Creating Semantic Search Campaigns

Experience Search provides a powerful way to enhance how users find products on your site through image-based Visual Search or text-based Semantic Search. This guide walks you through setting up and optimizing your Semantic Search campaigns to deliver highly relevant search results that drive conversions.

When you first access Experience Search, you'll see a launch message: Click Get Started, and then select whether to start with Visual Search or Semantic Search. 

Before you start: Define a default search experience variation

Before you begin working with either Semantic Search or Visual Search, you must create a default campaign. This campaign appears automatically when you first access each campaign type.

The default experience acts as a fallback — it's available to all users, all the time, ensuring that search is always active and returns results, even when no specific targeting rules are met.

Because of this, the default campaign doesn't support targeting conditions. Other than that, creating it follows the same steps as any other campaign, as described in this article.  

Now you're ready to begin your work with Experience Search campaigns.

Step 1: Create a new experience

1. Go to Experience Search › Semantic Search box › Edit.
2. Click the plus sign to create a new experience.
   
3. Name your experience, then click New Variation (or copy an existing variation). 
4. Set your targeting conditions, including:
   • User queries: Target specific queries and tailor their search results
   • Who: Target by user behavior, properties, technology, or geo-location
   • Where: Target specific pages or page types
   • When: Schedule the experience for a specific timeframe

5. Click Next to go to the Variation screen.

Targeting by search query

By default, every experience targets all search queries. If you want to target specific search terms and provide tailored search results, switch off the toggle (to the Users search for position) and enter your keywords in the queries field. Type or copy a word or term, and then press Enter to create a search tile.

• Semantic Search uses the OR condition for the search terms.
• Queries must be an exact match for words and spelling, but are not case sensitive.

Step 2: Create your variation and search strategy

Each experience can have multiple variations, each providing different scenarios for A/B testing. To create a new variation, click New Variation or Copy from Another Experience.

In the Variation screen, the left-side panel is where you define your search strategy, while the right side provides a real-time preview of the results in a dedicated playground—so you can see exactly how your changes will impact the user experience.

The Strategy tab

This is where you configure how your search results are displayed and ranked. 

Boost & Bury

Product attributes

Boosting or burying by product attributes enables you to influence search results based on attributes in your product feed. Select a feed column and the values you want to promote or demote, then adjust the boost or bury strength using the slider.

This makes it easy to prioritize specific brands, categories, colors, margins, inventory statuses, or other feed values, while keeping results relevant to the shopper’s query. 

Algorithm Studio

In Algorithm Studio you can create custom algorithms that address your unique business needs. You can boost products in search results based on performance metrics, such as views, purchases, or add-to-cart actions, or based on business data, such as profitability or stock levels.

• Step 1: Create your algorithm in Algorithm Studio
  Open Algorithm Studio and create the algorithm to use for boosting search results. For example, select a ranking factor of product views (high to low) to boost search results by popularity. 
  
• Step 2: Boost search results by your algorithm 
  1. In Experience Search › Edit Semantic Search › Edit Variation › Boost by, click Add and select your algorithm. 
  2. Use the slider to adjust the boost's strength. Find the right balance between ensuring relevance and maximum boosting of business factors. 
     

Personalized by affinity

Deliver hyper-personalized, 1:1 tailored search results based on each user’s affinity profile. By promoting products that better match user preferences, this feature improves search results relevance and can help drive stronger business outcomes, such as higher conversion rates and increased revenue.

Experience Search uses the same affinity attributes as the affinity recommendations algorithm. Configure these attributes in the product feed.

• Step 1: Configure your product feed
  In Assets › Data Feeds › Product Feed , set your affinity profile properties: 
  
• Step 2: Boost search results by affinity personalization
  1. In Experience Search › Edit Semantic Search › Edit Variation › Boost by, click Add and select your algorithm. 
  2. Use the slider to adjust the boost's strength. Find the right balance between ensuring relevance and maximum boosting of affinity. 
     

Note that the same restrictions that apply to affinity attributes in the affinity recommendations algorithm also apply to this feature

Filters

Filters are a powerful way to refine search results. Use them to:

• Include: Show only a specific subset of products
• Exclude: Remove certain products from search results (say, recently purchased items)
• Pin: Fix a product in a specific position in the displayed results.

To create a filter rule:

1. Click Add. 
2. Name the rule.
3. Select the rule type (Include, Exclude, Pin).
4. Select the products and values:
   • Include and exclude: Select a product feed column and values to include or exclude
   • Pin: Use the product name or SKU and select the desired position
     Important: Pin rules take precedence over include/exclude rules.
     If a product is excluded but also pinned, the pin rule wins—and the product is displayed.

Localized values

Localized values are especially useful for international sites. If a product attribute is localized (for example, product price), you can set a rule to automatically apply the localized value for each user's locale. For example, only show products that are in stock in the locale.

Real-time values

You can pass a dynamic filter value, such as a price threshold, directly from the search request. The rule adjusts instantly for each user. For example, if a user searches for "shoes," the rule can automatically filter results to show only shoes that the user can purchase with their loyalty points.

To enable this:

1. Select Real-time value, then select an evaluator from the dropdown.
2. Enter the key name to use to pass the value in the request.

The JSON tab

You can pass custom JSON values in the Search API response. This is a powerful tool for advanced users to trigger different behaviors or interface changes based on the user’s search query.

Here are a few common use cases:

• Redirect: Automatically send a user to a specific category page for certain queries. 
  Example: When someone searches for “dress,” redirect them to the “Dresses” category page.
• Content: Show a specific banner or message in a dedicated slot for certain queries.
• Look and Feel: Dynamically change the layout of the search results (say, grid vs. List), or run A/B tests on visual elements.

To enable custom JSON:

1. Click the JSON tab in the variation panel
2. Define your custom JSON schema.
3. Generate variables for easier reference and dynamic content (optional).
4. This JSON payload will be included in every search request that matches the experience.
5. You can localize JSON variables by clicking the flag icons—just like with templates in Experience Web.

The Playground

Use the Playground to test and preview how search results will look across different locales, while also seeing your merchandising rules in action. It’s a great way to validate how your configurations affect the user experience.