Menu
Feedback
Start here
Tutorials & Solutions
Developers Portal

Known Issues
Support Rules

Frequently Asked Questions
Announcements
Start here
VTEX IO
VTEX Intelligent SearchHow does VTEX Intelligent Search work?
How does VTEX Intelligent Search work?
8 min read

The Search Engine is the core of VTEX Intelligent Search's intelligence. It interprets the search term and displays a filtered and ordered list of products. This tool allows a more objective search through indexing and treatment of the catalog data.

This feature is included by default in VTEX Intelligent Search and covers the vast majority of scenarios. However, you can also adjust the Search Engine to improve search results. This guide covers the following:

Standard search behavior

The following are characteristics of Intelligent Search's standard behavior.

Autocorrect

Users can make spelling mistakes when searching. The tool interprets these misspellings and shows similar products.

The search engine performs an automatic correction for each search based on the items added to the catalog. For each search term, a specific number of wrong characters is allowed. The number of accepted misspellings is called fuzzy.

The fuzzy varies depending on the length of the search term. By default, if the term is 3 to 5 characters long, the fuzzy applied will be 1, which means that 1 misspelled character is allowed. For terms of 6 characters or more, the fuzzy applied will be 2.

The predicted misspellings considering fuzzy = 1 are:

  • Add 1 more character.
  • Remove 1 character.
  • Change 1 character.
  • Swap 2 characters that are next to each other.

Example: Suppose a customer wants to search for the term ball. Using the default configuration (fuzzy = 1 for terms of 3 to 5 characters), the search engine will return results for ball in any search that contains accepted errors, such as:

  • bawll (add 1 more character)
  • bal (remove 1 character)
  • balk (change 1 character)
  • abll (swap 2 characters next to each other)

Blank spaces are not considered in fuzzy, so base ball will not be corrected to baseball. We recommend using synonyms for such cases.

Minimum result

The minimum number of results displayed for any search is 1. If a search does not meet this minimum, autocorrect is applied.

In a search with multiple terms, the following logic occurs:

  1. If the minimum result is not reached even after the initial autocorrect, the search starts to include more comprehensive results containing any of the search terms. For example, if a user searches for "nike sneakers" and the catalog doesn't have any products matching that name, the search will return products with only the term "sneakers" instead.
  2. If the comprehensive search doesn't return the minimum result, the process will be repeated, applying autocorrect again to each term.

Searches with multiple terms are more efficient when customers sort the results by relevance, as such sorting can return more results than the initial search context. When sorted by relevance, items with multiple terms are prioritized. However, with other sorting criteria, such as price or bestsellers, the result ranking is based on these numerical values and does not necessarily prioritize items that match more search terms.

Pagination limit

VTEX Intelligent Search limits the search results to 50 pages, which encourages the customer to use filters, creating a more relevant and focused experience.

Search behavior configuration

There is a set of settings that change the products that are displayed in search results and their order. In the following sections, you will see the main definitions that affect search behavior.

Defining searchable fields

Determines which product information is searchable. From this configuration, every search in the search bar checks the content of these fields to generate search results. Check the following table to understand which information is searchable by default and the other configuration options:

InformationConfiguration
Product name, brand, product ID (ProductID), product reference code (ProductRefID), SKU ID (SKUID), SKU reference code (SKURefID), and EANThis information is searchable by default in VTEX Intelligent Search. To remove any of this information, please contact our Support.
Product and SKU specificationsTo include the desired specifications as searchable fields, you must configure the Searchable specifications option in the VTEX Admin, under Intelligent Search > Search Settings.

For example, if a shirt doesn't have the color in the product name or SKU, by default, Intelligent Search won't identify this attribute in a search for "blue shirt", displaying shirts of different colors as results. However, if the color specification is configured as searchable, the search will likely return blue shirts at the top of search results.

For more details about defining searchable specifications, check out the Search configuration guide.


Learn more about specifications in Product and SKU specifications.
Manufacturer code (ManufactureID)Please contact our Support to set this field as searchable.

