Recommended construction sequence for PIM structure

It is a good idea to build AI Commerce PIM in the order of deciding on store navigation, product base fields, selection model, and only then the attribute structure. This way, SKUs, variants, inventory lines, contents, and mass changes do not model the same information in parallel ways.

Choose the right structure for your data

AI Commerce PIM uses Product out-of-the-box fields for basic product structure or business logic. These include SKU, name, URL, status, base price, tax category, weight, availability dates, balance, and Google feed visibility. Do not create an attribute for the same information.

Categories place products in store navigation, product listings, and commercial groupings. A product can belong to multiple categories. Size, color, material, brand, compatibility, and technical value are usually considered attributes unless they actually form the customer's navigation path.

Product options are common selection axes in a catalog, such as size and color. Use the product Variants and Inventory Lines structures when the selection affects SKU, barcode, inventory, or price.

Custom selections are additional product-specific customer selections before adding to cart. Use them when the selection is not a general product attribute or a stocked variant.

Attributes describe technical, filterable, or comparable features of a product. Attribute sets organize attributes into logical groups. Attribute families assemble attribute sets into a product type-specific filling structure and define mandatory attributes for each channel.

Product groups group separate products into the same commercial presentation. They are not a generic variant model.

Data transfer is for CSV exports, previewed imports, and bulk updates. Use it when you are working with a large set of products, categories, links, attributes, attribute values, product groups, or inventory data.

Open the right maintenance view

The canonical maintenance views of AI Commerce PIM are:

• Categories /categories for the store's category tree.
• Products /products for product basic fields, content, SEO, categories, inventory, and attribute values.
• Product options /products/options for common selection axes and values.
• Product Options , Inventory and Variants , and Custom Selections sections for product-specific structures.
• Attributes for the /catalog/attributes attribute register.
• Attribute sets /products/attribute-sets for ordered groups of attributes.
• Attribute families /pim/families for product type-specific attribute structure.
• Product groups /products/groups for grouping separate products.
• Data exchange /data-exchange for CSV profiles, imports and exports.

Design a category tree

AI Commerce PIM category tree does not need to be symmetrical. Typically, use 1–3 levels depending on the size of the catalog and the customer's navigation needs.

As a soft usability guideline, less than 10 categories can work on a single level, less than 100 as a roughly 10×10 structure, and less than 1000 as a roughly 10×10×10 structure. These are not technical limits. The goal is to keep a single level understandable, not to force exactly ten branches.

Create main and subcategories based on customer search patterns. At a minimum, check the name, URL section, description, ranking, visibility, SEO information, and Google feed settings.

Create the foundation for products

The first products in AI Commerce PIM are created using the Create New Product function in the Products view or imported in a larger set using the Data Transfer product profile. First, fill in the SKU, name, URL, status, price, tax category, basic texts, visibility, availability, and necessary images.

Keep the product's built-in field canonical whenever it exists. For example, SKU, product name, URL, price, tax category, balance, weight, availability date, or Google feed setting should not be copied as an attribute.

Assign the product to one or more of the correct categories. The choice of status and required information determines whether the product is ready for retail use; a draft, unfinished product should not be promoted to In Use status just to test the design.

Decide on choices, variants and groups

Before attributes, decide whether the customer's choice will result in a stockable option. If so, create a common Product Option, add its values, and create variants and SKU-level inventory lines for the product.

A stock line can contain a Stock SKU, barcode, balance, shelf, and price per line. The product option value identifies the variant; the stock change is stored separately.

If the selection is only an additional selection for this product, use Custom Selections. If separate products are presented as a single commercial entity, use Product Group. Do not use Product Group to replace Product Options and Variants.

Build attributes, sets, and families

First, create individual Attributes. Choose the appropriate field type for your data, such as text, text range, decimal, yes/no, choice, or multiple choice. Configure visibility in the store, filters, comparison, and PDF as needed.

Attach attributes to an Attribute Set in the order you want them in the product specifications. Use sets to group together dimensions, materials, certifications, and compatibility information, for example.

Attach Attribute Sets to an Attribute Family and select the required attributes for each channel. The family's set list forms the attribute structure for the products that use it.

Open the product and on the Attributes tab, select the correct Attribute Family. Fill in the groups that appear and select Save Attributes . Missing mandatory attributes do not prevent this saving, but they reduce the Mandatory Data fill rate and are visible in missing mandatory data checks.

Make mass changes with Data Transfer

In AI Commerce PIM mass change, select the appropriate profile for the task from Data Transfer and upload the CSV. Always run Preview , correct all line-specific errors, and confirm the execution with only the reviewed data.

The impact of import profiles varies. Some update individual rows, but for example, the channel-specific mandatory matrix of an attribute family is replaced as a whole with the channel columns imported. Read the profile description before executing and save the export for reference.

Do not import unverified CSV into production content. Preview is not a write, but execution can replace values from a wide range of products.

Summary

AI Commerce PIM is first built from categories and product fields. Next, decide on Product Options, Variants, Inventory Lines, Custom Selections, and possible Product Groups. Only then do you add Attributes, Attribute Sets, and Attribute Families.

Use Data Transfer for bulk changes only through preview. A single piece of data belongs to one canonical structure: navigation to category, ready-made business data to product's own field, selection to option or custom selection, and descriptive property to attribute.