Release Summary 25.15 | Jul 24, 2025
The following key features and improvements, along with bug fixes, have been released in Algonomy CXP products in the release version 25.15.
Ensemble AI
Enhanced reporting for Ensemble AI with new metrics, visualizations, and reports
Ensemble AI reporting has been expanded with new metrics, visualizations, and dedicated reports to provide deeper insights into shopper engagement and system activity.
View and click-through rate (CTR) metrics have been added to the Style Performance and Merchandising reports (Product, Category, and Brand), offering a more complete view of how ensembles are seen and interacted with. These metrics are also now available as time series visualizations, helping teams monitor trends and performance over time.
Additionally, two new reports have been added to the navigation: the Ensemble AI Merchandising Report, which highlights engagement across key retail dimensions, and the Ensemble AI API Call Count Report, which tracks API usage. These updates make it easier to analyze effectiveness and optimize Ensemble AI performance.
Jira: ENG-30349, ENG-30456, ENG-30615
Conditional display of backgroundless image URLs in Ensemble AI
Ensemble AI now includes backgroundless image URLs in API responses only when transparent images are available for a product. This ensures that responses remain accurate and do not reference unavailable or missing assets. The images may originate from site configuration settings or be system-generated, and the URLs retain their original format such as .jpg or .png.
Additionally, the hosting location for these images has been moved from Cachefly to Cloudflare to support more efficient content delivery. Cloudflare access and directory permissions have been set up to manage this change securely.
Jira: ENG-30275
Social Proof
ATC Rate metric added to Social Proof MVT reports and API
A new Add to Cart (ATC) Rate metric has been introduced in the MVT reports and API for Social Proof tests. This metric helps merchandisers measure how often users add products to their cart after viewing a social proof message, offering a clearer view of its impact.
The ATC Rate is calculated using different approaches based on whether the “Filter by products shown social proof message” option is applied. It supports standard dimension filters and is available via the MVT API for use in dashboards and analysis. The metric also includes lift and confidence level calculations to support deeper evaluation of Social Proof performance.
Optimization training data is now populated when Social Proof optimization is enabled at site level
The Social Proof API now populates the optimization training data required for model training whenever optimization is enabled at the site level. Previously, this data was only generated at the variation level.
With this update, the training data is populated when optimization is enabled for a site, ensuring better training data and improved prediction accuracy during the initial days for sites using Social Proof optimization.
Jira: ENG-30362
Recommend
New Precision Mode introduced in User Affinity Configuration
A new scoring method has been added to improve recommendation accuracy by reducing the influence of common or generic attribute values in user affinity scoring. Using a TF-IDF approach, the system now down weights high-frequency values like "N/A" or "white" that could otherwise distort personalization.
To enable this, a Precision Mode option is now available in the User Affinity Configuration UI. When selected, it applies the adjusted scoring (inverse_document_frequency). The Blending Weight setting (formerly Weighted Mean Reciprocal Weight) and related affinity controls have also been reorganized under a new Default User Affinity Configuration section for easier access.
Category diversity filter added for Top Products with Category Context
A new filter option, “Diversify results by category,” is now available when using the Top Products model with Category Context as the personalization seed. This allows digital optimization managers to surface a broader mix of categories in recommendations, helping improve variety and relevance in product discovery.
Jira: ENG-30563
Improved clarity for AND/OR logic in Merchandising Rules
The UI for Rec Restriction and Boosting rules has been updated to make the use of AND/OR logic clearer when multiple conditions are applied. A summary phrase now appears above the rule conditions:
-
For OR logic: “Products matching any of the following (OR logic):”
-
For AND logic: “Products matching all of the following (AND logic):”
Jira: ENG-30537
User attribute targeting added to Strategy Rules
Strategy Rules now support targeting based on user attributes, enabling digital optimization managers to define rules that activate specific strategies for certain shopper segments. This provides more control over personalization based on known user characteristics.
A new ‘User Attribute’ tab has been added to the context definition section, allowing entry of attribute name-value pairs with autocomplete suggestions and optional exclusions.
Jira: ENG-30459
Find
Channel-based TS visualizations added to Zero Search Results report
Time series visualizations by channel have been added to the Find Zero Search Results report. Merchandisers can now view and analyze zero-result search trends by channel, helping identify areas that need attention.