For example: The "color" specification is set as searchable and the product was added with the name "Nike Total 90 Sneakers" and the color "black". If a customer searches for "Black Sneakers", the tool will return products that have the term "sneakers" and the color "black" at the top of search results.

You can include special characters ( - and /) when searching for a product.

Choosing the search fields affects the merchandising rule options, as only searchable fields can be used to set up the rules. For example, if you set only the SKUID field as searchable, you cannot promote a** ProductID** or a ProductRefID using merchandising rules.

Defining and sorting filter fields

In all search results, users can select filters. Some filters are pre-defined by default, but it is also possible to create more filters to meet store needs. The display order is also customizable. Example: In a search results list, there can be filters for Price, Categories, and Brand.

Configuring relevance

Relevance is the ordering principle of search results. In this section, you can assign weights and prioritize certain criteria in the Search Engine, such as popularity, release date, and promotions. The relevance of a product is composed by the combination of weights and priorities defined by each merchant.

For example: a merchant configured that "promotion" has greater relevance than other criteria. When a customer searches for "t-shirt", the tool might display the t-shirts on sale at the top of search results, depending on the weights assigned to other criteria.

For more information on how to customize the order of results to meet your store's needs, read the article Relevance settings.

When using Intelligent Search, two options define the sorting order of your collections. By using the map=productClusterIds command, you are setting the order of your collections to be the one pre-defined by the merchant, with its relevance criteria. If you choose to use productClusterNames, you are setting that you want your collections to follow Intelligent Search's own relevance standards.

It is possible to search by the ID of one or more items in the search bar or by adding search parameters (query) in the store URL.

Partial match

By entering the exact first digits of the product ID (ProductID), product reference code (ProductRefID), SKU ID (SKUID), SKU reference code (SKURefID), or EAN in the search bar, Intelligent Search will partially match products and SKUs active in the store, adding them to the search results.

For example, if the product ID is 123456789 and you search for 123, it will be listed in the search results. However, the product will not be displayed if the search is 234 and does not include any of the first ID numbers.

URL

To search from your store URL, add search parameters (query) at the end of the store URL, respecting one of the following structures:

  • Example for searching an item: [account address]/[id type]:[id_1]?q=[id type]&map=ft
  • Example for searching a list: [account address]/[id type]:[id_1];[id_2];[id_3]?q=[id type]:[id_1];[id_2];[id_3]&map=ft

The possible value types for the ID type field are product.id, sku.id,sku.ean, sku.reference, or id (ProductID, ProductRefID, SKUID, SKURefID, and EAN). All searched IDs should be of the same type.

  • Search by SKU ID: ?q=sku.id:<id> or ?q=sku:<id>
  • Search by EAN: ?q=sku.ean:<id>
  • Search by reference ID: ?q=sku.reference:<id>
  • Search by slug: ?q=product.link:<link>
  • Search by product ID: ?q=product:<id> or ?q=product.id:<id>

Indexing

Indexing is the syncing process between store and search catalogs. This feature allows you to keep your content up to date with continuous updates of the product index. It also performs data processing to be used in the search.

Initially, it syncs all the products from your VTEX catalog. After that, data changes — such as product name and status — are updated in real time by the search indexer.

In this sync process, the system queues up the information that has been updated to add it to the search catalog. Once all changes have been indexed, they will be available on the website.

The indexing process is safe and has a retry mechanism — if an item is not indexed on the first attempt, the system makes subsequent attempts until it is successfully completed.

Indexing is fundamental to analyze, process, and organize data as best as possible to improve your search results.

Check Indexing History for more information on how to monitor the indexing process.

Contributors
1
Photo of the contributor
+ 1 contributors
Was this helpful?
Yes
No
Suggest edits (Github)
Intelligent Search
« Previous
​​Availability
Next »
Contributors
1
Photo of the contributor
+ 1 contributors
On this page
Still got questions?
Request support from VTEX
For personalized assistance, contact our experts.
Open a support ticket
Ask the community
Find solutions and share ideas in the VTEX community.
Join our community
GithubDevelopers PortalCommunityFeedbackSite Map