Global Search, PLPs & Personalization

The existing SFCC assessment identified concrete limitations: manually maintained search suggestions, limited control over featured products in the flyout, autocorrect issues, keyword-pinning constraints and difficulty scheduling or isolating search changes. Search testing also included category traffic in the same pipeline, making results harder to interpret.

The objectives were to improve the relevance of suggestions and results, give business teams more useful configuration and testing tools, and maintain the filters, sorting and other storefront behaviors customers already used.

Brand grouping, color slicing and editorial grid content were established parts of Ralph Lauren’s category experience. Moving to Bloomreach could not mean losing them.

EMEA teams needed to preserve brand order while allowing products within a brand to respond to performance; expose color variants as separate tiles while retaining the relationship to one PDP; select different default colors by category; and merchandise products around content slots. Preserving these behaviors required deliberate changes across data mapping, backend logic, storefront behavior and vendor tooling.

Audience-specific search was evaluated separately from the platform migration. It used shopper signals to adapt results for defined groups, with its own product lead, test populations and measurement. This allowed the program to distinguish improvements from the core search engine, category merchandising and personalization.

I coordinated a six-vendor evaluation, including the RFP, vendor demonstrations, regional scorecards and technical assessment. The selection was a trade-off: Algolia scored higher for search, while Bloomreach scored substantially higher for merchandising. In the EMEA merchandising scorecard, Bloomreach received 43/50 versus Algolia’s 23/50.

The decision record favored Bloomreach’s regional merchandising fit, real-time publishing, business-user tools, segmentation, A/B testing and analytics. It also considered the vendor’s willingness to work through Ralph Lauren-specific requirements such as grid breaks and brand search placement.

Commercial evaluation covered onboarding, annual licensing and support, with scope defined through regional website and app demand, language catalogs, product/SKU volumes, and separate search-versus-category traffic. This made the usage assumptions and differences between vendor proposals explicit.

The resulting plan phased search ahead of the more complex merchandising migration, with PLP work sequenced by region. This kept the choice grounded in operating requirements and implementation risk, rather than a generic comparison of AI features.

Brand grouping was an existing capability to preserve. The design work evaluated multiple ways to retain the required brand hierarchy while allowing ranking within that structure.

The February 2024 approach review rejected a grouping-API implementation because of custom caching, pagination, SEO/performance considerations and API-call constraints. It recommended an attribute-weight approach using market-specific brand rankings and a single request. Earlier group-merchandising and sequential-lock screenshots document an evaluated option, not the final implementation decision.

My role was to make the business behavior precise enough for engineering and Bloomreach to assess: what must remain fixed, where the algorithm could operate, how markets could differ and what the solution would cost in complexity and performance.

Color slicing also predated the migration. Preserving it required mapping master products, variants, category-specific color choices and storefront links into Bloomreach’s data model and response behavior.

The design and enhancement discussions addressed how sliced variants receive behavioral learning and how they should appear in suggestions. A separate PLP tile, a variant in the index and the product identifier credited by a pixel event are connected decisions. They cannot be treated as a visual change alone.

The integration separated business-data ownership, platform behavior and storefront presentation. SFCC and upstream processes remained responsible for source product and category information. Bloomreach provided discovery responses and merchandising tools. RL implementation work handled how those responses became the web and app experience.

Where native capability did not match an existing requirement, the work required custom backend development, vendor-tool changes or a revised approach. The enhancement backlog made those dependencies explicit rather than treating the migration as a standard connector installation.

The initial search migration preserved the core UI. Subsequent search-flyout work considered entry before typing, incomplete and complete queries, multiple suggested terms, product and category suggestions, no-result behavior and the transition to all results. Research prioritized mobile and identified issues such as a false “no results” message while a customer was still entering the first characters.

Bloomreach powered query suggestions, product suggestions and category suggestions. Recent searches were part of the wider UX and were intended to be stored and served by the client; they should not be described as generated by Bloomreach. Static trending and popular-category content also had separate content ownership and implementation considerations.

