Director, Thought Leadership Marketing
Sorry, there is no results for this query
You type in a query while doing a Google search (or use another popular search engine like Bing), or one of your favorite ecommerce sites such as Amazon. You expect to get your information retrieval in the form of a search engine results page (SERP) that gives you exactly the right results, with the search engine taking an “educated guess” about the best item you’re imagining.
Unfortunately, many ecommerce stores don’t provide this level of search maturity and search-engine optimization.
In 2022, the Baymard Institute published an overview of types of ecommerce site search queries and reported that “42% of all sites perform below an acceptable ecommerce search UX performance…. To make matters worse, 8% of sites have a downright ‘broken’ ecommerce search UX…”.
A pretty scathing assessment, though not surprising when you consider the tough job that search engines must do: correctly parse informational, navigational, and transactional search queries, and accurately determine relevance.
What kinds of search queries? Sometimes shoppers know exactly what they want (something they saw on social media, for instance), but more often than not, they’re blindly searching for a possible solution.
Some of the query types Baymard identified include:
With such a broad array of query types, search engines can understandably struggle to interpret them and suggest the most relevant choices.
In this blog post, first we’ll look at the kinds of queries site search engines must contend with, and then how ecommerce businesses can meet demand with a rich search index. Finally, we’ll look at how machine learning is changing the game to improve search results for many query types. .
Most ecommerce search engines do a good job of delivering relevant results for exact search queries. For example, if a pet owner searches for “Greenies” on a pet-supply ecommerce website, they’re almost certainly going to be shown that brand’s dog and cat treats.
What type of product does the user want? Let’s say it’s a chewy dog treat. Search engines can have a tougher time when customers are searching for a product type (“chewy”) as opposed to a specific brand’s item. For instance, a search for “chewy treats” might or might not turn up the best results the user is envisioning.
At Algolia, our advanced search engine would address this issue with a strategy to provide more-accurate results: split the search for “chewy treats” into one for “chewy AND treats”. That means the acceptable search results must contain both keywords in the description or metadata.
(Spoiler alert: With vector-powered search, those kinds of synonyms and descriptive searches are built in. We’ll speak to AI more later in this article.)
What if rather than looking for a type of product, a shopper is searching for help with getting rid of a pesky symptom? As search data indicates, people are often myopically focused on a problem that they desperately want to solve, such as a:
They only know they have an urgent problem. Of course they’re looking for a solution, but they typically have no idea what particular product they need in order to get relief.
In this situation, a site visitor’s search results are generally broad based because the seeker is open to a range of possible solutions. For example, “Migraine headache” search results could offer up painkillers, natural remedies, cold compresses, head wraps, foam rollers, and more.
Different ecommerce search engines typically handle these types of queries slightly differently.
Some search engines assign a relevance score to each document in your index. The score ranges from 0 (no match) to 1 (perfect match), and the search results are ordered starting with the one with the highest score.
Algolia ranks items using a tie-breaking algorithm. It ranks all exact matches first. When more than one record matches exactly, it breaks the tie using a series of tests. First it ranks by geolocation (if the user has that feature turned on), e.g., closest restaurants first. Then it counts the number of words matching (the more matches, the higher the rank), until each record has found its place.
Regarding spelling, if a misspelled word matches a misspelling in the content, that’s considered an exact match. If not, Algolia allows up-to-two-letter misspellings, but in the tie-breaking order, misspelled terms are ranked lower.
Finally, the algorithm applies business-specific criteria to adjust rankings according to business objectives. For example, a company might want to promote its high-margin or trending items at the beginning of people’s search results.
If you want to improve the odds that your retail site search will return the best, most relevant result(s) and help improve your customer experience and search conversion rate, it all starts with building a rich search index. Whether you’re using a product information management system (PIM) or writing product descriptions directly in your online store CMS (such as Shopify), in order to provide high-quality results, you need to build a robust content index that includes product attributes such as:
Then, when someone searches for something, they’ll be more likely to get the right results. With something like “chewy dog treats”, the search engine will put them in the right product recommendations ballpark, as Amazon does here:
As Web marketer Neil Patel points out, many sites lack adequate descriptive data. Adding all the right information not only enhances your product pages but provides powerful cues for search engines to crawl your index.
