The Query Analyzer interprets what your customers are searching for in real time. Rather than looking for products that contain the same words as a search phrase, it breaks the query down into structured concepts and matches those against the Knowledge Graph to find products that genuinely fit the request.
How it works
Breaking down the search query
The Query Analyzer runs in real time, processing each search as it's entered. When a customer searches for something like "black Sony jacket for winter," it doesn't treat this as a block of text to match against. Instead, it identifies each part of the phrase as a separate concept: a brand (Sony), a color (black), a product type (jacket), and a contextual attribute (winter/occasion).
Each concept is weighted by importance. Product type and brand carry the most weight. If Elevate knows what kind of product a customer is looking for, it can return results with confidence. Secondary attributes like color or occasion are important but treated as less critical if a perfect match isn't possible. The identified concepts are then matched against the Knowledge Graph, checking whether products logically fit the search rather than just whether the same words appear in a product description.
Search relaxation when there's no perfect match
If no products satisfy all parts of a query, for example if you have Sony jackets in black but not in the specific shade the customer described, the Query Analyzer applies search relaxation. It removes the least important part of the query first, keeps the more critical parts, and returns these near-matches in a secondary list, clearly presented as not being exact results. This keeps search useful rather than returning nothing.
Multilingual and mixed-language searches
If a customer combines two languages in one query, the Query Analyzer attempts to interpret both parts using available dictionaries. Results will still be returned, though the precision for the non-primary language may be slightly lower.
Autocomplete and built-in synonyms
The same logic that powers search results also drives the suggestions that appear as customers type. The Query Analyzer considers not just what a customer is likely searching for, but which completions will lead to meaningful product results, steering them toward searches that will actually find what they're after.
Elevate also maintains dictionaries of synonyms across all supported languages as part of the Knowledge Graph. If a customer searches for "sneakers" but your catalog uses "trainers," Elevate handles that connection automatically without the need to configure it.
How to influence search results
Start with your product data
The Query Analyzer can only match against what's in your catalog, and that catalog is only as well-structured as the data you send. Accurate, consistent product data is the most impactful thing you can invest in.
Use keywords for campaigns and time-limited concepts
For searches that have nothing to do with a product's inherent attributes , such as Black Friday or a season sale, you can add keywords to a set of products directly in Elevate. Keywords attach to the products themselves, not just to a page. This means that if a customer searches "Black Friday jackets," they'll see only the jackets from your Black Friday selection, not every product in it. Keywords are exact matches, so if you want to cover "Black Week," "Black Friday," and "Cyber Monday" as separate search terms, each needs its own keyword entry.
Use synonyms to fix specific search gaps
If a particular search term is returning no results but clearly should be, you can add a synonym as a temporary fix. This is most useful for things that fall outside Elevate's built-in dictionaries, such as a brand abbreviation your customers commonly use. When you do add a synonym, it's worth also reporting it to support so that Elevate's team can assess whether it should be handled at the system level instead.
Be careful about overusing keywords and synonyms. Both features bypass Elevate's normal relevance calculation, which means they're powerful but blunt. If you use them for concepts Elevate should already understand, like standard product types, common colors or known brands, you may end up surfacing products in ways that don't serve your customers well. If something isn't working as expected, contact support rather than working around it with a synonym or keyword.
Adjust visibility with Boost and Bury
Boost and Bury settings influence which products surface near the top of search results and which are pushed lower. They work alongside Elevate's relevance calculation rather than overriding it, so a boosted product still needs to be relevant to a search to rank well.
Keep stock, pricing, and newness data accurate
Stock levels, pricing, and newness settings all feed into how products rank within search results. Products that go out of stock are automatically pushed down. Products with a reduced selling price relative to the list price receive a temporary boost. You can configure how aggressively and for how long newly added products are promoted. This is useful for getting new arrivals visible before they've built up any purchase history.
Article last updated