Release Summary 25.06 | Mar 20, 2025
The following key features and improvements, along with bug fixes, have been released in Algonomy CXP products in the release version 25.06.
Ensemble AI, Enterprise Dashboard
Ensemble AI: View Style-Specific Outfits for Shoppers
We have introduced a new option that allows shoppers to view outfits exclusively from a specific style. Previously, shoppers would see outfits from multiple styles ranked based on scoring. Now, they can select a preferred style and see all outfits curated under it, enhancing engagement and personalization.
Shoppers can switch between style-specific outfits and the default view. This option is only available when outfits come from multiple styles.
Jira: ENG-27902
Ensemble AI: Apply Layout Design to Dynamic Experience
We have introduced the ability to apply custom layout designs at the style level for Ensemble AI Dynamic Experience, ensuring that shoppers see ensembles arranged according to the Merchandiser’s intended aesthetic. The layout design defines product placement coordinates, ensuring a structured and visually appealing display.
The EnsembleAIRecs API now supports layout-specific rendering. If a custom layout design is available for a style, it will be applied automatically; otherwise, the system will default to the six-block layout. Additionally, transparent product images (if configured) will be used to maintain a seamless shopping experience, with a fallback to standard product images when necessary.
Jira: ENG-29432
Ensemble AI: rcs Parameter Support in API Calls
Enhanced the Ensemble AI API to support the rcs parameter, ensuring accurate tracking, reporting, and attribution. Similar to Social Proof and recsForPlacement API calls, rcs is now included in the response and must be passed back in subsequent API requests.
This improvement ensures consistent session tracking, enabling better engagement attribution. The rcs parameter is now available in Headers, Payload, and Preview, maintaining uniformity across the system.
Jira: ENG-29778
Ensemble AI - API Call Count Reporting
Added a new report to track Ensemble AI API usage and trends, providing visibility into API call volumes across different sites and regions. The Ensemble AI API Report is available only when Ensemble AI is enabled, ensuring relevant data is captured for analysis.
Users can view API call counts over a selected date range, analyze trends through graph visualizations, and download detailed tables containing site ID, site name, method, date, and total call count. This report helps merchandisers monitor API usage, optimize system performance, and make data-driven decisions.
Jira: ENG-29396
Enterprise Dashboard
Add Wishlist Event to User Affinity Configuration
User affinity scoring now considers wishlist events, allowing a stronger indication of shopper interest in ranking products and content. In the Advanced Tab of User Affinity Configurations, new scoring settings have been added for "saved to wishlist" events, with default values for new and old wishlist saves, along with a lookback period.
The wishlist events are visually differentiated in charts, using distinct colors, and API updates ensure wishlist-based affinity scoring values are included when saving configurations.
Recs Test Drive: Select Multiple Placements for Preview
Users can now preview multiple recommendation placements simultaneously in Recs Test Drive, allowing better simulation of pages with multiple placements. The placement selector has been expanded, enabling selection and reordering of multiple placements from the same page type, with a clear visual representation of selected placements.
The results section now separates recommendations by placement, with distinct headers and horizontal separators for better clarity. Additionally, rules are displayed per placement in separate tables. Email placements are limited to one selection per test when using the Email channel, with warnings and restrictions in place to guide users accordingly.
Jira: ENG-28482
Find
Find Search Terms Report - Zero Results Analysis
The Find Search Terms Report now includes two new tabs to analyze search terms that yield zero results, enabling search optimization by identifying and improving ineffective queries.
Zero Results (without facets): Displays search terms that return no results without any applied filters. Includes a bar chart visualization for the top 30 terms and a detailed table, allowing users to take corrective actions such as boosting, linking, or adding synonyms.
Zero Results (with facets): Identifies search terms that fail due to applied filters, helping refine facet visibility when no relevant products exist. This report provides search term breakdowns, filter counts, and actionable insights to improve search performance.
Jira: ENG-28527
Social Proof
Support for Multiple Location Selectors in Social Proof
Social Proof now supports multiple location selectors within a single experience, allowing merchants to display messages in multiple locations on the same page without creating separate experiences. This enhancement reduces the need for additional configurations and provides flexibility in managing messages across both web and mobile versions of a site.
By default, a single location selector is provided, but users can add multiple selectors as needed. When multiple locations are specified, the message will appear in all designated spots. If multiple experiences target the same location, the system will prioritize the experience with the highest importance.
