October - November 2026

A/B Testing for Recommendation Strategies

Take the guesswork out of recommendation optimization. Unbxd now supports widget-level A/B testing, allowing retailers to compare two recommendation algorithms and measure which one delivers better business results.

What’s New

• Added A/B test configuration under Experiences in the Recs console.
• Enabled traffic splitting between two variations, each using a different recommendation algorithm.
• Added configurable winning KPIs, including click rate, cart rate, conversion rate, average order value, and revenue per visit.
• Added reporting for impressions, clicks, carts, orders, revenue, and conversion rate.

Why It Matters

• Make data-driven decisions: Compare algorithms using measurable performance outcomes.
• Optimize conversions: Identify recommendation strategies that better support shopping and purchase goals.
• Measure what matters: Evaluate performance against the business KPI that matters most.

Note: Recommendation A/B testing is available on request and is not supported for every site key.

For more details, refer to the A/B Testing documentation.

Consistent Product Previews, Tailored to Your Catalog

Make recommendation previews more useful and easier to evaluate. Product Card Mapping now lets retailers customize how product cards appear across the Unbxd Recs console, from images and titles to prices, variants, and additional product attributes.

What’s New

• Added configurable hero image sizes, product titles, prices, and image URL fields.
• Introduced variant title mapping and controls to show or hide variants.
• Enabled up to five additional product metadata fields on recommendation cards.
• Added image URL prefix support for catalogs that use relative image paths.
• Applied the same mapping across Algorithm Preview, Experience Preview, and Recommendations Website Preview.

Why It Matters

• Clearer product previews: See the catalog attributes that matter when evaluating recommendations.
• More consistent evaluation: Use the same product card configuration across Recs previews.
• Simpler configuration: Map catalog fields to product cards without changing the live storefront.

Note: Recommendation Product Card Mapping is configured separately from Search Product Card Mapping. These changes affect console previews only, not the live storefront.

For more details, refer to the Recommendation Product Card Mapping documentation.

Smarter Cross-Sell, With More Control Over Every Recommendation

Make cross-sell strategies easier to manage and refine. Unbxd introduces AI-generated cross-sell rules for unmapped categories, configurable category priorities, and flexible product resolution to help retailers shape how complementary products fill recommendation slots.

What’s New

• Added AI-generated cross-sell rules for categories without existing manual mappings, with review and approval before activation.
• Introduced category priorities to control which companion categories fill recommendation slots first.
• Added relaxed and strict product resolution options for custom cross-sell algorithms.
• Introduced popularity-based product ordering in strict resolution mode.
• Updated category mapping limits based on existing usage.

Why It Matters

• Faster setup: Reduce manual effort when extending cross-sell coverage to unmapped categories.
• More relevant product combinations: Control which complementary categories appear first.
• Greater flexibility: Choose between broader fallback-based recommendations and strict category-specific results.

Note: AI-assisted cross-sell requires AI Complete The Looks mappings to be enabled. Strict resolution is available only for custom algorithms.

For more details, refer to the AI-Assisted Cross-Sell documentation.

Bring Product Variants Into Every Recommendation

Give shoppers more ways to find the right product. Unbxd Recommendations can now return variant details alongside recommended products, making it easier to surface options such as different sizes, colors, or styles in the shopping experience.

What’s New

• Added variant IDs, variant counts, total variant counts, and variant details to V3 Recommendations API responses.
• Introduced controls to specify how many variants to return for each recommended product.
• Extended console previews to display mapped variant fields through Product Card Mapping.

Why It Matters

• More relevant product discovery: Help shoppers explore product options directly from recommendations.
• Greater control: Choose how many variants appear with each recommended product.
• Easier evaluation: Preview variant information in the Recs console before implementation.

For more details, refer to the Variants: Recommendations documentation.

Store Based Recommendations: Know What's Available at Every Store

Deliver more relevant product recommendations with store-level availability. Unbxd Recommendations can now filter out products unavailable at a selected store, helping shoppers discover items they can actually purchase locally.

What’s New

• Added store-specific filtering to the V3 Recommendations API using store identifiers.
• Included store availability details in product responses, with support for combining store-level recommendations and variant data.

Why It Matters

• More relevant recommendations: Show products based on availability at the selected store.
• Better shopping experiences: Help shoppers discover locally available products.
• Greater flexibility: Combine store availability and product variants in a single request.

For more details, refer to the Store-Level Recommendations documentation.