Smart Data Modeler - Traditional channels

Smart Data Modeler analyzes an uploaded product catalog sample and suggests a structured Product Type model with Attributes. You can review, refine, and import the proposed model before using it in your Project.

When to use Smart Data Modeler

Smart Data Modeler is useful when:

• You want to create Product Types and Attributes from existing catalog data.
• You want suggestions for whether Product Attributes should be defined at the Product or Variant level.
• You want to review a proposed Product data model before creating it in your Project.

Smart Data Modeler works best for simple product catalogs, such as those in retail, rather than for complex scenarios like bundles or composite products.

Intended workflow

1. Open Smart Data Modeler from Settings > Product types and attributes.
2. Upload your existing product catalog. Supported formats are csv, json, and xlsx. Use the original xlsx or json export if available.
3. Provide business context. Before adding further instructions, select one of the predefined options to help the agent tailor the proposal.
4. Add optional instructions, such as required Variant Attributes, localization needs, and Attributes to exclude.
5. Generate and review the suggested Product Attributes. You can select which generated Attributes to include before accepting the proposal.
6. Confirm the list and continue the analysis to review the suggested Product Types and Attribute grouping.
7. Import the proposal to create the proposed Product Types in your Project, or download the proposal as a json file for later use.

Agent instructions

You can add custom instructions to tailor the agent's suggestions to your business case. For example:

• Product detail: "Product detail pages show brand, material, and care instructions."
• Search and filtering: "Customers must search by brand and filter by color and material."
• Downstream systems: "The ERP requires style_code on every apparel variant."

Custom instructions also let you work around known limitations, such as excluding pricing and analytics fields or defining Product-level and Variant-level Attributes.

Prepare your data sample

Provide a small, clean, representative sample that shows structure, not a full export.

Include:

• Representative products for each distinct product family or Category root.
• Unique identifiers to group Product Variants, such as Product key and Product name.
• SKU if it is the only reliable Product Variant identifier, for example when there is no Product key or other Variant grouping column.
• Differentiators, such as color, material, size, voltage, fit, brand, and format.
• Evidence of constraints, such as repeated identical values for sameForAll and unique combinations for combinationUnique.
• Complete or representative enum value sets.
• Multi-value examples where a Set type is needed.

Exclude or trim:

• Long marketing prose, FAQs, and promotional copy.
• Internal-only utility or audit fields.
• Pricing and core platform fields already modeled, such as price, slug, and sku.
• Fields only relevant to downstream systems if they are not needed for discovery, pricing, compliance, or checkout.
• Localized columns you do not plan to localize.

Limitations

The following limitations apply:

• The maximum file size for the uploaded file is 4 MB. Remove non-essential columns, such as category paths, images, stock, and prices, to reduce the file size.
• For large catalogs, run Smart Data Modeler separately per product family to stay within limits.
• Attribute search must be reviewed before adoption. Mark an Attribute searchable only if it must drive user search or filtering.
• The maximum length for a searchable Attribute value is 10,922 characters.

Test the Product Type model before adoption

• Review and adjust keys, enum completeness, and constraints.
• Validate results in a staging Project first.