Release Summary - Aug 10, 2023

ENG-25742

Enterprise Dashboard:

Attribute-Based Filtering for User View and Purchase History

We have introduced the ability to filter out products with matching attribute values based on a user's view and purchase history. As a digital merchandiser, you can now enhance recommendation strategies by preventing the recommendation of products that share the same attribute value as items the user has previously viewed or purchased.

The new API functionality allows you to specify the attribute name and a time frame within which the filtering should apply. If no time limit is desired, a value of -1 can be used.

This helps to ensure that product recommendations are more personalized and aligned with user preferences.

ENG-26360

Enterprise Dashboard:

Multi-Select for Guided Selling Questions

We have introduced the ability to make certain Guided Selling questions multi-select, enhancing the customization options for generating relevant recommendations. When configuring a Guided Selling screen, a new option "Allow multiple answers" has been added, specifically applicable to the Product Attribute question type. Once enabled, end-users can select multiple answers, effectively using checkboxes instead of radio buttons.

Note: This option is disabled and grayed out when the question type is not Product Attribute.

ENG-26312

Enterprise Dashboard / Find:

Fixed Search Test Drive Issues

We have resolved the issues with the Search Test Drive feature. The primary concern of not saving to production, which was only affecting integration, has been addressed.

Additionally, we have streamlined the user interface by eliminating unnecessary fields such as placements and languages that were causing confusion. Decimal value discrepancies have been rectified, enabling the successful addition of boosts. Notably, we've filtered out Engage placements from the dropdown, enhancing the user experience.

ENG-26370

Science:

ChatGPT Integration in Catalog Enrichment

In this release, we have introduced ChatGPT integration for Catalog Enrichment. This feature empowers optimization managers to enhance product attributes, aiding shoppers in finding the right products more effectively. By enabling ChatGPT through site configuration, users can leverage enriched product descriptions while maintaining cost efficiency. This integration streamlines the enrichment process by avoiding redundant submissions of product descriptions, and it allows for manual full catalog scans twice within a 90-day period. Additionally, attribute additions and manual scans are supported, providing a comprehensive solution for optimizing catalog enrichment efforts.

ENG-26211

Science:

Composite Outfit - fallback to product to category CP

Optimization managers can now provide outfits based on the probability that a product would be purchased or viewed with a category of products to increase the outfits presented.

ENG-25161

Science:

Composite Outfit - Update model approach

We have revamped the Composite Outfit generation approach to improve outfit relevance. Outfits are now curated based on both "purchased together" and "viewed together" scores, favoring purchased together suggestions by a factor of 5. Additionally, outfits are intelligently sorted by category considerations, enhancing customer engagement. Duplicate products, except for the seed product, have been removed from outfits, streamlining the shopping experience. This update ensures more appealing and tailored outfit recommendations for users.

ENG-25510

Science:

Composite Outfit - Design API to display outfits

In this release, we have designed a comprehensive API for displaying outfits, catering to different scenarios. Given a product seed, the API returns multiple outfits, ranked by their success score, including product IDs, part IDs, and style names. Outfits are de-duplicated, ensuring a unique user experience, and optional product de-duplication prevents repetition. The API supports style, category, brand, or no seed inputs, tailoring outfit recommendations accordingly. Notably, no merchandising rules are required, and outfits are created based on merchandising specifications, akin to Advanced Merchandising.

ENG-25511

Science:

Composite Outfit - Portal API to define and manage Styles

In this release, an essential enhancement has been introduced as part of the "Composite Outfit" feature. We have introduced a dedicated Portal API designed to define and manage styles for outfits. This addition empowers merchandisers by allowing them to create, customize, and organize styles with ease. Within this enhanced framework, merchandisers can establish styles, specify attributes, review and exclude products or outfits, and gain insights into their created styles. This streamlined process ensures a more efficient and intuitive approach to managing outfits, contributing to an improved overall merchandising experience.

ENG-26535

Recommend:

Log4j Update: Marathon Component Enhancements

We have made necessary updates to the Marathon components in response to a log4j update. These updates have been implemented and tested in our QA environment. The changes have been successfully completed and verified. Everything is functioning as expected.

ENG-20913

Recommend:

Create Recommend Rules with the image_url field

The optimization manager needed to create a do not recommend rule when the image URL is not a real image, to prevent it from being recommended.

Added an option to the ‘Do Not Recommend’ section to select Image URL and then identify the text that the image URL contains when it should not be recommended.

