FAQs: AI Semantic Search

Note: The feature previously known as AI Search is now referred to as AI Semantic Search. This update reflects a shift to a large language model (LLM)-based system.
The information in this document might not be accurate for both AI Semantic Search and the earlier AI Search RNN (legacy) version. For questions specifically about AI Search RNN (legacy), reach out to your Crownpeak contact.

This article will answer questions about our AI Semantic Search search pass add-on, including core functionality, requirements, implementation and A/B testing. 

What problem does AI Semantic Search solve?

AI Semantic Search enables customers to deliver more complete and accurate experiences for even the most complex and unstructured search requests. AI Semantic Search specifically addresses and solves long-tail searches, reducing the number of zero and low results searches, thereby increasing conversion. For more information, see Introducing AI Semantic Search.

 

How is AI Semantic Search different from FHR’s core search functionality?

AI Semantic Search is a complementary add-on solution to our core search functionality. It resolves unique and complex search queries that are not addressed with ID, Exact Match & Spellcheck search passes. You can find more information on how FHR Search and AI Semantic Search interact here.

 

Can AI Semantic Search be combined with existing search passes?

Yes, AI Semantic Search is a search pass add-on that slots into our Search Pass Orchestration System and can be used on top of the FHR core search functionality. You can find more information on how FHR Search and AI Semantic Search interact here.

 

What are the prerequisites for AI Semantic Search?

Fredhopper Access Server (FAS) 21.1 + Common Activities Data Pipeline (CADP) + Data Manager 3.0 or higher
For more information on how to prepare for using AI Semantic Search, see here.

 

What purpose does the Common Activity Data Pipeline serve?

As shoppers interact with your site, the AI deepens its understanding of your products, search terms, and customers to further optimize results over time. The Common Activity Data Pipeline (CADP) is needed to provide the model with the data it requires to enable potential re-trainings. Moreover, CADP enables FHR A/B testing as well as tracking the performance of AI Semantic Search against business KPIs.

 

Which languages does AI Semantic Search support?

AI Semantic Search currently supports English language catalogs. Multi-language support is coming soon.

 

A/B Testing 

What value does A/B testing for Search add?

A/B testing functionality for FHR Search and AI Semantic Search gives customers the ability to test and optimize search configurations and measure the impact these changes have on conversion. 

In FHR, search configuration is controlled within a search profile. Two separate search profiles can be configured. One profile will remain unchanged as a control. The other profile will use new or modified search passes: utilizing new technology, data, or weightings to complete the search.

 

What are the prerequisites for A/B testing for search?

Fredhopper Access Server (FAS) 21.1 & Insights + Common Activity Data Pipeline (CADP) + Installation of the FHR variant attribution model into your back end system (customer work required)

 

What is the effect on caching?

Caching is typically a blocker to A/B testing as if ‘A’ is cached, typically ‘B’ will not show. We’ve developed a solution to this, enabling customers to retrieve and store the running A/B tests so that multiple back end caches can be created for the pages running tests. A/B testing will still increase the cache, but only on pages that are being A/B tested and at most will double.

The only limitation to this is that some customers like to do CDN or other front-end caches. They can continue to do so, however, any Fredhopper request would have to be made available in their back end, where the FHR variant attribution model can assign the traffic.

For more information on the caching A/B testing integration, see here.

 

What trigger types are supported?

• Universe (applied to the parent scope/s)
• Locale (applied to the parent scope/s)

 

What trigger types are not supported?

• Keyword trigger
• Coverage trigger