Release Summary 26.21 | Oct 21, 2026
The following key features and improvements, along with bug fixes, have been released in ADA Global Digital Experience Personalization (DXP) in the release version 26.21.
Recommend
Start and End Times for Merchandising Rules
You can now set a specific time, not just a date, for when a merchandising rule starts and ends. This applies to Rec Restriction, Rec Boosting, Manual Recommendation, Strategy, and Advanced Merchandising rules. Each date field now has a time picker next to the date picker, the start date defaults to the current date and time, and the end date can be left open with the no end date option. Rules activate and deactivate at the exact time you set, based on the time zone in the site configuration. If no time is entered, the start defaults to 12:00 AM and the end to 11:59 PM, and existing rules work exactly as before. The rules list now shows start and end timestamps. This feature is available when it is enabled in the site configuration.
Merchandisers can schedule time-sensitive campaigns, such as a promotion that runs from 8 PM to 8 AM, without publishing or unpublishing rules by hand.
Blocked Items Supported in Advanced Merchandising Rules
Placements driven by Advanced Merchandising rules now honor the bi parameter, which lets you pass the IDs of products to exclude from recommendations. This applies to rules that have Enable system rules and filters turned on, and it works the same way as for strategy-driven placements. Ranking, boosting, and restrictions in the rule stay as they are, and existing backfill behavior applies if blocking items leaves too few results.
Products already shown elsewhere on a page no longer repeat in recommendations, giving shoppers a cleaner, more varied experience.
Jira: ENG-33219
Attribute Name and Value as a Seed in Recommendation Requests
Recommendation requests now accept a new parameter, seedAttribute, which takes a single attribute name and value pair as the seed instead of a product ID. For example, seedAttribute=best-use:Camping uses best-use as the attribute name and Camping as the value. The parameter applies to all placements in the request that use a Configurable Strategy with the attribute top sellers or attribute top viewed model and the context product seed, and it returns recommendations for that attribute value. Requests that use a product ID work as before.
This suits pages without a natural seed product, such as author, seller, or category listing pages, and removes the need to maintain a placeholder product for each attribute value.
Jira: ENG-33385
Engage
Fixed Content Order in Campaigns
Engage campaigns now let you choose how selected content is sorted: Experience Optimizer (the default), User Affinity, or Fixed. With Fixed, you choose a content attribute with numeric values, such as rank, and an ascending or descending sort order. Content is returned in that order, and only content that has the attribute is returned. The preview shows content in the fixed order, and the selected option appears in the Rule Summary and is saved with the campaign.
You can now set up a single campaign for ordered, curated content instead of one campaign per content item. For example, a fixed order campaign can serve as the fallback for shoppers with no history, with a higher priority affinity campaign layered on top for shoppers who show interest in a category. The content list API also accepts a sort attribute and direction, so previews match what shoppers see. This simplifies setup and lets merchandisers curate the experience at scale.
Jira: ENG-33211, ENG-33042, ENG-33217
Engage Page Views and Usage Reporting
Engage reporting now includes page views for all page types, so you can see how much shoppers use Engage. Trend charts are available for all key metrics, including AOV, CTR, clicks, clicks per visit, orders, RPV, sales, units, page views, and page views per visit. A new Engage Usage Report also shows page views, the basis for billing. You can filter by date range with preset options, region, and channel, and in the performance reports by currency as well. Tables show the data for each selection, including by page type.
Merchandisers get a clear view of Engage usage and performance across pages and date ranges.
Ensemble AI
Virtual Try-On Jobs Now Follow the Ensemble AI Model Option
The virtual try-on job now runs only when Ensemble AI is enabled in the site’s model options. The whole Ensemble AI pipeline, including the LLM job, the Ensemble AI job, the recommendations step, and virtual try-ons, runs together when the option is enabled and stays off when it is disabled.
To stop outfits and virtual try-ons for a site, simply disable the option in that site’s model options. Other sites are not affected, and no processing is spent on sites that do not use Ensemble AI.
Jira: ENG-33441
Social Proof
Social Proof Refinements for Interval Selection, Recency, and Message Display
Interval selection for Social Proof messages now uses a cleaner layout that scales better than the previous long list of pills. The default minimum for recency messages is now zero, and when less than a minute has passed, messages show “1 min ago” for viewed, added to cart, and purchased messages.
