Table of contents :

How to Prepare a WooCommerce Product Catalog

100-days-of-amazon-personalize-woocommerce-031

Table of contents :

A recommendation model is only as useful as the product catalog it can interpret. In a WooCommerce implementation, catalog preparation is not a one-time export of product names and prices. The catalog must provide stable identifiers, meaningful metadata, accurate availability signals, and a repeatable synchronization process so Amazon Personalize can connect product attributes with customer interactions.

Define the item represented by each catalog row

Start by deciding whether an Amazon Personalize item represents a parent WooCommerce product, a product variation, or both. This decision must match the item identifiers recorded in interaction events such as views, cart additions, and purchases.

  • Parent-product model: Use the WooCommerce product ID when recommendations should operate at the product level and variations are mainly pricing or fulfillment choices.
  • Variation model: Use the variation ID when size, color, material, or another variation attribute materially changes the recommendation.
  • Hybrid model: Use separate identifiers only when the application can reliably translate recommendations back to a useful storefront experience.

For most catalogs, the variation model is appropriate when customers browse or purchase specific variants. The parent-product model is simpler when users think of all variations as one product. Do not mix parent IDs and variation IDs without a documented mapping, because the model may learn interactions against items that cannot be rendered consistently in the storefront.

Create stable and unique item IDs

The ITEM_ID field is the link between the WooCommerce catalog and Amazon Personalize interaction data. Use a stable identifier that does not change when a product title, slug, or price changes. WooCommerce database IDs are usually better candidates than URLs or post titles.

A catalog export might contain values such as product-1842 or variation-9321. Use the same exact value in interaction events:

{"userId":"customer-447","itemId":"variation-9321","eventType":"purchase","sentAt":"2025-02-14T15:21:00Z"}

Keep the identifier format consistent across historical imports, real-time events, recommendation requests, and catalog updates. If a migration requires changing IDs, build an explicit translation layer and decide how historical interactions will be handled before sending the new data.

Select metadata that supports recommendation quality

Amazon Personalize can use item metadata alongside interaction history. Export attributes that describe how products differ from one another and that are available reliably across a meaningful portion of the catalog.

Common WooCommerce fields include:

  • Product name or a normalized title.
  • Product category and subcategory.
  • Brand, manufacturer, or collection.
  • Product type.
  • Material, style, compatibility, or use case.
  • Price or price band.
  • Release or creation timestamp.
  • Availability and catalog status.
  • Season, department, audience, or merchandising group.

Do not export every WordPress custom field automatically. Internal notes, supplier references, temporary campaign labels, and highly sparse attributes can add noise or create unnecessary schema maintenance. Work with the merchandising team to identify fields that influence product discovery and purchasing decisions.

Map WooCommerce fields to an Amazon Personalize item schema

Create an item dataset schema with ITEM_ID as the required identifier and add only the fields needed by the use case. Classify each metadata field according to its actual meaning. A category, brand, or product type is categorical data; a numeric price or rating should remain numeric rather than being exported as formatted text.

For example, a logical source record might be mapped as follows:

ITEM_ID,PRODUCT_TYPE,BRAND,CATEGORY,PRICE,IN_STOCK,CREATION_TIMESTAMP
variation-9321,hoodie,Northwind,

Trending posts
You might also like