Release Summary 26.19 | Sep 21, 2026
The following key features and improvements, along with bug fixes, have been released in ADA Digital Experience Personalization (DXP) in the release version 26.19.
Social Proof
Configurable Exploration Settings for Social Proof Optimization
Social Proof optimization now supports configurable exploration settings instead of relying solely on fixed defaults. Previously, the system applied a fixed initial exploration period of 2 days and an ongoing exploration rate of 10%. These values can now be configured directly through the portal, giving clients the flexibility to set values that better fit their traffic patterns and business needs, with the training and underlying model applying the configured values automatically.
This gives clients more control over how aggressively the system explores new message variants versus exploiting known top performers, tuned to each site's traffic and business priorities.
Jira: ENG-32310
Flexible Context Parameters for Social Proof Optimization
Merchandisers can now choose which context parameters to include when optimizing Social Proof messaging, rather than being required to use a fixed set. Optimization now runs correctly with or without any context parameters selected, and day of week has been made an optional parameter rather than a required one.
This gives merchandisers more control over how social proof messaging is optimized for their specific use case, allowing optimization to be tailored to the context that best fits each site's needs rather than a one-size-fits-all configuration.
Jira: ENG-33123
Page Type Tracking for Social Proof Messaging
The track experience API call for Social Proof messaging now captures the page type on which a message was served, distinguishing between item pages and list pages. This means every tracked message can be tied to whether it appeared on a single-product page or a multi-product page, without yet identifying the specific page type beyond that distinction.
This additional context makes it possible to analyze social proof message performance by page type, helping to understand how messaging performs differently on item pages compared to list pages. It lays the groundwork for more detailed performance analysis and reporting as page type tracking is refined further in future updates.
Jira: ENG-33074
Views Set as the Default Metric in Social Proof Message Reporting
Social Proof message level reporting now defaults to Views as the primary metric, ensuring a consistent starting point regardless of which page types are enabled for a given site. List page clicks and list page click‑through rate have been moved to the end of the metrics list, keeping the report focused on the metric most broadly applicable across item and list pages.
This gives users a more consistent and reliable starting view when reviewing social proof message performance, reducing the need to manually adjust the metric selection based on how a site's page types are configured.
Jira: ENG-33134
Enterprise Dashboard
Ensemble Exposed in the API Response for Recommended Outfits
Recommended outfits returned through the API now include ensemble score. Previously, these values existed internally but were not consistently exposed for server-side consumption.
This allows merchandisers and developers using server-side integrations to sort and rank ensembles by relevance directly from the API response, without needing to re-compute scoring on the client side, making server-side integrations faster to build and easier to maintain.
Jira: ENG-33133
Exclude Brands from Ensemble AI Outfit Generation
Merchandisers can now exclude specific brands from outfit generation in Ensemble AI, ensuring those brands do not appear in the resulting virtual try-ons. An "Exclude brands" option lets merchandisers select one or more brands from the client's catalog to exclude at the seed product level, applicable to both structured and free-form style definitions. For free-form styles, an additional "Apply for all parts" option extends the exclusion to every part of the outfit, not just the seed product.
Once "Exclude brands for all parts" is configured, no product from an excluded brand appears in any outfit generated within that configuration's scope. Existing outfits containing products from newly excluded brands are regenerated or filtered on the next refresh cycle, so exclusions take effect without requiring manual cleanup.
This gives merchandisers precise control over brand representation in generated outfits, supporting merchandising decisions such as honoring brand exclusivity agreements or keeping certain brands out of specific outfit contexts.
Jira: ENG-33205
Reports Tab Added to the New Dynamic Experiences Interface
Digital optimization managers can now view performance reports directly within the new Dynamic Experiences interface, without navigating to a separate page. A new Reports tab offers the same reporting experience already available for Social Proof experiences, including date pickers, a channel selector, a currency selector where applicable, table results, and graphs, along with the full set of metrics from the existing Dynamic Experiences report.
This gives digital optimization managers a consistent, complete view of experience performance without leaving the page they're working in.
Jira: ENG-31592
Extended Lookback Period for New Arrivals Model
The lookback period for the New Arrivals model can now be set up to 180 days, up from the previous limit of 31. This gives merchandisers with slower-turnover catalogs, such as furniture or appliances, the flexibility to surface genuinely new products instead of an empty or stale new arrivals set. The increased range applies to both configurable strategies: New based on recency, and New based on rank of top sellers among new products.
