Smart Recommender identifies and ranks the products in your Product Catalog by their item_id. When the item_id in your XML feed differs from the one your system rules collect, or when several products share one item_id, the algorithms cannot connect analytics data to the right product. As a result, recommendations fail or show products that do not reflect real performance.
This article explains how Smart Recommender uses item_id, the two most common item_id issues, and how to fix each one.
Before you begin, make sure you are familiar with the following articles:
This article covers the following topics:
How Smart Recommender uses item_id
Smart Recommender reads item_id from two sources:
XML feed: Adds and updates products in your Product Catalog.
System rules: Collect the item_id on your website and pass it with analytics events such as purchases and page views.
The algorithms use item_id to link these analytics events to catalog products, then identify and rank products such as top sellers, best matches, and category leaders. Insider One identifies each product by its item_id and locale together. For recommendations to work, two conditions must be true:
The same product has exactly the same item_id in your XML feed and in your system rules.
Each product and each variant has its own item_id within a locale.
When either condition fails, you see one of the issues below.
Issue 1: item_id mismatch between your XML feed and system rules
Symptoms
Smart Recommender shows no recommendations, or shows poor-quality recommendations that do not reflect real product performance.
Products update correctly through your XML feed but have no analytics history. These products are known as ghost products.
Cause
Analytics events are attributed to products by the item_id that system rules collect. When this value differs from the item_id in your XML feed, the events cannot be linked to the correct catalog product.
Impact
Ghost products have no analytics history, so the algorithms cannot score them.
The algorithms cannot correctly identify top sellers, best-matching products, or category leaders.
How to fix
Update your XML feed so that each item_id exactly matches the value your system rules collect and pass.
Allow time for the catalog sync and algorithm jobs to run before you check the result.
Go to the Recommendation Algorithms page and confirm that the feed is populating. For details, see Recommendation and Algorithm Settings.
Once the values match:
Analytics events are attributed to the correct catalog products.
Algorithms receive accurate data to generate and rank recommendations.
Top seller and category-based algorithms reflect actual product performance.
Issue 2: Several products share one item_id
Symptoms
A product shows another variant's price or stock.
Recommendations are less accurate.
Price Drop journeys use a price that belongs to a different variant.
Cause
Insider One identifies each product by its item_id and locale together. When you send two products with the same item_id in the same locale, Insider One stores them as one product. The values sent last replace the earlier ones, which affects every feature that uses your catalog.
How to fix
Give each product and each variant, such as a different color or size, its own item_id.
Give all variants of the same product one shared groupcode.
For more information on structuring products and variants, see Decide Your Product Catalog Structure.
Check your item_id setup
Use this checklist to prevent both issues:
Each item_id in your XML feed exactly matches the value your system rules collect.
Each product and each variant has its own item_id within a locale.
Variants of the same product share one groupcode.
Related articles
Aligned item_id values keep your catalog, analytics, and recommendations accurate. As a next step, see Handle Out-of-Stock Items.