Vexture Search in Miva Merchant
Smarter, More Relevant, and Context-Aware Search for Your Customers
The Vexture search feature is a subscription based upgrade that enhances online store searching and product recommendations by using the same kind of artificial intelligence powering today’s most advanced language models to understand meaning, not just words. This means your customers can search naturally, just like they think, and still find exactly what they need.
Vexture offers a powerful new way to search by moving beyond traditional keyword matching searches to understand the meaning behind customer queries and product descriptions. This allows for a more intuitive, effective shopping experience for your customers.
Key Benefits
- Native Integration: No fragile third-party dependencies.
- AI Understanding: Handles misspellings, intent, and broad/niche queries without manual synonym lists.
- Fast: No external API calls — runs entirely on your Miva infrastructure.
- Flexible Testing: Branch-level preview before going live.
- Scales for Large Catalogs: Especially powerful for automotive, industrial, and fitment-based catalogs.
How Traditional Database Searches Work
Until now, Miva’s product search relied on traditional based keyword matching.
When a customer typed a term, the system looked for exact matches in product names, SKUs, or descriptions.
Example: If they typed “brake pads,” it looked for the exact phrase “brake pads.”
- Typos, synonyms, or related terms often meant no match or irrelevant results.
- The system couldn’t “understand” relationships between terms “hatchback” and “Civic SI”.
This worked well for precise searches for things like part numbers, but it struggled when shoppers used conversational, semantic, or less specific searches.
How Vexture, Semantic Search Works
Vexture Search uses semantic search powered by vector embeddings to find products by meaning, not just matching letters.
When you enter a search term:
- The words are converted into a vector array, an array of 768 numbers representing the concept in a multi-dimensional “semantic space.”
- All products in your store are stored in the same space.
- The AI measures the “distance” between a search’s meaning and each product’s meaning.
- The closest matches are returned even if they don’t share the same exact words.
Example:
- Search for “dog food” → also finds “golden retriever chow” or “puppy kibble.”
- Search for “EK hatch parts” → understands that’s a ’90s Honda Civic SI and returns relevant Civic parts.
This is possible because the AI is trained on billions of real-world text examples from online conversations like Reddit threads or encyclopedias, so it understands context and relationships between terms.
Traditional Search
- Matches exact keywords, substrings, or equality checks in a database.
- Fails when the customer uses synonyms, slang, typo, or different phrasing.
- Works best for precise lookups like SKU or part numbers.
Vexture Search
- Matches based on meaning, not just letters or numbers.
- Can understand relationships (e.g., “EK hatch” → ’90s Honda Civic SI parts).
- Supports typos, variations, and vague descriptions.
- Uses heuristics to decide how many results to show quickly with Specific, Broad, or Sparse Results.
- Can blend results: exact matches for precise searches plus AI semantic matches for broader searches.
Blending the Best of Both Worlds
Some searches like exact SKUs, part numbers, or unique product codes are still best handled with traditional keyword matching. Vexture processes natural language searches through the AI index, while also using key product fields for traditional keyword matching.
The Vexture Search blends these approaches:
- Exact matches from traditional search are treated as a 100% match.
- AI-powered semantic matches fill in the rest, ensuring relevant suggestions even when there’s no exact match.
- Merchants can tune blending rules to prioritize traditional results, AI results, or a mix of both.
This “best of both worlds” approach means the new release doesn’t replace traditional search, it enhances it with AI capabilities for cases where a traditional search alone would fail.
Blending Vexture and Traditional Search occurs when certain triggers are met:
- If exact matches are found in configured fields (such as code, SKU, or name), they are prioritized over AI-powered semantic results.
- If no exact matches are found, results rely entirely on AI semantic search.
- Merchants can tune (adjust) cutoff values and blending rules, so behavior may vary by store.
Exclude Products from Search
Products with Exclude from Search Results enabled are removed from:
- AI-powered (Vexture) search results
- Traditional keyword search results
- Blended search results
These products are also removed from the AI search index during indexing.
Note: Changes to this setting may not appear immediately, as search results are updated during indexing and cache refresh cycles.