The lookback field now shows its valid range and provides clear feedback when a value is too low, too high, or left empty. A related issue where clearing the field could silently save an invalid value has also been fixed.
Jira: ENG-33291
Granular Run Status and History for Data Science Workbench (DSW) Strategies
The Strategies list and Detail panel in DSW now show detailed, per-stage run status instead of a static checkmark. Each strategy row displays a status pill, Updated, Running, Failed, Canceled, or Queued, with a caption showing the current stage and elapsed time, the failure point, or a countdown to the next run. All timestamps now display in the site's configured time zone instead of being hardcoded to Pacific Time.
Clicking a strategy opens a redesigned Detail panel showing configuration and schedule details that can be edited directly, along with the latest run's status, timing, and a three-step timeline covering the Airflow job, model build, and model publish stages. A run history section lists prior runs, each expandable to show its full chronological event log.
Jira: ENG-33019
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 26.19.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Social Proof: Segment Support for Social Proof Optimization Context Parameters |
Social Proof optimization now supports segment as an additional context parameter for training data, alongside the previously supported primary category, price, and channel. Segment can be configured similarly to the other parameters, and training data missing any of these parameters is simply excluded from that data point. This allows social proof messaging to be optimized with a more complete view of shopper context, helping deliver more relevant messaging based on the segments a shopper belongs to. |
21-Sep-26 |
|
|
Social Proof: Improved Calculation of Context and Action Feature Interactions in Social Proof |
We have refined how context and action features are weighted and combined in Social Proof messaging, improving the accuracy of how these interactions are calculated. This results in more accurate and reliable social proof messaging, ensuring the underlying calculations better reflect the intended weighting between context and action features. |
21-Sep-26 |
|
|
Enterprise Dashboard: More Accurate Section Tracking on the Online Help Landing Page |
Usage tracking on the Online Help landing page now reads the section group directly from the updated breadcrumb structure, rather than relying on a hard-coded mapping. The complete breadcrumb path is also captured as a custom property. This makes usage tracking more accurate and easier to maintain as documentation content is reorganized, since section tracking now stays in sync with the actual content structure. |
21-Sep-26 |
|
|
Enterprise Dashboard: Placement Profile Page Refinements |
Several usability and accuracy fixes have been made to the Placement Profile page. Sites using multiple currencies can now filter the Performance tab by currency, currency now displays correctly in price range columns, labels and attribute values display correctly (with a "+n more" indicator for longer lists), sitewide restriction and boost rules show consistent context information, and an EXPIRED badge now appears next to PROD and INT badges for expired rules. A loading delay that caused the Performance tab to flip back to Strategies has also been fixed. These refinements make the page more reliable and easier to read when reviewing and managing placement configuration. |
21-Sep-26 |
|
|
Enterprise Dashboard: Complete Tracking for Multiple Social Proof Messages |
When multiple social proof messages are enabled and shown to a shopper, all displayed messages are now tracked as part of the track experience call, comma separated, rather than only the first message. The corresponding message templates are tracked alongside the actual messages, also comma separated when multiple messages are shown. This gives a complete, accurate view of which messages were actually displayed, allowing reporting and optimization to account for the full set of messages rather than just the first one shown. |
21-Sep-26 |
|
|
Platform: Asynchronous Checkpoint Event Processing |
Checkpoint event execution in NCA now runs asynchronously on a dedicated thread pool, with control over the number of concurrent executions, instead of blocking other events while it runs. This lets NCA continue processing other events during checkpoint execution, improving throughput and preventing checkpoint operations from becoming a bottleneck. |
21-Sep-26 |
|
|
Recommend: Boost Rules Performance Reporting Added to the Left Navigation |
Boost Rules Performance reporting is now accessible directly from the left navigation under Reports > Recommend > Boost Rules Performance, matching the label shown on its landing page. This makes the report easier to find, letting users navigate to it directly instead of needing to know its URL or reach it through another path. |
21-Sep-26 |
|
|
Platform: Improved Accuracy of Sync API Item Counts |