Users can view aggregated data across all channels by default or filter by specific channels for a more focused view. These visualizations are available across regions and cover all search terms, including those with and without facets.
Jira: ENG-30470
Facet-only queries excluded from Find reports
Find reports now exclude search queries where the response style is set to "FacetOnly." This ensures that only meaningful result-driven queries are reflected in reporting metrics.
The update applies to Find Search Terms Reporting, Find Reporting, and Find Response Reporting, helping maintain cleaner data and more accurate insights.
Jira: ENG-30437
Enterprise Dashboard
Margin Model added to Model Options page and API
The Margin Model is now fully integrated into both the Model Options page and its supporting APIs, allowing digital optimization managers to enable and configure it without backend assistance. This self-serve enhancement improves visibility and accessibility of model-based strategies.
Users can view and manage the Margin and Profit model settings, choose the appropriate attribute for each, and access supported strategies such as MarginClickCP, MarginViewedPurchased, and MarginPurchaseCP. These strategies are now also listed in the relevant APIs, with configuration support included for seamless integration and customization.
MVT API updated to use default currency from portal
The MVT API now uses the default currency configured in the portal when no currency parameter is provided in the request. This aligns MVT reporting behavior with the rest of the reporting system, ensuring consistency in how currency values are handled and displayed.
Jira: ENG-30162
Ecommerce Chatbot
Chatbot Update: Regional Currency and Language
The Ecommerce chatbot now uses the shopper’s region to display product prices in the correct currency and initiates chats in the appropriate language. Language settings are also passed through the API to ensure consistent responses.
Jira: ENG-30463
API Support for Affinity-Based Product Ranking
The /affinityScoresByConfig API now supports sorting one or more product lists by user affinity scores. This enables future chatbot experiences to deliver personalized product recommendations based on user preferences.
Jira: ENG-30511
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 25.15.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Ensemble AI: New Ensemble AI API to Identify Products with Outfits |
A backend API is now available in Ensemble AI to help merchandisers identify which products have generated outfits for a given style, streamlining the review process, especially when many seed products are involved. The API provides the number of ensembles created, the seed products contributing to them, and supports filtering by category, region, and style status. It also includes the last model run time in the client’s time zone. Note: Front-end integration is not yet complete and will be added in an upcoming release. |
24-Jul-25 |
|
|
Find: Findability and CTR Metrics Introduced for Find Configuration MVTs |
New metrics, Findability and Find CTR, are now available for Multivariate Tests (MVTs) on Find configurations. These help optimization managers compare how different setups impact product discovery.
Note: The rollup is complete, and UI visibility will be enabled in a future update. |
24-Jul-25 |
|
|
Find: KNN set as default algorithm for hybrid vector search |
The default algorithm for hybrid vector search has been updated from vector similarity to KNN. This change improves retrieval behavior by using KNN as the standard approach for similarity scoring. |
24-Jul-25 |
|
|
Find: Personalization disabled when default sort is overridden |
To improve performance, personalization is now automatically disabled in Find when the default sort is overridden with a custom sort field. Since personalization boosting does not apply in such cases, this update prevents unnecessary processing when the primary sort field is not based on score. |
24-Jul-25 |
|
|
Chatbot icon moved to top right corner of the page |
The chatbot icon has been repositioned to the top right corner of the dashboard to avoid overlapping with key actions like saving rules. It now appears next to the Help links and remains visible while scrolling, improving overall usability. |
24-Jul-25 |
|
|
Historical Offline Feed migrated to Airflow |
The Historical Offline Feed process has been migrated to Airflow, streamlining job orchestration and improving reliability and scalability of data processing workflows. |
24-Jul-25 |
Bug and Support Fixes
The following issues have been fixed in the release version 25.15.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Find: Search Service: Latency Metric Tagging Issue Fixed |
Search request latency metrics were being incorrectly tagged with the cluster value "local" for all requests. This has now been corrected, and metrics will reflect the appropriate cluster based on the actual request origin. |
24-Jul-25 |
|
|
Catalog Embedding Job Failure Resolved |
The catalog embedding job was failing due to a data type mismatch in the description_meta column and missing input files. These issues have now been resolved, ensuring the job runs successfully across all sites. |
24-Jul-25 |