Release Summary - Aug 26, 2024 (24.16)
The following key features and improvements, along with bug fixes, have been released in Algonomy CXP products in the release version 24.16 during Aug 09, 2024 - Aug 26, 2024.
Enterprise Dashboard, Recommend
User Affinity Sort Enhancements for Configurable Strategies
We've introduced enhancements to the User Affinity sort within Configurable Strategies, providing more control over how and when strategies are applied based on user history.
The new “Play Strategies Only If Reordered by Affinity” option allows strategies that are re-ranked by User Affinity or Size-based sorting to be played only if the user's history influences the results. This ensures that the strategy's performance is more accurately evaluated. If the user has no history, and this option is selected, no recommendations will be returned.
Social Proof Product Badging with Attribute-Based Messaging
We have enhanced the Social Proof Product Badging feature to support dynamic badge text based on product attributes. Merchandisers can now create more impactful and personalized messages for shoppers by incorporating attribute values directly into the badge text.
For example, you can now display a message like "40+ people rated this product 4 and above," where "40+" and "4" are dynamically populated based on the product's number of ratings and average rating attributes. This new capability allows for more relevant and engaging messaging, directly influenced by product attributes.
Jira: ENG-28744
Boost Recommendations Based on User Affinity
A new option has been added to the Recommendation Boosting rules, enabling digital merchandisers to boost products based on user affinity scores. This enhancement allows for the reordering of recommendations according to the affinity configuration selected, making it possible to apply affinity sorting across any recommendation strategy, not just those within Configurable Strategies.
Merchandisers can now select the "Based on User Affinity Scores" option under the Boost section of the User tab. A dropdown menu allows for the selection of the desired User Affinity Configuration, with the 'default' configuration selected by default. Additionally, there's an option to limit the number of products that are boosted, providing further control over the recommendation process.
Social Proof - Dynamic Messaging and API Integration for User Engagement
Merchandisers now have enhanced capabilities to deliver tailored and impactful messages through Social Proof, with the ability to define multiple message types based on user or event count ranges. For example, a product viewed by 10 to 50 users might display "Popular," while higher engagement levels such as 51 to 100 views could trigger "Hot Product." As engagement continues to rise, products with 101 to 200 views might be labeled "Today's customer favorite," and those with over 200 views might display "Flying off the shelves! Grab your own before it's gone."
This enhancement allows for flexible configuration both in the UI and via the Social Proof Output Response API. Users can set messages by a single threshold or across multiple user/event count ranges. The appropriate message is then dynamically displayed based on a product’s engagement level, ensuring that social proof messaging is relevant and compelling for customers. This feature supports both single and multiple message types, offering a more customized approach to customer interaction. The API integration further enables automated and scalable delivery of these dynamic messages, ensuring consistent and timely communication across various channels.
Find
Introduction of Hybrid Search in Find
We have introduced Hybrid Search within the Find. This functionality combines lexical (keyword-based) and neural (vector-based) search methods to deliver more relevant and effective search results. This new capability generates results from both traditional keyword matching and K-Nearest Neighbors based vector search, combining them to maximize relevance for the user.
Key benefits of Hybrid Search include:
-
Reduced Zero-Search Results: Vector search understands the semantic meaning of words, providing results even for terms not explicitly in the catalog.
-
Improved Relevance and Synonym Handling: A search for "running shoes" will also return related items like "sneakers" or "athletic trainers."
-
Effective Typo Handling: Vector search can find relevant items despite minor errors or misspellings.
-
Better Results for Long Queries: Longer search terms benefit from vector search's understanding of word relationships.
-
Multilingual Support: Built-in support for all Find languages, integrating with facets, boosts, redirects, and B2B functionality.
Our hybrid search leverages clients' catalog and historical queries to generate vectors via OpenAI. We combine the generated vectors with traditional keyword searches to enhance overall search relevance. Consultants must explicitly enable Hybrid Search for it to be active on a site (on our portal configurations page).