Messages that reference a category, brand, or other attribute now appear only when that value exists for the product, so shoppers never see partial or blank messages.
Jira: ENG-33097
Discover
Discover Performance Report
A new Discover Performance Report is now available under a new Discover menu, covering New Discover and Discover powered by recommendations. You can filter by date range with preset options, currency, region, and channel.
The report includes visits, overall visits, views, clicks by page position, exits, attributable orders, sales, and conversions. KPIs include conversion rate, revenue per visit, average order value, exit rate, findability, and weighted CTR. Every metric has a trend chart, and conversion rate, revenue per visit, and average order value also compare Discover with non-Discover visits. Table data is shown for all selected filters. Merchandisers can see how Discover affects engagement and revenue, and fine-tune their configurations.
Enterprise Dashboard
Strategy Rules Sorted by Most Recently Updated
The Strategy Rules list now opens sorted by most recently updated, so the rules you created or changed recently appear first. This makes recent rules quick to find, saving time when managing strategies.
Jira: ENG-32784
Privacy Mode Captured in Visit Data
Visit data now records the privacy parameter (privm = true or false) whenever it is passed. It is also recorded with views and clicks when available.
This lays the groundwork for reports that show how privacy mode affects attribution and performance, helping merchandisers understand its impact on their results.
Jira: ENG-33089
Other Feature Enhancements
The following feature enhancements and upgrades have been made in the release version 26.21.
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Social Proof: Daily Data in Social Proof Aggregated Campaign Report |
The aggregated Social Proof campaign report now includes daily data across all campaigns, in addition to the overall totals. This gives you a breakdown by each date in the selected range, matching the variation level report. Merchandisers can use the daily data to view trends over time. |
21-Oct-26 |
|
|
Social Proof: Product Coverage Count in Social Proof Campaign Reports |
Social Proof campaign reports now show the number of unique products that displayed a social proof message. The count is available for each variation and across all campaigns, so you can see how much of your catalog is covered by social proof messaging. The report API also returns this count in its response. Merchandisers can quickly see the reach of each campaign and variation, and spot gaps in product coverage. |
21-Oct-26 |
|
|
Ensemble AI: Daily Outfit and Virtual Try-On Count Report in Ensemble AI |
Ensemble AI now includes a report that tracks how many outfits and virtual try-ons are generated each day. For each style, it shows the overall count and the incremental count compared with the previous day, for both outfits and virtual try-ons. Each outfit is identified by its combination of products, so the same outfit is recognized from one day to the next, and virtual try-ons are matched through their outfits. This separates new outfits from existing ones and gives an accurate picture of daily growth. Merchandisers can see how much variety is being created for shoppers and how it changes over time, and then adjust their styles as needed. |
21-Oct-26 |
|
|
Find: Enrichment Attributes for Find Facets and Filters |
Find can now use enriched product attributes, including occasion, festive, length, and style, for facets and filters. These attributes can be used to narrow results and are stored with the product data. Shoppers can refine search results using richer product attributes, which helps them find the right products faster. |
21-Oct-26 |
|
|
Recommend: Run Status for Initial Builds of Unscheduled DSW Strategies |
The one-time initial build for a DSW strategy without a recurring schedule now reports its status the same way scheduled runs do. Its progress through each stage and its final status appear in the strategy list and in job history, so you can see how the build went instead of seeing no data. Strategies with a recurring schedule are unchanged. Users get a complete view of every strategy build, whether scheduled or not. |
21-Oct-26 |
|
|
Enterprise Dashboard: Faster Loading for Dynamic Experiences, Guided Selling, and Social Proof |
The new Dynamic Experiences page, Guided Selling, and Social Proof no longer make an unused request when they load. This reduces unnecessary background activity so these pages load more efficiently, with no change to how they look or work. |
21-Oct-26 |
|
|
Recommend: Campaign Tracking for Email Recommendations |
The campaign parameter (cpi) sent in email page recommendation requests is now carried through to the click URL. Views and clicks are recorded under the campaign you pass, and the data appears under the correct campaign in the Recs in Email report. Requests without a campaign parameter work as before. Marketers using Active Content can now see email recommendation performance by campaign. |