We have improved the accuracy of item counts returned by the Sync API. Some sites were seeing negative values for recommendable and non-recommendable counts in the item count data, caused by counters being updated before initialization and race conditions between backfill and live ingestion. Validation has been added to prevent negative values, ensure recommendable and non-recommendable counts always sum to the total, and correct any corrupted counts automatically after ingestion completes. Counts are also updated correctly when items are deleted, and the Sync Count API response format has been corrected. This gives a more reliable, accurate view of catalog item counts, reducing confusion caused by incorrect or negative values. |
21-Sep-26 |
|
|
Platform: Improved Performance for the Async Count API |
The Async Count API now uses the same efficient logic as the newer Sync Count API to calculate catalog counts, replacing the heavier processing it previously relied on. This reduces unnecessary load on streaming components, particularly for large catalogs, while keeping the existing response structure and processing pipeline unchanged. This gives faster, more efficient catalog count retrieval without requiring any changes on the integration side, since the response format and behavior remain fully compatible with existing usage. |
21-Sep-26 |
|
|
Platform: Improved Performance for Catalog Enrichment Processing |
We have improved the performance of catalog enrichment processing by introducing a lightweight index table for enrichment items, replacing a query pattern that was causing timeouts while processing datasets with high item volumes. The new index is updated automatically as items are ingested and cleaned up when items or datasets are purged, scoped to the affected dataset so other datasets remain unaffected, with re-ingestion of the same item handled without creating duplicate entries. This reduces the risk of processing timeouts during enrichment, making dataset processing more reliable and consistent. |
21-Sep-26 |
Bug and Support Fixes
The following issues have been fixed in the release version 26.19
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Discover: Fixed Assortment and Package Boost Not Applied in Browse Results |
We have fixed an issue where products with an assortment or package boost applied were not being ranked at the top of Browse API results as expected. Boosted products now correctly receive their configured score boost and are ranked accordingly. |
21-Sep-26 |
|
|
Enterprise Dashboard: Fixed Contextual Help Opening the Wrong Document After Switching Sites |
We have fixed an issue where switching sites on certain portal pages could leave a leftover query parameter in the URL, causing contextual help to open the wrong Online Help document or the OLH home page instead of the correct one. This has been resolved centrally across all affected pages, so contextual help now consistently opens the correct document after a site switch. |
21-Sep-26 |
|
|
Enterprise Dashboard: Fixed Missing Usage Tracking on Additional Dashboards |
We have fixed an issue where usage tracking was not working correctly on the Advanced Merchandising, Engage, and dmportal dashboards due to an environment detection gap. Tracking now works consistently across these dashboards, reporting to the same project already used for the main portal dashboard. |
21-Sep-26 |
|
|
Enterprise Dashboard: Fixed Experience Deletion Overlay and Group Name Localization on New Dynamic Experience |
We have fixed two issues on the New Dynamic Experience interface. Deleting an experience now displays a full-page overlay that blocks interaction until the operation completes, matching the existing behavior for deleting a variation. We have also fixed the group name on certain templates, which was hardcoded and did not localize correctly, so it now renders correctly across all supported locales. |
21-Sep-26 |
|
|
Social Proof: Fixed Default State of the Aggregate Metrics Option in Social Proof Coverage Report |
We have fixed the default state of the aggregate metrics option in the Social Proof Coverage Report, which was checked by default even though most clients do not use aggregate metrics. The option is now unchecked by default, so clients see their report metrics immediately without needing to uncheck it first. |
21-Sep-26 |
|
|
Streaming Catalog: Fixed Items Being Silently Dropped Due to Missing Property Definitions |
We have fixed an issue in NCA where a failed property definition lookup would silently return no result, causing items to be dropped without any indication of the underlying issue. Failures are now logged properly, failed calls are retried, and the consumer can optionally shut down to allow the application to restart and recover. |
21-Sep-26 |
|
|
Streaming Catalog: Fixed Category Ordering Mismatch Between View and rfadsFind APIs |
We have fixed an issue where the category array for a product could return in a different order from the Find API than the order preserved by the View API. Category order is now synced consistently, ensuring the Find API returns categories in the same hierarchy order as originally ingested. |
21-Sep-26 |
|
|
Platform: Fixed Enrichment Sidekick Restart Failure When Reprocessing a Deleted Dataset |
We have fixed an issue where the enrichment sidekick could fail to restart with a "Dataset is not available for deletion" error. This occurred when the sidekick was restarted during a narrow window where a cancelled dataset had already been purged while another dataset was still being published, causing it to reprocess a stale delete event for a dataset that no longer existed. The sidekick now restarts reliably in this scenario without erroring out. |
21-Sep-26 |