A search experience can take many forms, from site search to an intranet search or a store locator. However, every Yext search experience is built on three essential components that work together to deliver results. At its core, an AI-powered search experience relies on natural language understanding and a sophisticated multi-algorithm approach to fulfill user intent, no matter how a query is phrased.
The Three Components of Search
A Yext Search experience is made up of these three parts:
- Knowledge Graph: The central index that stores your brand's data. You control this data by adding and updating entity types and fields.
- Search Backend: The backend contains the configuration and algorithms that determine which results are selected for a given query. You set the business logic to configure how the search engine behaves — for example, which fields are searchable, synonyms to broaden queries, and result boosts to highlight specific entities.
- Search Frontend: The frontend determines how the results are displayed to the user. This includes the page structure, the design of each results card, and interactive features like facets and sorting.
The Search Quality Loop
The quality of your search results is affected by the Search configuration, the Search algorithms, and the Knowledge Graph. Consider a user searching for "financial advisor who speaks Mandarin":
- Knowledge Graph: The Knowledge Graph must contain financial advisors, a field for languages, and an advisor with "Mandarin" specified in that field.
- Search Configuration: The Search configuration must have a vertical defined for financial advisors and be configured to search the languages field.
- Search Algorithm: The Search algorithms must use the business rules from the Search configuration to identify and return the correct results from the Knowledge Graph.
While you control the data in the Knowledge Graph and the rules in the Search configuration, Yext manages and continuously optimizes the Search algorithm.
Frontend vs. Backend
It's important to understand the fundamental difference between what's searchable and what's displayed. For example, you could make a field like "business hours" searchable in the backend to influence search relevance, but not display the hours on the results card in the frontend. Conversely, you might display a photo of a product on its results card without making the photo's URL searchable in the backend. This separation of logic gives you the flexibility to design a search experience that is both effective and visually appealing without compromising performance.
The Search Query Journey
Now that you know about the different components of Search, here's what happens under the hood in the milliseconds from after a user runs a search to when they get an answer.
Vertical search searches across a single vertical, while universal search searches across multiple verticals. The query journey looks almost identical on both types — universal search simply aggregates results across verticals at the end.
Step 1: User Submits a Search Query
The moment a user clicks into the search bar, Yext Search is hard at work. The search bar guides users to suggested queries using both hardcoded prompts and popular queries. You'll learn more about both in the Query Suggestions unit.
Step 2: Algorithm Routes Queries via Federated Search
Once a user submits a query, the algorithm uses federated search to run the query in each vertical. Learn more about federated search in the federated search reference doc.
Step 3: Algorithm Evaluates Results in Each Vertical
Based on the settings in the Search configuration and the relevant content, the algorithm produces a set of results — if they exist — for each vertical. For verticals using content from the Yext platform, Search has a predefined index of entities and attributes based on the rules provided in the Search configuration. For third-party verticals, Search calls the respective search endpoints using any business logic or access information specified in the Search configuration.
The algorithm not only evaluates which entities should be returned, but also the order of the results within each vertical.
Step 4: Algorithm Aggregates Results and Sends Them via the Search API
The algorithm puts the verticals together in a cohesive order using any vertical ranking set in the Search configuration, then sends all the data via the Search API.
Step 5: Frontend Assembles Results for the User
The frontend uses the result set from the API and the results page and result card layouts to build the user-facing result.