PLAT-3529, PLAT-3530, PLAT-3532, PLAT-3535

Platform > Streaming Catalog:

New enrichment components

We have introduced several new components, including enrichment calculation, enrichment subscription, enrichment ingestion, and enrichment view components in Enrichment Version 2. These components utilize the streaming catalog enrichment APIs for Enrichment Version 2.

PLAT-3567

Platform:

Enhanced Snapshot View with Subscription Sections

We have introduced a significant improvement to our streaming-snapshot feature, allowing users to gain deeper insights. By utilizing the new query parameters "sections=subscriptions" and "subscriptionState," you can now view subscriptions associated with a specific snapshot. When using "sections=subscriptions," the response will display active subscriptions, offering valuable real-time data. This feature works seamlessly for creating, complete, and active snapshots, enhancing your snapshot analysis capabilities.

PLAT-3573

Find:

Enable republish dataset by site via a REST call

We have enhanced the dataset republishing process by introducing a new capability: republishing datasets by site through a REST call. This improvement addresses an issue where the enrichment ingest API encountered errors due to a missing parameter in the calculation service HTTP call. As a result, dataset synchronization during republishing was incomplete.

This enhancement ensures the successful republishing of datasets, facilitating seamless synchronization between various components.

ENG-26588

Find:

e2e WOC feature with Enrichment V2

We have successfully integrated the end-to-end Wisdom of the Crowd (WOC) feature with Enrichment V2. This integration marks a significant transition of dataset management from calculation to the site.

ENG-26508

Enterprise Dashboard:

Placement Selection and Freezing Screen Issue

We have resolved an issue that prevented users from selecting placements and caused the screen to freeze while creating restriction rules. Previously, when attempting to select a placement during the creation of a Do Not Recommend Rule, the screen would become unresponsive and stuck in a loading state.

ENG-26405

Enterprise Dashboard:

Variation Duplication Issue in Social Proof Campaigns

We have resolved an issue where duplicating a Social Proof campaign/experience was failing to copy the associated variations. This bug resulted in incomplete duplication, impacting the smooth creation of campaigns. We have addressed this by ensuring that when duplicating campaigns, the variations are now fully copied, mirroring the entire experience without publishing the variations.

ENG-26560

Enterprise Dashboard:

Resolved Unexpected Error in Social Proof Item Page Template

We have fixed an issue that was causing unexpected errors in the item page template of Social Proof (SP). Users were encountering errors while interacting with the item page template. The issue has been fixed now.

ENG-26518

Enterprise Dashboard / Find:

Incorrect Data Source in "Search Test Drive"

We have resolved an issue where the "Search Test Drive" feature in the QA environment was fetching data from the staging environment instead of the intended QA environment. This issue occurred due to an incorrect internal API call that directed to the wrong URL. We have fixed this issue by ensuring that the "Search Test Drive" functionality correctly fetches data from the QA environment, with the API call now correctly directed to qa.richrelevance.com.

ENG-26296

Recommend:

Resolved Issue with "Do Not Recommend"

Rule in Configurable Strategies

We have resolved an issue related to the "Do Not Recommend" rule functionality within Configurable Strategies. Previously, the restriction rule "Products Viewed" was not functioning correctly, causing recommendations to appear for items that users had viewed in the past X days. With the necessary fixes in place, the restriction rule now operates as intended, ensuring that recommendations align with user behavior.

ENG-26504

Recommend:

API Publish Issue Resolution

We have successfully resolved the issue related to the API call for publishing on the search page. Previously, this API call exhibited a 500 error for certain clients, while it performed successfully for others. We have rectified the problem, ensuring that the publish API now functions as intended for all clients. The issue has been fully resolved and confirmed with a 200 OK status code in the API response.

ENG-26320

Recommend:

Affinity sorting seems apply to only the default configuration

When selecting User Affinity sorting, the recs should be sorted based on the User Affinity Configuration that was selected. However, it appears as though sorting is always based on the default configuration.

It’s possible that this is an expected behavior. It all depends on what products are in the base model. If there aren’t products from those categories in the base model, then we wouldn’t expect to see them in the affinity version. The affinity results are just the same products as the base model, but they are re-sorted based on the affinity score. It sounds like this is working as expected.

ENG-26544

Data Engineering:

Strategied Detailed report In QA throwing empty row with ‘0’ value

Records with strategy_id=-1 were filtered out from all the Site Analytics related reports, including the Strategies report, where the issue was observed. This issue was not visible in OnPrem as it was applying inner join for non-admin User. The issue has been fixed now.