21-Oct-26 |
|
|
Engage: Central Data Store for A/B Testing |
A new data store now holds the foundation for A/B testing. Each test is linked to one Dynamic Experience and one JIW that holds the test configuration. It records the test name, site, primary and secondary success metrics, and a lifecycle status of draft, active, paused, or stopped. This gives A/B tests a single place to be defined and managed, and supports the upcoming A/B testing capabilities. |
21-Oct-26 |
|
|
Engage: Static A/B Tests as a New Dynamic Experience Type |
Dynamic Experiences now support a new type, STATIC_AB_TESTING, for storing static A/B tests. These tests are not listed on the Dynamic Experiences page by default, and existing types are unchanged. A performance review was also completed to keep response times fast for these tests, since the page waits for them to load. This lays the groundwork for static A/B tests without adding a separate data store, while keeping page loads quick. |
21-Oct-26 |
|
|
Ensemble AI: Existing Ensembles Protected When Image Downloads Fail |
When the image download job cannot download most of the images for seed products, the job now stops with an error. The Ensemble AI job and the Image Similarity update also stop for that day instead of running, so existing ensembles and similarity results are kept as they are. This prevents a temporary image download problem from erasing ensembles for many seed products, keeping recommendations stable. |
21-Oct-26 |
|
|
Ensemble AI: Site-Specific Limit for Ensembles per Seed Product |
The maximum number of ensembles per seed product can now be set for each site instead of as one global value. The limit applies across all styles for the site, and one site is now set to two ensembles per seed product. This gives each site control over how many ensembles are generated, helping keep outfit variety focused and manageable. |
21-Oct-26 |
|
|
Platform: Quality Validation for Enriched Product Attributes |
The multi-modal enrichment pipeline now includes a validation step that checks the quality of attributes generated from product images and text. The validator reviews each attribute against the product’s image, text, and captured attributes, and ranks it by confidence. Only attributes that meet the confidence level are added to the enrichment results. The confidence level is configurable and defaults to 90%. This keeps enriched product data accurate and reliable, which improves search, filtering, and recommendations. |
21-Oct-26 |
|
|
Platform: Stable Status Tracking for Jobs with Missing Update Times |
The Common Status service now handles jobs that have no recorded update time. The Common Status API returns a response for these jobs, and the service stays healthy instead of restarting. All status events continue to be processed as normal. This keeps job status tracking reliable and avoids service interruptions. |
21-Oct-26 |
|
|
Platform: Faster Item Lookups in Streaming View |
Item lookups and listings in Streaming View now use a more direct data lookup, which improves performance and keeps it consistent as the catalog grows. Getting an item by internal or external ID, and paging through items, work as before, and existing page cursors remain compatible. Item retrieval stays fast and reliable for large catalogs. |
21-Oct-26 |
|
|
Discover, Find: Smoother Global Rank Updates |
The global rank publisher job now updates the existing live dataset by default instead of creating a new dataset for each run. This removes the gap in global ranking and user affinity data between delete and insert updates, and reduces processing backlogs. Sites that need a new dataset for each run can still be configured to do so. Global ranking stays continuous and updates are more reliable. |
21-Oct-26 |
Bug and Support Fixes
The following issues have been fixed in the release version 26.21
|
Jira # |
Module/Title |
Summary |
General Availability |
|---|---|---|---|
|
Enterprise Dashboard: Intermittent Recommendation API Errors Resolved |
We have fixed an issue where the recsForPlacements recommendation API occasionally returned an internal error for a short period, typically 10 to 15 minutes, before recovering. Recommendation requests are now served reliably. |
21-Oct-26 |
|
|
Recommend: Content Items Excluded from Compatibility Recommendations |
We have fixed an issue where content items could appear in Compatibility recommendations. Compatibility results now include only actual compatible products, and other results are unchanged. |
21-Oct-26 |
|
|
Recommend: Environment Name Now Set Correctly in Portal Apps |
We have fixed an issue where the environment name was not set in the Advanced Merchandising, Engage, and Datamesh portals. These portals now identify their environment correctly, so features that depend on it work as expected. |
21-Oct-26 |