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 24.16 during Aug 09, 2024 - Aug 26, 2024.
Jira # |
Module/Title |
Summary |
General Availability |
---|---|---|---|
Enterprise Dashboard: Bundles Template in Dynamic Experiences
|
We have introduced a new Bundles Placement template in Dynamic Experiences, allowing optimization managers to easily add bundle placements to their sites for cross-selling complementary products. This template is fully customizable with various styles and configuration options. The Bundles Placement template includes settings for font styles, region-specific configurations, and placement options, ensuring that it seamlessly integrates with your site’s design and functionality. It leverages the code from the Boots bundles experience as a foundation, with adjustments to accommodate the new variables and requirements. |
26-Aug-24 |
|
|
Recommend: Enable Category Recommendations in recsUsingStrategy API
|
The recsUsingStrategy API now supports returning category recommendation data. When the "Enable Strategy to Recommend Categories" option is activated for a Configurable Strategy, and the appropriate parameters are included in the request, the API response will now provide detailed categoryRec information. This includes the category ID, name, link URL, and image URL.
This enhancement applies to both legacy and Configurable Strategies. |
26-Aug-24 |
|
Recommend: Filter products the user has already viewed in recommendations but not clicked |
Digital merchandisers now have the ability to filter out products that a shopper has frequently seen in recommendations but has not interacted with, ensuring that customers are exposed to fresh product options. This new filtering option, available under the Do Not Recommend (DNR) rules in the User tab, allows merchandisers to specify the number of times a product can be shown in recommendations within a set lookback period (e.g., 10 days). If a product has been recommended more than the specified number of times without being clicked, it will no longer be shown. Currently, this filter can be configured only at email placements level. The rule resets if a product is clicked, allowing it to be recommended again, ensuring that only relevant and engaging products are continuously presented to the shopper. |
26-Aug-24 |
Enterprise Dashboard: Customizable Start Screen for Guided Selling 2.0
|
Guided Selling 2.0 has been enhanced to offer greater customization options for the start screen, allowing it to be displayed in a different location from the main quiz and with a unique background image. Digital optimization managers can now set separate background colors, title, and description font colors specifically for the start screen. Additionally, text can be placed in an inner div with customizable backgrounds to ensure readability over images. |
26-Aug-24 |
|
Enterprise Dashboard: Guided Selling Template Styling Enhancements
|
We've refined the Guided Selling template styling for a better user experience. The start screen text is now centered within a container with adjustable opacity to reveal the background image. Error messages when no products are found are also centered in a container, with an added button to restart the quiz using the same start screen styling. For mobile views, we've widened the screen container, added padding to question text, and expanded the progress bar and answer buttons. Recommendation styling has been tightened with reduced margins around images and smaller fonts for product names and prices. A new button in the top right of the recommendations allows users to restart the quiz, hiding results and starting over. |
26-Aug-24 |
|
Enterprise Dashboard: Guided Selling 2.0 - UX Enhancements Implemented
|
Several UX enhancements have been made to the Guided Selling 2.0 experience, specifically addressing the Quiz step. These updates improve usability and design consistency, ensuring a smoother user experience. Key changes include hiding the Back button on the first quiz question, adjusting the main window height to keep the Save & Proceed button visible, and improving the color picker functionality. Additionally, clickable "Step" buttons have been made more intuitive, and the issue with the Start Screen reappearing after being disabled has been resolved. |
26-Aug-24 |
|
Enterprise Dashboard, Find: UI Upgrade for Publish Button on Search and Browse Merchandising Page
|
The UI for the publish button on the Search and Browse merchandising pages has been upgraded. Previously, if all items were deleted from the boosts or links page, there was no option to remove them from Solr due to the absence of a publish button. With this enhancement, the publish button is now available even when no items are present on the page, similar to the functionality on the Visual Merchandising page. This ensures that changes can be effectively published regardless of the number of items on the page. |
26-Aug-24 |
|
Discover, Find: Enhancement to Product Snapshot Property Definitions |
We have enhanced the product snapshot property definitions by adding find attribute settings to the "recommendable" field. This update ensures that the "recommendable" field is now properly configured with find attribute settings, making it facetable and filterable, among other properties. Additionally, the "querytag_name" field has been updated to be sortable, allowing for more flexibility and precision in handling product data.
