Release Summary 25.12 | Jun 12, 2025
The following key features and improvements, along with bug fixes, have been released in Algonomy CXP products in the release version 25.12.
Ensemble AI
Use shopper history to identify relevant outfit styles
Ensemble AI can now use a shopper’s purchase or view history to identify matching products from selected styles. This helps marketers create more relevant outfit recommendations for specific themes, such as summer entertainment or back to school, by focusing only on styles related to those campaigns.
Based on the number of seeds specified, the system identifies viewed or purchased products that have ensembles and ranks them by score. Users can select specific styles to receive tailored ensemble recommendations. If no style is selected, the system automatically finds matching products across available styles, ensuring that the displayed outfits align with user preferences and campaign objectives.
Jira: ENG-30173
Style-Specific Outfit Selection in Ensemble AI
We have enhanced the Ensemble AI experience by allowing shoppers to focus on outfits from a specific style when multiple styles are available for a product. A new “Show only ensembles of this style” button and a style selector appear when applicable, helping shoppers explore more relevant combinations. This makes it easier for them to stay within a preferred look or theme, improving engagement and shopping efficiency.
Additionally, merchandisers can choose to hide style names from the shopper experience. When this is enabled, the style-specific options are removed to deliver a cleaner interface. These updates give businesses more control over how style-driven ensembles are displayed, improving both personalization and user experience.
Jira: ENG-30216
Improved Click Behavior and Layout Enhancements in Ensemble AI
Shoppers can now seamlessly navigate to the product detail page in a new tab when quick view is not available, ensuring a smoother browsing experience. This change applies to all ensemble templates where quick view support is absent.
We have also made layout refinements for better balance in four-product displays and ensured similar products become clickable after being swapped. Carousel views have been optimized to fully display all ensembles.
Jira: ENG-30257
Recommend
Cart-Based Recommendation Model in Configurable Strategies
A new recommendation model, "User’s Cart", has been added to configurable strategies to help optimization managers re-engage shoppers with products currently in their cart. The model uses up-to-date cart state information, considering items added, removed, or purchased since the last cart view, ensuring highly relevant recommendations.
It supports multiple seed options including brand and category contexts, and allows personalized sorting by user affinity and Smart Shuffle. Filter and diversification options by brand and category are also available.
Find
Automated Query Enrichment with LLM in Find
We have introduced an automated process to enhance shopper search queries in Find using AI. This process helps improve search relevance by expanding poorly performing or zero-result queries with more meaningful variations.
The system runs daily, applies predefined rules, and stores the enriched results for improved future searches.
Excluding Facet-Only Queries from Zero Search Results Report
The Zero Search Results report now excludes queries where only facets are requested (ResponseStyle = FacetOnly). This ensures that the report reflects only genuine zero-result searches, providing more accurate insights for merchandisers and search analysts.
Jira: ENG-30436
Social Proof
Enabling Social Proof Optimization
With this capability, social proof optimization is enabled for client-side Integrations.
Social Proof Optimization leverages AI to select the highest-performing messages from a set, tailored to specific message types, intervals, and thresholds. It empowers users to either display a single top-performing message or present multiple messages in order of their performance for a chosen Key Performance Indicator (KPI).
Core Mechanism
At its heart, Social Proof Optimization employs a sophisticated contextual bandit algorithm. This algorithm dynamically balances exploration (testing new messages) and exploitation (using proven messages) to optimize for metrics such as Revenue Per Visit, Conversion Rate, or Add to Cart Rate.
User Experience and Automation
Users provide a list of messages through social proof campaigns and enable the optimization feature. The AI autonomously identifies the best-performing message while continuously exploring to adapt to changes in user behavior, ensuring sustained performance for the selected KPI.
Jira: ENG-29493
Social Proof Reporting: View Visits with or without Messaging
A new checkbox has been added to the Social Proof reporting interface, allowing users to filter visits based on whether the social proof message was shown or not. When selected, the report displays a clear breakdown of visits per variation, distinguishing between those where the message was shown versus not shown. This enhancement helps merchandisers better evaluate the impact of social proof on user behavior.
Jira: ENG-30269
Enterprise Dashboard
New Configuration Options for Shopper Feedback in Ensemble AI
A new "Shopper Feedback" configuration page has been added under the Optimization section in the portal. This allows merchandisers to customize how shopper feedback scores are calculated to fine-tune which ensembles appear.
Users can now select product attributes like brand and total price for scoring, define price ranges per currency, and set scoring rules for shopper actions such as upclicks and downclicks. Additionally, the decay factor that influences how feedback impact fades over time can be adjusted, helping tailor the experience based on shopper interaction patterns.
Jira: ENG-29478
Support for showOnlyQualified Parameter in Experience Calls
An optional setting has been added in client.js to support the new showOnlyQualified parameter for experience targeting. When enabled, this parameter ensures that only contextually relevant experiences are returned, helping reduce the size of the response and improve efficiency.
This feature can be activated via a URL parameter and, once enabled, the system skips the separate targeting call and directly fetches qualified experiences.
