You are here:
AI-Powered Search for Commerce
Improve store search experiences with Einstein semantic search in D2C store. Einstein semantic search reduces the occurrence of “no results found” pages, lowers bounce rates by providing a smoother shopping experience, and improves conversion rates. Einstein semantic search uses natural language processing to interpret queries and matches the information in the Product Name, Product Description, and Category Name fields with the terms entered in the query.
Required Editions
| View supported editions. |
Which Salesforce Commerce Product Do I Have?
Einstein semantic search populates the search results page by matching and returning terms that are semantically related to a query. When you query for product bundles, search results only match queries for the bundle’s parent product. Einstein semantic search uses exact queries for product SKUs.
With Einstein semantic search, your store is better equipped to handle a range of customer queries, although improvements vary based on your store’s catalogs and customer behavior. For example, gains are especially apparent if your customers more frequently search for broad terms like “furniture” or “couch” rather than exact product SKUs.
The table describes how query results differ based on whether Einstein semantic search is enabled.
| Query Entry | Example | when Einstein Semantic SEarch is Disabled | when Einstein Semantic SEarch is Enabled |
|---|---|---|---|
| Synonyms | Couch and sofa | Search results don’t include synonyms of the term queried. | Search results include synonyms of the term queried. |
| Alternative spellings | Adapter and adaptor | If the spelling of a term that the user enters doesn’t exactly match the spelling in the product catalog, search results could be incomplete. For example, the system might not recognize that “adapter” and “adaptor” refer to the same product. | If the spelling of the term that the user enters is slightly different from the spelling in the product catalog, search results could include the variations. For example, if the user searches for “adapter” and the catalog lists “adaptor,” the search results include the catalog term. However, if the spelling difference is significant, relevant items are overlooked. |
| Abbreviations | AC and Air Conditioner | If the abbreviation that the user enters isn’t in the product catalog, relevant items are overlooked, and the search results could be incomplete. | If the abbreviation that the user enters isn’t in the product catalog, the search results could include relevant items. However, the quality of the search results depends on how common if the abbreviations are. |
| Slang | Men’s shades | If the slang term that the user enters isn’t in the product catalog, relevant items are overlooked, and the search results could be incomplete. | If the slang term that the user enters isn’t in the product catalog, the search results could include relevant items, but the completeness depends on how common the slang term is. |
| Misspellings | Shert and shirt | Search results are accurate if the user enters words with minor spelling errors that are within the auto-correction parameters. See Commerce Search and Auto-Correction Management. | Search results are accurate if the user enters words with minor spelling errors. However, the quality of the search results depends on how common the misspellings are. |
| Product themes | Wedding gift | If the user enters a product theme, search results could be incomplete. For example, the system can find items containing the exact words “wedding” and “gift,” but it can miss suitable gifts that aren't explicitly described this way. | If the user enters a product theme, search results are likely complete. For example, the system can generally understand the concept of “wedding gift” and return relevant items even if they aren’t explicitly labeled. However, the quality of the search results depends on how common the product theme is. |
| Product features | Long USB cable | If the user searches for product features using specific terms like “long,” “USB,” and “cable,” the system can miss items that are described with synonyms such as “cord” or specific lengths like “3m,” leading to incomplete search results. | If the user searches for product features using specific terms like “long,” “USB,” and “cable,” the system generally understands the desired features of the product, such as length and type. It can find cables described as “long” using specific measurements like “10ft” or “3m” or synonyms like “cord.” However, the quality of the search results depends on how common the product features are. |
| Product types | Women’s jackets | If the user searches for a product type, the search results include items that explicitly match the term entered in the query, but it can miss related items. For example, if the user enters “women’s jackets,” the search results show exact matches, but it misses related items like “coats” and “blazers” and related outerwear like “ladies’ jackets.” | If the user searches for a product type, the search results include items that are relevant to their query. For example, if the user enters “women’s jackets,” the results include exact matches as well as related outerwear, like “coats” and “blazers.” |
| Multiple languages | Wireless headphones for running (English) Auriculares inalámbricos para correr (Spanish) |
If the user enters a non-English term in an English-language store, no search results are returned. | If the user enters a non-English term in an English-language store, or uses non-Latin characters, the system matches the term with its English equivalent and the search result is complete. This is also true if a user enters non-Latin characters, such as those used in Chinese. |
| Brand names | Adidas shoes and Plorra dress | Search results are returned if the brand name entered is in the product catalog. | Search results are likely complete for well-known brands and incomplete for less common or newer brands. |
| Ambiguity | Apple charger | If the user enters an ambiguous query, no search results are returned. | If the user enters an ambiguous query, the search results show the most reasonable matches. |
| Related products | 20" TV and 21" TV | Search results don’t reflect the similarity or relationship between products. For example, if the user enters a query for a 20" TV, a 20" TV appears first as an exact match. However, other TV sizes can be returned in a random order, or a 20" monitor could be listed. | Search results reflect the similarity between products and the difference in size. For example, if a user searches for a 20" TV, a 20" TV shows first as the closest match. TVs with similar sizes, like 19" or 21", appear higher in the results, while a much larger size, like 55", appears lower. Monitors, even if they are 20" are less likely to appear because the system distinguishes between product types, such as TV and monitor. |
| Exact keywords | iPhone 14 Pro Max | Search results return the exact keyword. | Search results return the exact keyword. |
| Descriptions | Table with blue-greenish color | Search results return items that match the exact words that the user enter in the description. For example, if the user searches for “Table with blue greenish color,” the results include blue and green tables but not teal. | Search results return the most reasonable items. For example, if the user searches for “Table with blue greenish color,” the results include teal tables as well as blue and green tables. |
Note When you use Einstein semantic search, sort by Best Match to return the expected results.
See Add a Sort Rule.
Did this article solve your issue?
Let us know so we can improve!
