully mapped.
  • Gray bubbles — Optional field. Can be left unmapped.
• AI Source indicator — Marks fields that feed into AI-powered steps later in the pipeline (such as Extraction or Classification). Knowing which fields are AI sources helps you prioritize accurate mapping for those fields.

The screenshot above shows the Mapping module interface in SDM, with fields listed in the center panel. Required fields are marked with a red asterisk, successfully mapped fields show a green checkmark, and optional fields appear with gray bubbles. The AI Source indicator highlights fields used by AI models in subsequent steps.

AI-powered mapping suggestions

What is AI mapping?

AI mapping in SDM automatically suggests which columns from a supplier's source file should be mapped to each target field. Instead of reviewing every column manually, the AI analyzes the source data and proposes the most likely matches — so you can validate or adjust suggestions rather than start from scratch.

How AI mapping works

When you click Generate AI mapping in the Mapping module, the AI examines:

• Column names from the source file
• Sample values from each column (actual data rows)
• Target field names, types, and descriptions from your taxonomy

Based on this analysis, the AI suggests a source column — or a combination of columns — for each unmapped target field.

Fields that are already mapped are left untouched. Only unmapped fields are analyzed.

The screenshot above shows the AI mapping flow in SDM after clicking Generate AI mapping. Each suggestion appears alongside its confidence score, and already-mapped fields are unchanged.

Confidence scores

Every AI suggestion includes a confidence score between 0% and 100%, indicating how certain the AI is of the match. Suggestions below 70% are automatically excluded — the AI only surfaces matches it considers plausible.

The confidence score appears alongside each suggestion in the Mapping module interface, so you can quickly decide whether to accept, adjust, or ignore it.

Why a field may not receive a suggestion

The AI may not suggest a match for every field. This can happen when:

• No source column is similar enough (confidence would be below 70%)
• The field is already mapped from a previous import
• The target field's name or description does not have a clear counterpart in the source file

In those cases, the field remains unmapped and you can fill it in manually.

Tips for better suggestions

AI mapping in SDM performs best when:

• Column headers in the source file are descriptive (for example, product_name rather than col_A)
• Sample data is present and representative of the full dataset
• Target fields have clear names or descriptions in your taxonomy

If a source file uses generic or cryptic column names, the AI may produce fewer or lower-confidence suggestions, and manual mapping will be more reliable.

Limitations

• Column limit — AI mapping supports source files with up to 2,000 columns. If a file contains more than 2,000 columns, AI mapping is not triggered. You need to map fields manually in that case.

Frequently asked questions

How does the AI understand our data?

The AI is LLM-based and interprets column names, sample values, and target field descriptions to identify the most likely matches. It does not retain your data between sessions — each mapping request analyzes only the current file and your taxonomy.

Can the AI map a single target field to multiple source columns?

Yes. If data is physically split across multiple source columns (for example, a first name column and a last name column that should combine into a full name field), the AI can suggest mapping multiple source columns to a single target field.

What happens if all AI mapping batches fail?

If the AI is unable to generate suggestions — for example, due to a service interruption — the Mapping module remains fully functional for manual mapping. No data is lost and your existing mappings are preserved.