These changes enhance the functionality and manageability of product snapshots, ensuring that key fields are configured to meet the evolving needs of product recommendation and search operations. |
27-Aug-24 |
|
Discover, Find: Enhancement to RuleIndexBuilder Job for Pinned Product Rules
|
We have updated the RuleIndexBuilder job to save pinned product rules directly to Find Solr via the streaming catalog. This adjustment streamlines the process of managing pinned product rules. |
27-Aug-24 |
|
Find: Implementation of Case Insensitive Search for Vector Data
|
We have enhanced the search functionality by implementing case-insensitive search when fetching vector data from the query tag collection. This improvement ensures more consistent and accurate search results, regardless of the case used in the search queries. |
27-Aug-24 |
|
|
Discover, Find: Adjustments to Play Service for Pinned Products Merchandising Rules
|
The play service has been updated to support the saving of Pinned Products merchandising rules for both Find and Discover. |
27-Aug-24 |
Find: Hybrid Search - Query Tag Name Retrieval Fix
|
We have resolved an issue where the query tag name was incorrectly fetched from the query tag schema instead of the query vector schema in the search service during vector search queries. This issue has now been fixed. |
27-Aug-24 |
|
Find, Science: Creation of Vectors for Search Queries |
We have enhanced the system by adding a job that converts search queries into vectors for use with Find. This job reads input queries and generates corresponding vectors, which are then used to search Solr for product vectors closely matching the search query vector. This enhancement ensures that search vectors are created similarly to product catalog vectors, utilizing OpenAI embeddings. Vectors for previously converted queries are now efficiently sourced from the cache. |
27-Aug-24 |
|
Find, Science: Schema Per Site for Storing Catalog Embeddings |
We have implemented a schema-per-site approach for storing catalog embeddings, ensuring better organization and management of data. A dedicated schema, such as catalog_0xxxx.find_catalog_embeddings, is now created for each site. This enhancement involved creating the necessary change scripts, updating table names in the data science catalog embeddings job, and adjusting the Find catalog vector job accordingly. Data is now stored in the appropriate site-specific schemas. |
27-Aug-24 |
|
Find: LLM Access Service Integration for Enrichment Service |
We have implemented a job that uses the LLM access service to generate related search suggestions from historical queries. These suggestions are then ingested into the Enrichment Service and stored in Solr, enhancing search capabilities with more relevant and accurate results. |
27-Aug-24 |
Bug and Support Fixes
The following issues have been fixed in the release version 24.16 during Aug 09, 2024 - Aug 26, 2024.
