B2B Commerce Search and Auto-Correction Management

Which Salesforce Commerce Product Do I Have?

Commerce Search matches prefixes against standard fields to return results. Prefix matching is enabled by default for all stores starting in Summer’25, and it can be disabled by contacting Salesforce support. Typo correction is enabled by default for Commerce Search, but it’s executed as a secondary search function. The secondary search updates the search phrase to include candidate corrections and then repeats the search. You can’t disable typo tolerance.

When a customer initiates a term search:

• The initial filter and search phrase execution occur simultaneously, and then Commerce Search determines if the term exists in the catalog.
• If Commerce Search finds the original search term, it returns the results. The results are based on the match between the prefix entered and an existing prefix in the standard fields. Prefix matching is applied only for terms that are three to seven characters long.
• Typo tolerance isn’t applied when a user searches for a SKU.
• If no results are found for a search term, Commerce Search internally initiates a typo correction search and finds the closest match to the original term in the catalog.
• Using the corrected term, Commerce Search runs the search again and returns any results.
• To improve search efficiency, the language analyzer filters out certain common words, known as stop words. For example, in the search phrase “the best pizza in New York,” the language analyzer filters out the stop words “the” and “in,” allowing the search engine to focus on “best,” “pizza,” and “New York.” Stop words vary from language to language.

Keep these considerations in mind for Commerce Search, typo tolerance, auto-correction, and Einstein Search suggestions.

• Stemming is supported. Commerce Search delivers results for any matches to a word’s stem, for example, coffees finds coffee and running returns run.
• Commerce Search provides typo tolerance and auto-correction only when it finds terms that differ by one or two letters or numbers from the original term. For example, searching for coff results in the corrected term coffee. But searching for cof provides no results because it takes more than two letter corrections to reach coffee.
• Typo tolerance looks for the closest match within two corrections of the original term that returns results.
• Typo tolerance applies to tokens with three or more characters. For example, if a customer query for XY-1234 doesn’t return an exact match, typo tolerance searches for the closest match to the second token, 1234, but not the first, XY.
• Typo tolerance looks for replacement terms before filters or entitlements are applied. Typo tolerance thus doesn’t guarantee a specific replacement for a misspelled or not found term, even when candidate terms are within the two correction distance.
• Search uses the exact one or two correction distance from the original term only when looking for corrections in non-localized terms.
• When searching localized terms, the original term is first analyzed using regular search functions. Depending on the localization language, analysis can include stemming searched terms and potential corrections, resulting in terms of shorter length.
• If a localized, misspelled term can’t be properly stemmed, the distance between the term and the stem of its corrected spelling can be greater than two corrections. This scenario can result in some longer misspelled words not getting corrected.
• Einstein Search suggestion uses different algorithms from Commerce Search typo tolerance or auto-correction. Einstein Search suggests type-ahead options only after the site builds up enough search history to use as a reference input.