Jira: ENG-30263
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 25.12.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Recommend: Airflow Enabled for All DSW Clients |
Airflow has been enabled for all existing clients using Data Science Workbench (DSW) following successful stability tests. Moving forward, Airflow will also be configured by default for any new DSW clients. Monitoring has been set up to ensure consistent performance across all sites. |
12-Jun-25 |
|
|
Enterprise Dashboard: Secure Links for Email Recommendations in Recs Test Drive |
Email recommendations in Recs Test Drive now default to using https for all image and anchor links. This update ensures secure rendering of content and aligns with customer expectations for secure communications. The changes apply to both the rendered view and the HTML Code tab. |
12-Jun-25 |
|
|
Enterprise Dashboard: Hide Discovered Segments Tab for Non-Engage Sites |
To avoid confusion for clients without Engage, the “Discovered Segments” tab is now hidden on the Segment Library page if Engage is not enabled for the site. The tab remains visible for sites with Engage access. |
12-Jun-25 |
|
|
Enterprise Dashboard: Alert for Lookback Period in Top Sellers Model |
An alert now appears on the Model Options page if the Top Sellers lookback period is set to less than 7 days. This warns users that it may impact the Movers and Shakers model, which compares trends over the past week. Users can choose to cancel or proceed, helping avoid unexpected reporting gaps. |
12-Jun-25 |
|
|
Enterprise Dashboard: UI Improvements to User Affinity Configuration Page |
Several UI updates have been made to improve usability on the user affinity configuration page. These include updating the label for “Weighted Mean Reciprocal Weight,” fixing display issues with long attribute names in hover tooltips, enhancing border visibility for input fields, and adding a '+' icon to the “Add attributes” placeholder. |
12-Jun-25 |
|
|
Enterprise Dashboard: Social Proof Reporting: Breakdown by Message Shown vs. Not Shown |
Social Proof reporting now includes a clearer breakdown of how many visits actually displayed the social proof message within each variation. This helps distinguish overall variation performance from the subset where the message was shown. Metrics are now available for both message-shown and not-shown visits at each variation level. |
12-Jun-25 |
|
|
Platform: Timestamp Added for Cart and Wishlist Items |
Each item in the cart and wishlist now includes a timestamp to reflect when the action occurred. For cart items with multiple add events, the total quantity is consolidated and the timestamp reflects the most recent event. |
12-Jun-25 |
|
|
Exclude Purchased Products from Cart State |
The cart’s current state has been enhanced to exclude products that have already been purchased. Once a confirmed order is received, the corresponding items are removed from the cart state calculation to ensure accuracy in tracking shopper intent. |
12-Jun-25 |
|
|
Recommend: Product Details Now Included for Wishlist Items in Dashboard UPS Response |
The dashboard API has been enhanced to include product details for items associated with wishlist activity. This ensures product names and images display correctly for events like add to wishlist, remove from wishlist, and wishlist views. Details are also now returned for items removed from the cart, improving the completeness and clarity of the user profile display. |
12-Jun-25 |
|
|
Recommend: Disable In-Session Personalization in Privacy Mode |
To support privacy mode, in-session personalization has been disabled by ignoring the provisional user ID. When privacy mode is enabled, the system now ensures that no user-specific history is used for personalization. This helps maintain compliance with shopper privacy expectations, especially for clients not using DTA. |
12-Jun-25 |
|
|
Science: Switched to Azure API for Generating Vector Embeddings |
The system has been updated to use the Azure OpenAI endpoint for generating vector embeddings from the text-3-embedding model. This change helps avoid quota limitations from the standard OpenAI endpoint and ensures smoother execution of production jobs. |
12-Jun-25 |
|
|
Science: Shopper Feedback: Attribute and Scoring Configuration |
A configuration API is now available to set how shopper feedback scores are calculated. Business users can choose product attributes (like brand or price buckets), define price ranges per currency, and adjust scoring weights (up/down click and decay). Validation ensures ranges don’t overlap or leave gaps. These settings help fine-tune how feedback influences re-ranking. |
12-Jun-25 |
Bug and Support Fixes
The following issues have been fixed in the release version 25.12.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Social Proof: Long Load Time on Social Proof Message Types Screen |
The issue causing delays in loading the Social Proof message types screen has been resolved. The screen now loads consistently without needing manual refresh or navigation, improving usability during experience setup and customer demos. |
|
|
|
Social Proof: SP Messages API Not Updating RCS with Treatment Assignments |
The SP Messages API has been updated to correctly assign treatments and update RCS cookies. This ensures consistent treatment tracking and behavior across shopper |
|
|
|
Ensemble AI: Unable to Enable Ensemble AI Model |
Fixed an issue where the "Enable Ensemble AI" option was not working. The problem occurred due to missing configuration mappings, which caused errors during activation. A validation check has been added to handle missing configs and ensure smooth model enablement. |
12-Jun-25 |
|
|
Query Vector Job Failure |
Resolved an issue where the query vector job failed for a particular site due to data-related errors in the output file. The job now correctly processes the output CSV with the SOH delimiter, and the issue has been fixed |
12-Jun-25 |
|
|
rrserver: API recsUsingStrategy Uses Site Configuration as Intended |
The recsUsingStrategy API has been updated to correctly apply the enable_request_based_response_domain site configuration. The response now follows the expected behavior based on this setting. |
12-Jun-25 |
|
|
Recommend: Wildcard Refinements with Multiple Conditions |
Resolved an issue where wildcard refinements using the same attribute name with OR logic did not return any results. This fix ensures multiple conditions with wildcards now work as expected, allowing accurate product filtering. |
12-Jun-25 |