FAQ about Smart Recommender

Prev Next

Product Catalog and Integration

Does Smart Recommender work for non-retail verticals?

Yes. Smart Recommender can be used for recommendation use cases beyond traditional retail product catalogs, including industries such as airlines and banking.

The standard Smart Recommender experience is designed around retail-style product catalogs. For non-retail use cases, the recommendation setup may require additional configuration or a tailored implementation based on the type and structure of the data available.

For example, airline use cases can include recommendations for fares, ancillaries, or other travel-related offerings, while banking use cases can include recommendations for financial products or services.

If you are implementing Smart Recommender for a non-retail use case, contact the Insider One team to determine the appropriate recommendation setup for your data and use case.

How do you collect product catalog data?

Insider One provides four different ways of Data Collection:

  1. Click-Stream (ProductFeedETL): Captures data automatically as users browse pages.

  2. XML: Scheduled uploads (hourly to weekly) via an XML file hosted on your end. Best for stable catalogs managed via ERP/PIM.

  3. Catalog API: A REST API for real-time synchronization. Best for large, frequently changing catalogs.

  4. Shopify: A direct, automated sync for stores built on Shopify.

What is the update frequency of the product catalog?

  • Catalog API: Updates are real-time and immediate.

  • Shopify: Updates are near real-time.

  • XML: Dependent on the synchronization period you selected (ranging from 1 hour to 5 days).

  • Clickstream: Updates occur based on user visits. Popular products update faster than less visited ones.

How are "Out of Stock" items handled?

  • API/Shopify: Updates are instant.

  • XML: Updates occur during the next scheduled sync.

  • Clickstream: Uses Product Hit Check. If a product receives no views during a defined window, the system marks it as out of stock automatically.

Why did my product update fail or get rejected?

A product is rejected if it's missing a required attribute (name, item ID, image URL, or price) or if a default attribute fails validation, for example, an incorrect data type or a value longer than its character limit (such as a description over 1,024 characters).

If a custom product attribute fails validation (it isn't defined, has the wrong data type, or is too long), the product is still ingested, but that attribute is skipped.

To see which products were rejected and why, open Catalog Settings > Integration Health Center and check the Error Report.

Can I rename or change the data type of a custom product attribute?

No. Once a custom attribute is created, you cannot rename it or change its data type (e.g., from Text to Number). You must deactivate the old attribute and create a new one with the correct configuration.

What is the character limit for product attributes?

Most string fields (like name or brand) have a limit of 512 characters. However, description and category fields can hold up to 1,024 characters, and the tags array can hold up to 4,096 characters. You can visit the Product Attributes page to see the character limits.

What happens if I update an active XML file?

You do not need to pause campaigns. The system automatically sets the file to "passive" and begins validation. During this time, the catalog continues to use the last valid data. Once validated, it becomes active automatically.

Do I need a separate integration for each country/language?

Yes. You must establish a product catalog integration for each Locale (e.g., en_US, tr_TR). You can also configure Stores under locales (e.g., en_US:NewYork) to manage inventory segmentation by warehouse or region.

Recommendation Widgets & Campaigns

What does the "Test" status mean for a campaign?

When a campaign status is set to Test, it is not visible to regular visitors. It can only be viewed by generating a Test Link from the InOne panel, which allows you to preview the widget on your site before going live.

Why is my Smart Recommender widget not showing any products?

Common reasons include:

  • Filters are too strict: No products match the criteria (e.g., filtering for "Red" shoes in a category with no red shoes).

  • Stock Issues: The matching products are out of stock.

  • Algorithm Health: The selected algorithm in your strategy lacks sufficient data (check the "Prerequisites" on the Recommendation Algorithms page).

  • Integration: The Insider Tag or Product Catalog integration is incomplete.

What is the difference between Inline and Overlay widgets?

  • Inline: Embedded directly into the page layout. Best for seamless browsing.

  • Overlay: A popup layer. Best for capturing attention before a user leaves.

Why is my widget width not changing in the editor?

The widget's width is inherited automatically from the HTML container on your website where it was appended. You cannot force a width wider than the container it sits in. You can only adjust the spacing and number of cards within the available width.

Recommendation Strategies

You can see the rules each strategy follows and test them:

  • Strategy details: Open the strategy to see how it selects products, including its algorithms or bundle slot settings, filters, exclusions, and fallback settings where applicable.

  • Recommendations Preview: See a sample of the products the strategy returns for the inputs you select. For strategies that use browsing history, test with a known user who has browsing history. The preview is a test, not a record of what a shopper saw.

  • Algorithm health: In Components > Recommendation Algorithms, select a locale to see whether each algorithm has the data it needs. Click an algorithm’s name to see which prerequisite is missing.

  • Results: Smart Recommender Analytics shows performance by campaign, campaign variant, product, and category. In Product Metrics, click a product to see which algorithms recommended it and generated revenue.

Can I recommend items that aren’t products, such as offers, events, articles, or plans?

Yes. Add them to your Product Catalog as items. Each item needs a unique item ID, a name, an image URL, a destination URL, and a locale. Additional required fields depend on your catalog feed type. For example, a Published-Time Based feed for articles requires a publication or update timestamp, while a Start–End-Time Based feed for events or time-limited offers requires start and end dates. Neither feed type requires a price.

To recommend only one type of item, create a custom text attribute such as content_type, enable its Filterable toggle, and give each item a value such as Offer, Event, Article, or Plan. Then add a Static filter to your recommendation strategy. For example, content_type + is + Offer limits the results to items with that value.

How fresh are recommendations?

It depends on the algorithm and where you use it. For example, Live Intent can adapt recommendations during the same browsing session based on what a shopper views. In Architect, user-based recommendations refresh once a day between 00:00 and 06:00 UTC. Architect uses those recommendations to build the message before it sends.