Jira: ENG-29856
Social Proof Badging: Limit the Number of Displayed Badges
Users can now define the maximum number of badges displayed on product pages when multiple badges meet the criteria. Badges are selected based on priority, ensuring only the most relevant ones appear. If no limit is set, all qualifying badges will continue to be shown for backward compatibility.
The client-side will respect this configuration, displaying only the specified number of badges. By default, up to three (3) badges will be shown unless configured otherwise.
Jira: ENG-29759
Social Proof Optimization: Variation-Level Control and Multiple Messages
Social proof optimization can now be enabled at the variation level, allowing AI to prioritize and display messages based on key metrics such as Revenue Per Visit (RPV), Conversion Rate (CVR), or Add to Cart Rate (ATC Rate). Users can toggle optimization for each variation, provided the feature is enabled at the site level.
When multiple messages are enabled, social proof will display messages in priority order. If optimization is off, messages meeting the threshold will be shown based on predefined ranking. If optimization is enabled, messages are selected based on probability scores from the prediction API, ensuring the most relevant messages appear first. If only one message is returned by the prediction API, additional messages meeting the threshold will be displayed in order of priority. When only one message qualifies, the prediction API is not called. Logging enhancements now capture probability scores, model IDs, and other key details for improved tracking.
Social Proof - Activity Log
The Social Proof feature now includes an Activity Log, allowing merchandisers to track changes made to social proof campaigns. This helps in understanding modifications over time and comparing performance before and after changes.
The Activity Log records key details such as date of change, change type, user, and specific modifications. Logged events include adding, modifying, deleting, or cloning experiences, adjusting traffic allocation, updating message types, and making display design changes.
Jira: ENG-28088
Recommend
Improved Advanced Merchandising Rules in RecsForPlacementsContext API
The RecsForPlacementsContext API now fully supports Advanced Merchandising Rules, ensuring that merchandising logic is correctly applied when generating recommendations. Previously, these rules worked in RecsForPlacements API but were not applied in RecsForPlacementsContext API.
With this enhancement, merchandisers can now use Advanced Merchandising Rules seamlessly within RecsForPlacementsContext API.
Jira: ENG-26257
Engage
Engage Debugging: Campaign and Affinity Score Visibility
Engage debugging now provides enhanced visibility into content selection by including campaign details and affinity scores in the response. This update helps optimization managers better understand which campaigns influenced content selection and how affinity scoring impacted ranking.
If affinity scoring is used, the response now includes the calculated affinity score for each content item.
Jira: ENG-29842
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 25.06.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Find: Hybrid Search: Enforcing Minimum Query Length |
Hybrid Search now includes a minQueryLength configuration, ensuring that searches execute only when the query meets or exceeds the defined character limit. By default, queries with fewer than three characters will not trigger Hybrid Search. If a query does not meet the minimum length, the system logs a debug message indicating that Hybrid Search conditions were not met, improving efficiency and preventing unnecessary processing. |
20-Mar-25 |
|
|
Find: Support for Dutch (NL) in Streaming Context |
Streaming Find Indexer now supports the Dutch (NL) language container for B2B use cases in Streaming Context. This enhancement ensures compatibility with Dutch-language data processing, improving multilingual capabilities for businesses operating in the region. |
20-Mar-25 |
|
|
Find: Boosting for Spellchecked Query Terms |
Find API now ensures consistent search results when a query undergoes spell correction. When a misspelled query is corrected, all existing boosting rules—QU boost, Boost & Bury, and Personalization Boost—are applied to the spellchecked term, aligning results with those of an initially correct query. |
20-Mar-25 |
|
|
Find: Hybrid Search Query Result Cache Optimization |
A fix has been implemented for Solr hybrid search to ensure that repeated queries with preFilters are now served from the cache instead of being reprocessed from scratch. Previously, a bug in Solr’s source code prevented result caching, impacting performance for vector search. With this enhancement, repeated queries now leverage caching, significantly improving query performance. |
20-Mar-25 |
|
|
Find: Arabic Language Support |