This separated search-generated suggestions from client history and editable content, giving teams a clear implementation responsibility for each part of the experience.

The submitted search-results experience retained filtering and sorting while replacing the underlying retrieval and ranking service. The value of the migration was more relevant results and improved configuration and testing, not the invention of department filters or other controls that already existed.

Requirements covered valid facets and sort options, synonyms, redirects, correction behavior, product eligibility and no-result handling. RL remained responsible for storefront display logic, while Bloomreach supplied the results and associated facet values.

I owned the integration specifications and mapping requirements connecting SFCC data with Bloomreach. The work distinguished product and SKU entities, preserved identifiers and relationships, and defined transformations for localized brand information, online eligibility and regional availability.

Category, product, inventory and price inputs had separate update requirements, including full feeds and deltas. Regional indexing also needed language-level content with market-level views for differences such as price, currency and availability.

Completed integration tasks included publishing SFCC files and availability inputs to S3, delivering catalog deltas to Bloomreach ETL over SFTP, and sending daily full pricebooks. A dedicated record-type attribute removed the need for ETL to infer product type. Regional follow-up work included EU feed validation, Global-e dynamic pricing and Middle East pricebook exports. These were distinct delivery paths and transformations, rather than one generic feed.

The original system-context and EMEA index diagrams make that complexity visible. They show why this was a commerce-platform integration, not only a frontend search change.

The PLP implementation needed to preserve several distinct behaviors:

App requirements extended the integration beyond the web search box. They covered shared suggestion payloads, valid filters and sorts, eligibility restrictions, pagination, redirects and fallback behavior.

The implementation records include MuleSoft data mapping, exposing sort options, routing controls and the ability to switch between Bloomreach and SFCC. One integration decision used an SFCC proxy for Bloomreach autosuggest to preserve hidden-product and customer-entitlement checks. App discussions also addressed mapping Bloomreach facet values to the SFCC fields expected by existing clients.

These were compatibility and operational decisions, not simply API connectivity. The records support the app-integration scope without implying that every app surface remained permanently on Bloomreach.

I worked with the personalization product lead and regional stakeholders as segmented search was tested and expanded. The documented approach distinguished manually managed boost/bury rules for defined audiences from additional automation and real-time-segmentation opportunities.

The historical default and Big & Tall T-shirt comparison shows how the same query could produce different results for different audiences. The Women’s High Sale Affinity experiment was measured independently, with its own cohort and commercial outcome.

I coordinated global product and technology teams, regional search and merchandising stakeholders, Bloomreach and implementation partners. The stakeholder, access and training register spans more than 80 people across global, NA, EMEA and APAC groups, including the business users involved in adoption.

The initial vendor agreement committed approximately $1.2M over two years for subscriptions, implementation, training and strategic services. That is the vendor commitment, rather than total program spend or a direct-report team budget.

UAT had to validate the complete journey: feed identity and eligibility, pricing and inventory, regional content, search suggestions, results, filters, sorting, category order, color behavior, analytics and fallback controls.

The work included checking that the new service preserved existing capabilities and that the metrics represented the experience being tested. For example, the EMEA merchandising test started with eight PLPs at 50/50 allocation, then expanded after reporting and configuration work. Search, PLP and audience experiments remained separately measured.

The Bloomreach enhancement backlog covered category facet controls, brand ranking, curated grids, sliced-product behavior, multilingual relevance, staging/production workflows and related dependencies.

I used these requirements to coordinate what could be delivered through RL implementation, what required Bloomreach changes and what needed a later phase. The RFP and demo scorecards, design decisions, integration maps and enhancement records connected the original business requirements to the delivery plan.

The November 2024 EMEA PLP A/B test recorded approximately +1.5% revenue per visit versus SFCC. The November 1–28 result was $4.89 for Bloomreach versus $4.82 for SFCC.

The program combined commercial platform selection, customer-experience design and detailed integration ownership. Its complexity came from improving discovery while retaining the regional merchandising practices and operational controls already embedded in Ralph Lauren’s business.

The results belong to different test populations and windows. They should be presented separately rather than averaged into a single global uplift.