Search Merchandising in SAP Commerce: The Business-User Levers on Solr

The Solr guides in this library cover search from the engineer's side: architecture, indexing, infrastructure, query tuning. This guide covers the other half, the surface a merchandiser or search manager operates every day without writing code, because on-site search is a merchandising instrument as much as a technical one, and the business users who tune it need to understand their levers. Relevance, autosuggest behavior, synonyms, redirects, and audience-specific results are all configurable from Backoffice and SmartEdit, and a team that leaves them at defaults is leaving conversion on the table. This is the merchandiser's control panel, with the technical notes that keep the business changes from surprising the engineers.

Indexed Attributes and Weighting: What Search Can See#

Search can only rank on what is indexed. The indexed attribute list is set in Backoffice and can include any product attribute, whether it comes from typing or from classification. Two consequences business users control:

• If an attribute or media item must appear on the search results page, it must be in the index. A results page cannot show a badge, a price, or an image the index does not carry. When merchandising asks for a new element on the results tile, the answer often starts with "add it to the index."
• Weighting drives relevance. Indexed attributes are weighted (through Free Text Query configuration in Backoffice, and/or XML), and the weights decide which attribute matters most when Solr scores a keyword match. A product name match should usually outrank a description match; getting the weights right is the difference between "searching for the brand shows the brand's products first" and a frustrating scatter. This is merchandising work informed by relevance testing, not a set-and-forget.

Which indexed attributes feed spell check, autocomplete, and faceting is also a Backoffice setting, so the same attribute can be searchable, suggestable, and a filter, or any subset, by configuration.

The Configurable Search Bar#

The Search Bar Component in SmartEdit exposes autosuggest behavior directly to business users:

• Whether and how many keyword suggestions show.
• Whether and how many product suggestions show, and whether they include imagery.
• The minimum characters before Solr suggests, and the delay after that minimum before it fires (a tuning knob balancing responsiveness against Solr load).
• Whether facets show, letting customers refine after a search or on a category page.

Facets deserve merchandising attention: any attribute set as a facet in Backoffice becomes available for faceted search, and facets can be limited per category in the Commerce Search Perspective. Showing the right facets per category (size and color for apparel, capacity and voltage for appliances) is a merchandising decision that materially affects findability, and dumping every facet on every category is the lazy default that overwhelms customers.

Stopwords and Synonyms#

Two Backoffice-managed features shape which products a query returns:

• Stopwords: a list of irrelevant words ignored during search ("the", "these", generic terms like "product") so results stay relevant. Managed in Backoffice on top of the standard Solr mechanism.
• Synonyms: extend results with products that do not match the query literally. A customer searching "trousers" also sees "jeans" if you define them as synonyms. This is one of the highest-leverage merchandising tools, because it captures the vocabulary gap between how customers search and how your catalog is described.

Two technical notes that keep business changes working:

• Stopwords and synonyms require export to Solr and a full index run to take effect; they are not instant, and a merchandiser who edits a synonym and does not see it live needs to know a reindex is pending (coordinate with the operations schedule, the Solr infrastructure guide).
• They only work on fields configured with the appropriate filter (the Hybris stopwords filter) in the Solr schema; a plain string field with no filter ignores them. And when exporting, Commerce needs access to all Solr nodes including replicas, or the files do not propagate. These are the seams where a merchandiser's change quietly fails without engineering awareness.

Keyword Redirects#

Keyword redirects override default search behavior: when a customer's query matches a defined keyword, they go to a specified URL instead of a results page. Supported match types are EXACT, STARTS_WITH, ENDS_WITH, CONTAINS, and REGEX. The merchandising use cases are the point:

• "help" redirects to the FAQ or contact page.
• A brand name shows a curated CMS brand landing page instead of raw results.
• An exact product name goes straight to the product detail page.

Redirects turn the search box into a navigation and campaign tool, and they are pure business configuration.

Variant Grouping#

For catalogs with many variants (a shirt in eight colors), showing every variant clutters results. Variant grouping collapses results by a chosen variant level, usually the base product, to de-clutter. The merchandising decision is which level to group on: base product for the cleanest results, or a lower level if customers should see all options, or if you want per-variant control (a boost rule for red products, say). Group at the level that gives the most manageable, most conversion-friendly result set for your catalog shape.

Query Templates and Stemming#

Query templates let different search contexts use different query configurations: different indexed attributes, different facets, different scoring for exact/fuzzy/wildcard matches. The canonical use is as-you-type suggestions, which do not need facets and can be optimized for speed, versus the full search, which does. Templates are created in the Facet Search Configuration area of Backoffice, and they let a merchandiser optimize each search context independently.

Stemming decides how aggressively Solr reduces words to roots. The classic example: with stemming, a search for "run" matches "ran", "running", "runs", "runner"; without it, only literal "run". SAP Commerce defaults to the most aggressive stemming (Porter), which is often too aggressive for a product catalog (it can over-match and hurt precision), so the common advice is to dial it back to something that mainly handles pluralization. The algorithm flavors (Snowball/Porter, PorterStem, Hunspell dictionary-plus-rules, KStem which is faster and less aggressive) trade precision against recall, and the right choice depends on your catalog and language. This one crosses into engineering (it is a schema configuration), but the decision is a merchandising one about how forgiving search should be.

Adaptive Search: Different Results for Different Audiences#

The most powerful business-user search capability is Adaptive Search (the recommended module; the legacy Commerce Search module was deprecated years ago). Adaptive Search lets business users build search profiles that deliver dynamic, context-specific search configurations to achieve business objectives: different search behavior for different audiences, categories, or contexts, managed by merchandisers rather than developers. This is where search becomes personalization (the personalization and recommendations guides): boost seasonal products for one audience, promote high-margin items in one category, curate results for a campaign, all as business configuration. Understanding search profiles is what separates a merchandising team that shapes results to business goals from one that ships Solr's raw relevance and hopes.

The Merchandiser's Checklist#

• Indexed attribute list covers everything search must rank on and everything results pages must display
• Attribute weights tuned by relevance testing, not left at defaults
• Facets configured per category, showing the right filters, not every attribute everywhere
• Autosuggest (keyword/product suggestions, imagery, min-chars, delay) configured for the experience
• Synonyms capturing the customer-vocabulary gap; stopwords removing noise; both reindexed to take effect
• Keyword redirects turning the search box into a navigation and campaign tool
• Variant grouping at the level that de-clutters without hiding needed options
• Query templates optimizing suggestions versus full search independently
• Stemming dialed to the right precision/recall for the catalog, not left at aggressive default
• Adaptive Search profiles delivering audience- and context-specific results toward business objectives

On-site search converts when it is merchandised, not just indexed. Every lever here is a business-user control that shapes what customers find and buy, and the teams that own them (weighting, synonyms, redirects, profiles) turn search from a utility into a merchandising channel. The engineers keep Solr fast and correct; the merchandisers make it sell, and the handful of technical seams (reindex-to-apply, filter-configured fields, replica access) are where the two need to talk.