- Site Search features that impact shoppers’ engagement
Next-gen customers have access to multiple shopping channels and low tolerance to latencies. They expect retail sites to provide best-in-class experiences along with complete and relevant search result for the products they are looking for. Retail giants like Amazon and Walmart are giving top notch experiences in site search and customers expect the same or better experiences from everyone else.
Based on the complexity of implementation and impact on user engagement, search features can be divided into 3 layers.
- Standard Features
- Intelligent Features
- Advanced Features
Standard Features
These are the features that help users with faster search experiences – these are like hygiene factors that today’s users expect from all the e-commerce search engines.
- Auto Correct Many a times while typing a search query, users make spelling mistakes. System should be able to automatically correct these and other similar mistakes like hyphenation, punctuation etc. and use the corrected search query for further processing and rendering search results
- Auto Spacing If users tend to omit spaces between words, the engine should automatically correct them and give results for the corrected query. This makes the entire searching process more intuitive for the user. E.g. When user types “samsungoledtv”, It should automatically be converted to “Samsung oled tv” for further search query processing
- Keyword Redirection Keyword redirection is used to redirect user’s search queries to specific webpages/URLs. For optimized user experience, most of the websites have specially designed landing pages for various categories, assortments & brands. Keyword redirect lets business user associate various keywords like names of categories and brands to specially designed landing pages. When users search for keywords that have such associations or URL redirection assigned to them, search engines should take users to the assigned URL for better experience
- Recent Search Recent search gives the list of all search queries that the user has recently searched for. As user puts her cursor on the search bar, site should give a list of recent searches by the user. This enables the user to resume from the last search session without having to enter same search queries again
- Type-ahead Typeahead or autocomplete predictions are the list of search queries that user could potentially type. The list is predicted based on the characters entered by user, and it keeps getting refreshed with every new letter that user enters. This helps users select any of the predicted queries to get the search results
- Breadcrumb Negation The idea of giving removable breadcrumbs with search results is that users can negate one of the search keywords but still can keep the overall search query
Intelligent Features
These are features driven by AI/ML models and understand users intent and context to give relevant results. Some examples are:
- Facet Recognition For many search queries, users does not get relevant results because search engine is unable to map keywords to the correct attributes/dimensions. Search engine can give more relevant results and can rank products more effectively if it can map keywords to the correct facets/dimension
- Category type identification System is unable to identify exact category/product type when user enters queries with multiple details or uses natural language. It is important for a search engine to identify context of the search query and recognize the intended category type to give relevant results to the user. We can use category type identification to determine context of the search and improve relevance and ranking of the products
- Intent Identification Users tend to use connecting words, prepositions, long term semantic terms in the search query. Search engine should identify user’s intention and interpret search query correctly to give correct and more relevant results
- Examples of prepositions in determining facets and range: “Product below(under) 50” means price range should be < US$50
- Examples of Connecting words are “Camcorder with microSD card” means Category Type is “Camcorder” and Filter for MicroSD card
- Relevance Ranking For search queries, Search engines should be able to render correct and complete result set with most relevant products coming on first page. Engines should be able to rank products in such a way that the user can find most relevant products in an easy manner. Engine should be able to proper query pre-processing, identify intent & context to understand users
- Intelligent Type-ahead Typeahead or autocomplete predictions should be based on search queries that user could type. The list is predicted based on the characters entered by user, and it keeps getting refreshed with every new letter that user enters. User can select any of the predicted queries to get the search results. Instead of using just catalog entries like category and brand names to generate type-ahead list, this feature could take the list of actual user search queries also into account to generate more relevant type-ahead predictions.
Advanced Features
Advanced features create engine’s own understanding of product and search queries driven by user behaviour – these features read between the lines to capture user’s interest and give engaging products even though user might not have mentioned it explicitly through the search query
- Auto Tagging Content creators are usually responsible for all the content that appears in product detail page. The language and terms that content creators use might not necessarily be the same that actual users use. If content creators use “washing machine” and if end users search with the term “washer”, with normal keyword search the catalog might not be able to return any results because that exact term might not be available in the catalog
Any advanced search engine, which strives to understand the meaning of user’s search query, it is very important to learn that washer is same as washing machine based on user behaviour and store this learning about new association between “washer” and “washing machine” for future use. So that next time whenever user searches for washer, the engine knows that she’s searching for products tagged as washing machine in the catalog. - Behaviour Based Ranking Search engine can also make associations based on user behaviour to improvise its understanding of user search queries and products in the catalog. Nowadays normal keyword search is unable to give end-user the experience that they expect. Search engine must understand what the user might be trying to find and give nearest matches if exact match is not available.
If for example user is searching for “mauve dress”, search engine can make interpretations like, is she searching for a dress in color mauve? Does the portfolio have that exact color with that exact name? if not, are there any products which might be tagged as purple or pink which can qualify as color mauve?
If the search engine can make such connections and interpretations and use that to render proximal products, just imagine the level of satisfaction that user will get. With such advanced interpretations, users will never return empty handed when catalog has relevant products - Domain Dictionary Based on user behaviour specific to the domain of the site, search engines can create domain specific dictionaries. “Apple” in an eCommerce store in electronics domain would mean gadgets of Apple brand, while same in an eCommerce store dealing in groceries and fresh farm supplies would mean Apple fruit. A312-ds-2323 might not mean anything in fashion domain but that would mean a laptop’s model code in electronic domain. Search engines should observe user interactions and behaviour in a session and make connections through which a domain specific dictionary can be created, which can be user to understand user’s need with a deeper level.
Conclusion
Search is fast emerging as the centre of all eCommerce stores. More and more people are using search to find relevant products instead of relying on traditional menus and navigations. With deeper understanding of user’s intent and interpretations derived through user behaviour using advanced AI algorithms, search engines are making bigger strides than ever. With domain specific use cases, eCommerce companies will be able to give seamless, satisfying and engaging user experiences
How can AIE Help?
AIE can strong experience in building custom search engine solutions that are robust and bring relevant, lightning-fast search capabilities to your application. If you are interested in deploying a lightning-fast search solution for your customers, then contact us for a free consultation and quote