Introduction
These release notes provide an overview of the resolved issues and improvements of the most recent release of the FAS 21.1 line.
FAS 21.1 supports Oracle Java 8 VMs. For detailed setup instructions please refer to Install Java instructions.
Version: 21.1.0
Release Date: January 22, 2021
FAS 21.1 is a major release. The Query and Data APIs in this version are backwards compatible with versions FAS 7.5.x, 8.x.y., 19.x.y, and 20.x.y. The Configuration API is backwards compatible with FAS 8.x.y, 19.x.y, and 20.x.y
Highlights of the release are AI-Powered Search capability, A/B tests for search profiles, improved drag-and drop user experience in Visual Merchandising, refined relational search, and multiple performance improvements.
Improvements & Resolved Issues
Search & Navigation
[New] AI-Powered Search is now available. It can use all the product data (text and images), as well as historical user behaviour to self-learn to provide more relevant search results to the user. AI-Powered Search is best suited in scenarios in which traditional keyword matching returns poor or no results. It can be added as an additional search pass, replacing either the fuzzy matching logic (SPELLCHECK-pass) or the partial match (OR-pass) on top of the existing exact match search capabilities. AI-Powered Search requires an additional license.
[Improved] [Open Beta] Relational search is refined. The functionality is behind a feature switch in system.xml and it is disabled by default. It is especially useful when item variants have multiple attribute values with similar localizations (names within the locale), and search is performed in the locale with words that are among the overlapping words in the similar values. The functionality has two modes:
- The first one ("top") is more lightweight and fixes exact match of the attribute value name in certain edge cases.
- The second one ("multi") does more sophisticated refinement but could have impact on the query response time.
[Improved] Response times for search queries on multi-variant items (relational search) are decreased.
[Improved] Performance when using conflated attributes is improved.
Insights & A/B Testing
Note: to use A/B testing, your platform needs to integrate with Fredhopper Insights.
[New] This release introduces A/B testing of Search Profiles. It is behind a feature switch in system.xml. By default, the feature is disabled.
- From the A/B tests page in the Merchandising Studio you can create an A/B test for Search profiles.
- A Search A/B test for two or more profiles can be created. The traffic among the rankings being tested can be allocated. Each alternative ranking can receive 10% to 90% of traffic. Search profiles can only be configured for a selected universe/locale, so the same applies to the Search A/B tests. They can only be created in a scope that has a universe and locale (no wildcard accepted), and there can only be one Search A/B test per universe/locale.
- Specifying a search profile in the FAS Query will overwrite any running Search A/B tests.
- After an A/B test is published, it is possible to review and analyze the results of a test in on the Fredhopper Insights page. This analysis enables the user to pick a winning test. Please refer to the Fredhopper Insights documentation for details.
- Once a winning option has been identified, the configuration of the winning Search profile can be applied to the default profile.
Visual Merchandising Interface
[Improved] The UX of drag-and-drop in Visual Merchandising is improved. Selection of multiple items is enabled.
[Improved] New look-and-feel for the items and cards is introduced.
[Improved] Performance improvement for long locations in the navigation in Visual Merchandising.
Merchandising Studio & Preview Pages
[Improved] Enhanced picking of locale for "all" languages/countries.
[Fixed] Result modification blocks all items when added on attribute with no values.
Known Issues & Limitations
The Search A/B tests cannot be previewed on the pre-published environment.
Kommentare
0 Kommentare
Bitte melden Sie sich an, um einen Kommentar zu hinterlassen.