This enhancement ensures support for Arabic language attributes within the client’s retail catalog, addressing previously reported issues. The implementation enables proper handling of Arabic content in product data and catalog processing. |
20-Mar-25 |
|
|
Find: Language-Specific Vectors for Product Catalog and Data Science Embeddings |
The Data Science Catalog Embedding Vector Job now supports language-specific vector generation for improved multilingual search and recommendations. Product data, including product names and category names, is retrieved language-wise from the database, and vectors are generated accordingly. A new column, language_tag, has been added to the find_catalog_embeddings table, enabling the data publisher job to pick vectors based on language and push them to the enrichment service via language overrides. This enhancement ensures that vector models differentiate product embeddings based on language, improving search accuracy and recommendation relevance. |
20-Mar-25 |
|
|
Find: Exponential Backoff for Enrichment Service Requests |
To prevent failures from intermittent 503 errors, an exponential backoff retry mechanism has been implemented for catalog-vector, globalrank, and woc jobs. Previously, retries occurred without delays, causing repeated failures. The retry handler used for vector jobs in streaming ingest has now been applied to EnrichmentRestClient, improving stability and recovery across relevant jobs. |
20-Mar-25 |
|
|
Find: Handling Max Boolean Clause Error in Solr 9 |
To prevent Solr timeouts caused by excessive synonym expansion, a new configuration has been introduced: If the max boolean clause limit is reached, synonym expansion will be disabled, ensuring Solr performance stability across multiple clients. |
20-Mar-25 |
|
|
Updated Support Customer Portal Link in RR Portal |
Following the migration of JIRA to the cloud, the Support Customer Portal link in the RR Portal has been updated. The previous link (https://jira.richrelevance.com/jira/servicedesk/customer/portal/1) has been replaced with the new cloud-based link: https://algonomy.atlassian.net/servicedesk/customer/portal/4. |
20-Mar-25 |
Bug and Support Fixes
The following issues have been fixed in the release version 25.06.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Enterprise Dashboard: Fix for Category Selection in Advanced Merchandising (AM) Rules |
We have resolved an issue where selecting a category in Advanced Merchandising (AM) Rules would incorrectly switch it to its super category upon selection. This was due to a recursive loop modifying the ancestor categories array, preventing proper loading of child categories. With the fix, selected categories now display correctly, and all child categories appear under their respective parent categories without unexpected modifications. |
20-Mar-25 |
|
|
Enterprise Dashboard: Fix for Recs Test Drive - Forced Strategy Behavior |
We have fixed an issue in Recs Test Drive where recommendations were displayed even when the returned strategy did not match the Forced Strategy selected. Now, if the returned strategy differs from the Forced Strategy, recommendations will not be shown, ensuring accurate strategy validation. |
20-Mar-25 |
|
|
Ensemble AI : Fix for Ensemble AI Report - Detail Report Loading Issue |
We have fixed an issue where the detail report was not loading when selecting a specific style in the Line Chart of the Ensemble AI Report. Now, the report loads correctly when a particular style is selected, ensuring seamless data visualization. |
20-Mar-25 |
|
|
Find: Fix for CatalogEmbedding Job Not Creating Path in Hadoop |
We have resolved an issue where the CatalogEmbedding job for query vectors was failing due to missing paths in Hadoop, causing the QueryVectorJob to fail. The root cause was improper path creation during execution, leading to missing directories and failed data retrieval. With this fix, the CatalogEmbedding job now correctly generates the required paths in Hadoop. |
20-Mar-25 |
|
|
Data Reporting: Fix for Email Recommendations Report Loading Issue |
The Email Recommendations Report was not loading due to a MIME type error, preventing the proper application of styles and causing a pop-up displaying '1'. This issue has now been resolved, ensuring the report loads correctly without errors. |
20-Mar-25 |
|
|
Enterprise Dashboard: Fix for Home Page Carousel Issue |
The Home Page Carousel issue, where the compound strategy was not appearing in the Strategy Rules page despite being linked, has been resolved. The strategy now correctly displays in the portal, even when grid data is null. |
20-Mar-25 |
|
|
Invalid Date Logging in OmniChannel API |
The OmniChannel API was logging events with structurally valid but semantically invalid dates (e.g., the year 35), resulting in incorrect Hadoop directory structures. This caused log data to be stored under erroneous timestamps, affecting event tracking and analytics. A validation mechanism has now been implemented to check and prevent invalid date values, ensuring accurate logging and preventing issues with data organization in Hadoop. |
20-Mar-25 |