Jira # |
Module/Title |
Summary |
General Availability |
---|---|---|---|
|
Enterprise Dashboard: Prevent Single Line Comments in Layouts
|
We have addressed an issue where single line comments in layouts were causing unintended display of products from different customers in email recommendations. Since our layouts compile into a single line, all content following a single line comment was being commented out, leading to these errors. To prevent this, the layout editor will now check for single line JavaScript comments (//) before allowing a layout to be saved. If detected, an alert will prompt the user to modify comments to the multi-line format. |
26-Aug-24 |
Message Type Page Not Loading When Creating Variation
|
We have resolved an issue where the "Message Type" page was not loading correctly during the creation of a social proof campaign. Users encountered a persistent loading message after selecting the context and clicking "Save and Proceed." The issue has been fixed, and the Message Type page now loads as expected. |
26-Aug-24 |
|
Enterprise Dashboard: Social Proof Messages - Reordering Issue Resolved
|
We have fixed an issue on the Social Proof message page where reordering message types was not being saved correctly. After changing the order and saving, the message order would revert to its original state upon returning to the page. The issue has been addressed, and the message order now persists as expected after saving. |
26-Aug-24 |
|
|
Enterprise Dashboard: Disabled "Use Region As A Seed" Option for Unsupported Models
|
A bug has been fixed where the "Use Region As A Seed" option was incorrectly available for certain configurable strategy models that do not support this feature. The affected models include Top Sellers Offline, Top Sellers Omni Channel, DSW Strategy, NLP Similarity, NLP Cross Sell, and User Browse History. The "Use Region As A Seed" option is now properly disabled and unchecked for these models, ensuring the correct configuration settings are applied. |
26-Aug-24 |
|
Fixed Issue with TopProducts Azkaban Job Pipeline Execution |
A bug has been addressed where the TopProducts Azkaban job failed to generate files when the aggregation part was skipped. This issue caused the MR pipeline in the TopProducts job not to trigger, leading to incomplete processing. The problem has now been resolved, ensuring that the pipeline runs correctly even when the aggregation step is skipped. |
26-Aug-24 |
Find, Recommend: Fix for Restriction Rule Not Working in Find
|
A bug affecting the Restriction Rules in Find has been resolved. The issue prevented rules from working as expected when trying to exclude certain products, such as "colete" (vest), from search results for "casaco feminino" (women's coat). The problem occurred both when using ProductIDs and Product Attributes in the Restriction Rules. The configuration for "enable recommendations restrictions for find" is now functioning correctly, ensuring that exclusion rules are applied as intended. |
26-Aug-24 |
|
Discover, Recommend, UPS Fix for Restriction Rules Not Filtering Recommendations
|
A bug affecting Restriction Rules in the recommendation engine has been resolved. The issue prevented the rules from properly filtering out products that did not match the specified user attributes, such as EligibleSalesOrgs and PromoEligibleSalesOrgs, during recommendation requests. As a result, products that should have been excluded based on these attributes were incorrectly included in the recommendations. The issue has been addressed, and the Restriction Rules are now functioning correctly, ensuring that only matching products are recommended. |
26-Aug-24 |
|
Enterprise Dashboard: Fix for Browse Configuration Page Not Loading
|
An issue that prevented the Browse Configuration page from loading has been resolved. The problem was caused by the removal of "charser=utf-8" from the discoverConfig.js file. The necessary corrections have been made, and the page now loads properly. |
26-Aug-24 |
|
Enterprise Dashboard: Fix for Missing Information in Restriction Rules
|
An issue where critical information, such as the name, date, and other details, was missing when opening certain Restriction Rules has been resolved. The problem was identified when clients attempted to view specific rules, and errors were returned in the console. The necessary corrections have been made, and the information is now displayed correctly when accessing these rules. |
26-Aug-24 |
|
Enterprise Dashboard: Fix for Sorting by User Affinity Option in Engage Campaigns |
An issue preventing the deselection of the "Sort selected content by User Affinity" option in Engage campaigns has been resolved. Previously, when users attempted to uncheck this option and save the rule, the option remained selected. This issue has now been fixed now. |
26-Aug-24 |
|
Find: Correction in Autosuggestion Job for Scoped Action Handling |
A bug in the Autosuggestion job caused errors due to incorrect handling of scoped actions during business rules updates. The issue stemmed from improper handling of the Snapshot/ScopedAction object. This issue has now been fixed. |
27-Aug-24 |
|
|
Find: Sitewide Product Boost Rule Fix
|
We have resolved an issue where the sitewide product boost rule caused certain products to always appear at the top of search results, regardless of the search query. For instance, a product like "dell lattitude abc" would consistently appear first, even when searching for unrelated items like "mango" or "fork." This issue has now been fixed. |
27-Aug-24 |
Find: Hybrid Search - Query Tag Name Handling Issue Resolved |
We have resolved an issue where the query tag name was incorrectly fetched from the query tag schema instead of the query vector schema in the search service during vector search creation. This fix ensures that vector searches now retrieve data from the correct schema. |
27-Aug-24 |