Ecommerce search engines are like the navigators on a ship. Just as a navigator helps the captain chart a course through rough waters, ecommerce search engines help shoppers find their way through the vast and often overwhelming sea of online products. But just as navigators must be able to handle various challenges, from storms to uncharted territory, ecommerce search engines must be able to handle a wide range of queries from shoppers.
In this article, we'll explore some of the different types of queries that ecommerce sites may encounter, and how they can be handled to ensure a seamless user experience.
How do search engines go about handling different types of queries in general?
When a user enters a query into a search engine, the search engine begins the process of finding and ranking relevant results in order to provide the most useful and accurate results to the user.
It all starts with a detailed search index. Your index will determine your user's search experience. The search index is basically a giant library comprising all your products, product information, metadata, synonyms, etc. Whenever a search is made, the engine refers to this index to retrieve information about the product, and uses the included product attributes to rank the results.
Before a query can be processed, it must be "preprocessed" by the search engine. This involves breaking the query down into its component parts, such as individual words or phrases, and removing any unnecessary words (called "stop words") that are commonly used but do not carry much meaning (e.g. "the," "and," "but").
Processing means understanding the query at a semantic level. The search engine must try to understand what type of query it is and its intent in order to provide relevant results. This is best accomplished by search engines that rely on natural language processing (NLP) rather than string-matching.
For example, if a customer types in a query such as “red shirt for important work meeting”, an NLP–based search engine will be able to figure out that it should display red shirts that are formal, rather than half-sleeved red shirts or red shirts with some images or text on them. In contrast, an engine based on string-matching might fail to return too many relevant results for such a query.
Once the initial query has been processed, the search engine must build an optimal query tree to provide relevant results – the search engine uses its index to retrieve a list of relevant products. This process is known as "information retrieval," and it involves matching the words and phrases in the query with the content of the product in the index.
To do this, the engine evaluates different combinations of filters, sorting criteria and other parameters such as location or price range to create a query structure that best matches what the user is looking for. This process requires advanced algorithms which have been carefully tuned over time to provide precise results in minimal time.
In the context of an ecommerce search index, product attributes may include things like the product's name, brand, price, color, size, material, and other features that are relevant to the shopper.
Having a comprehensive set of product attributes in the search index allows ecommerce search engines to accurately match queries with the relevant products and provide a more refined and relevant set of search results to the user.
Once the initial query has been processed, the search engine must build an optimal query tree to provide relevant results. To do so, the algorithm evaluates different combinations of filters, sorting criteria and other parameters such as location or price range to create a query structure that best matches what the user is looking for. This process requires advanced algorithms which have been carefully tuned over time to provide precise results in minimal time.
The search engine then ranks the retrieved products based on their relevance to the query and their overall quality. Relevance is determined by how well the product info matches the words and phrases in the query, while quality is determined by other factors.
The search engine then presents the ranked list of products to the user through search results. These results typically include the title, image and a brief description of each page, as well as a link to the page itself.
Now that we know what happens behind the code when a search query is made, let’s look at the different types of search queries for eCommerce.
Types of search queries
- Exact name – query is an exact match of the product or service name e.g. ‘PS5’
- Product type – searching for a type of product e.g. ‘laptops’
- Symptom – solution oriented queries e.g. ‘leaky faucet’
- Feature specific – queries that mention product attributes e.g. ‘vegan hot sauce’
- Thematic – revolving around a theme e.g. ’Birthday present for 7 year old girl’
- Compatibility – looking for products that work with other products e.g. ‘iPad keyboard’
- Slang – queries using slang terms or abbreviations e.g. ‘forces’ for ‘AF1 shoes’
- Non-Product – queries where the searcher is looking for information instead of an item e.g. ‘next sale’
Of these, we’ll deep dive into the four major query types and how they are handled below.
How search engines handle queries
· By exact name
Exact name searches are straightforward and typically handled well by most search engines. A search for ‘Apple iPhone 13 Pro Max’ will yield the right product. Both string-based and AI-based engines can handle exact name queries with ease.
· By product type
Here’s where it starts getting trickier. Searching for a product type or category, such as "trousers" can be more challenging for ecommerce sites to handle than exact name searches. The search function must be able to understand and interpret the user's intent and return relevant results even without the entire context.
For one thing, if the search term matches an existing category on the site, it can be a good idea for the search engine to redirect to the relevant category page. Secondly, the search engine also needs to take synonyms into account here; for example, the site might have a category page named ‘pants’ but not ‘trousers’, and the search engine has to know that they are equivalent. Lastly, it’s important for a good site search engine to provide filters and facets that are appropriate for the product category.
Modern AI-based search engines use vectors to properly handle alternative category names and synonyms, so that the displayed results are always highly accurate.
· By symptom
Sometimes, your customers don’t even know what they want; they just have a problem and want something that’ll fix it.
Handling queries related to symptoms, such as "heel pain while running," can be particularly challenging for ecommerce sites. This is because these types of queries often require a deep understanding of the user's problem and the available solutions, and the search function must be able to match the problem with the product.
As a result, this is where search engines based on artificial intelligence (AI) and machine learning (ML) truly shine. Such search engines can make use of techniques based on what’s called natural language processing (NLP) to figure out that the most likely reason for typing in such a query is that the searcher is looking for a solution to a problem.
Another way that search engines can guide the user so as to help them get more relevant results is by providing suggestions when they start typing (e.g. suggest ‘heel pain remedy’ when the user types in ‘heel pain’); this feature is known as autosuggest or predictive search.
· Non-product searches
Maybe what your customer is looking for isn’t even a product. Sometimes people want quick answers to their questions, and a user base used to Google expects your search to function similarly.
Users frequently make non-product searches such as ‘electronics return policy’ or ‘next sale’’.
With non-product searches, customers typically want a concise and single answer. So search engines have to be configured to recognize certain common searches like the ones above. Even if the query isn’t an exact match, engines with AI and ML capabilities can understand the context and will display non-product information before the results page.
Overall, ecommerce search engines play a crucial role in the success of an ecommerce site by helping shoppers find what they are looking for and facilitating a seamless and enjoyable shopping experience. By being able to handle a wide range of queries, ecommerce search engines can help turn casual shoppers into loyal customers.
With Zevi, ecommerce search engines are able to provide relevant and accurate results for every type of query — from simple keyword searches to complex structured queries — ensuring customers have an effortless shopping experience while helping retailers maximize their sales. Try out our Shopify App, or book a free demo today!