Most website visitors don't just type in one or two keywords when they use a search engine – they use multiple, sometimes lengthy, queries. This is especially true for ecommerce sites, where shoppers might search for something specific like "men's size 10 running shoes" or "women's velvet green dress." It can be difficult for a traditional search engine to handle these longer, more specific ones that the engine is optimized to understand.
Keywords and how they relate to search
Not all search queries are the same. With the advances in search technology, consumers have grown used to platforms that can understand their queries even when they're presented in a natural form, in mixed languages, and even with typos and misspellings.
Most searches fall into one of three categories: fat head, chunky middle, or long tail.
Fat head searches are the most common and are usually one or two-word queries about a broad topic. Chunky middle searches are slightly more specific, often three to four-word queries about a particular aspect of a general topic. Long tail searches are the least common but can be the most specific, consisting of five or more words and often including particular product or brand names.
Interestingly, long-tail queries comprise a significant chunk of buyer search queries (over 56% of buyers input queries with three or more words).
Search engines like Google use artificial intelligence to try to understand the intent behind each search and match it with the most relevant results. It is no wonder, then, that half of Google's revenue comes from long-tail advertisers.
Types of search queries
Let's take a look at some of the major types of queries search engines work with every day -
Broad searches: Broad searches are query types encompassing a wide range of search terms. They usually return a large number of results, making them great for exploring a new topic or finding general information. For example, someone looking for information on cars might do a broad search for "cars." This would return results on a variety of topics related to cars, such as "car insurance," "car dealerships," and "car reviews."
Exact searches: Exact searches are when a customer enters a very specific query, typically a brand name or item number. For example, if a customer enters the query "Nike women's running shoes size 7," they are looking for a specific item. This customer is much further along in the purchasing journey than someone who simply enters "running shoes."
Compatibility searches: Compatibility searches are query types that allow you to find products or services that are compatible with your existing ones. This can be helpful when upgrading or expanding your current setup.
Feature-based searches: Feature-based searches are query types that let you find products or services based on the features you're looking for. This is helpful if you have specific requirements or preferences and want to narrow your options.
Concept searches: Concept searches are query types that let you explore a topic by finding related concepts and ideas. This approach can be helpful when you're trying to learn about something new or better understand an existing topic.
Symptom-based searches: Symptom-based searches are query types that focus on finding solutions to problems based on the symptoms you're experiencing. This is helpful when the searcher is troubleshooting an issue or trying to solve a specific problem.
A good question is whether the nature of search query matters for ecommerce businesses trying to offer a smooth shopping experience. The short answer is – it does.
Long-tail keywords - the nemesis of classic keyword search
Traditional keyword search engines do a decent job at fast searches when it comes to single-word queries or short phrases (like "blue pen"). However, they often fall short when it comes to long-tail search queries.
For a business, a long tail query such as "women's black leather high heels in size 10" would typically mean that the user has a rather specific idea of what they want (and a higher intent). Your site search engine should be capable of handling these queries and matching them with the most relevant and accurate search results.
With a string matching-based search, the search engine will surface limited results even if many more products match the user's need. This is because traditional keyword search requires a word-to-word match for the results to surface. It also only accounts for synonyms, and spelling errors, if built explicitly for these functionalities. So, if a user searches for "acne spots" or "pimples" on an online cosmetic store, a keyword-based search would fail to make sense of the query unless those synonyms have been written rules for synonym handling. For long-tail queries, it is next to impossible to add all the plausible synonyms for all the queries.
This means that the user looking for black leather high-heeled boots would see fewer options, with many relevant options remaining unexplored. This is bad news for the ecommerce store.
Vector search - Making sense of long-tail keywords
AI-driven search is equipped to deal with long-tail queries in an intelligent, concept-focused way. AI search offers relevant results for unique, misspelled, or highly specific queries by focusing on the user's entire query – not just individual keywords – to determine the meaning and intent behind it. But how does it make it happen?
The answer is vector search. Vector search is a more sophisticated approach that uses machine learning to understand the meaning of queries by uncovering the relationships between words and associating them with real-world concepts. It's based on the idea that words that are semantically similar (for instance, "acne," "dark spots," "hyper-pigmentation," and "sensitive skin") have similar (or closer) vectors.
Vectors are essentially mathematical representations of text to help search engines make sense of the millions of unique search queries. By analyzing the vectors of words in a query and comparing them to the vectors of products in a catalog, vector search can more accurately match products to queries—even when those queries are long-tail or highly specific.
What makes vector search superior to classic keyword-based search is its lack of dependence on words to deliver relevant results. So an ecommerce store that sells skin care can push products that are effective for acne or sensitive skin over others in the search results.
This is a powerful on-site search approach, especially when dealing with a massive product catalog and several categories. It can handle misspellings and variations of a query (e.g., "jogging pants" and "athletic wear"). Vector search can also account for synonyms and related terms (e.g., "black dress shoes" and "men's loafers")
AI search engines are powering a new generation of search
AI and ML are advancing the search experience in a way that no other search technology has accomplished. AI search uses various techniques to make sense of long-tail queries, such as natural language processing and semantic analysis. This allows businesses to quickly find the products that match a customer's query and show them the most relevant results. This means highly relevant and accurate search results and a seamless user experience.
AI-driven search and discovery solutions help businesses transcend the limitations of building on-site search functionality from scratch. Platforms like Zevi offer advanced search capabilities like natural language search, autocomplete, typo tolerance, synonym handling, search merchandising, and personalized search results. Zevi is built on NLP and, therefore, surpasses string-based search engines in both performance and search relevancy.
To unlock the power of intelligent search for your business, book a free demo with us.