The process of categorizing and tagging your products can improve users’ search results, plus help people browsing the site through category navigation instead of the search box. Ideally, the category page and/or tag labels will be present in the data when it’s being indexed to provide additional metadata for the search engine. Categories can be added as metadata or even inferred by your site’s URL structure. For example:
Search engines can extract useful category and subcategory data (such as “swimwear” and/or “bikinis”) from URLs like these for returning accurate results from the outset.
Do you offer faceted search? Categories are also used to build filters and facets on your search results pages.
A good product taxonomy doesn’t just aid with on-site search, it helps with SEO and web search engines. Crawlers work best when they’re provided with structured hierarchical data. Additional tagging (e.g., adding color as an attribute) can greatly improve results.
Some things to consider when building rich descriptive data:
Note: It’s possible to categorize and tag too much; you need to find a balance between adding just enough to be helpful but not so much as to affect browsing.
Ecommerce site visitors are looking only for products, right? Not always. Sometimes they just want the search tool to help them find shipping information, the return policy, support, tracking, jobs, or something else. So if your site search is indexing only your product information, it’s missing the proverbial forest for the trees.
Your search provider can:
People may search by entering synonyms for words that you typically use to describe your products. Or they may type in the abbreviation “in” when they mean “inch”; which can confuse a search engine. That’s why we recommend you leverage your search analytics — whether they’re from Google Analytics or the metrics package included with your search platform — to find out what other keywords your visitors are using in their searches.
With robust synonym management and good machine-learning capabilities to develop an understanding of searchers’ intentions, you can offer them a better search experience.
One caveat: developing synonyms for each and every word in a query — especially for terms that are very nuanced — isn’t really feasible. But over time, if enough people are using abbreviations or colloquial terms in queries, an AI-powered search engine can make adjustments to deliver better-converting results.
So far we’ve covered the importance of improving your search index. Let’s turn to some of the ways you can improve search relevance for queries.
AI-based search offers tremendous power through continuous and automatic improvements with intelligent feedback loops: the more data that’s produced from searches and sales, the more effectively a search engine can improve results automatically over time.
Built-in search AI collects search history, learning over time what visitors are looking for and buying, whether it’s a brand name or specific feature or long-tail keyword query like “stunning fall outfit for mother of the bride”. By knowing which searches lead to conversions, a search engine can automatically deliver higher converting results for similar searches.
Soon, the best ecommerce search engines will include vectors — a mathematical approach to representing words that effectively encapsulates the meaning of text and can deliver dramatically better results than standard keyword search. Vector search enables more-relevant long-tail, non-keyword-focused queries.
Search language is often ambiguous because a user’s intent is not always apparent. “Bank” is a classic example: — does the searcher mean a financial institution or the side of a river?
For some ecommerce site search use cases, customers may type in symptoms or adjectives to find answers. Without added context or mention of a specific product, it’s difficult to know exactly what they need.
Natural language processing (NLP), the process of analyzing unstructured text to infer structure and meaning, is one technique available for improving search results. In this context, “structure” means information that is highly defined, for example, a category or a number, much like fields in a database. It can also represent relationships between things. Common examples include sizes, colors, places, names, times, entities, and intent.
NLP is most valuable when the underlying data has significant structure that can be mapped from the queries. For example: “mens size 14 nikes under $75”. In this case, the data can be structured and filters applied automatically on product type, gender, size, price, and other attributes.
If someone visits your site and searches for “warm jacket”, you can deliver different results tailored to their specific individual needs with search personalization. Generally speaking, the more data you have about someone (e.g., pages visited, purchase history, gender, age), the more you can personalize their search results.
Even if visitors are anonymous, a search for “warm jacket” can be personalized using available data such as the browser type, IP location, time of day or year, mobile search vs. desktop, and other attributes.
In the example of “warm jacket,” you might promote different clothing to someone in Florida vs. someone in Minnesota. While you may not know anything about the user, you can still personalize search results for them based on their IP location.
Most search services provide a feature that lets you add rules (or adjust the search algorithm) for handling different types of searches.
Rules can help for difficult search queries, and they offer another way to provide results that match a retailer’s criteria for displaying products — for example, best sellers, high or low inventory levels, the user’s location, price, merchandising, buying history.
Adding autocomplete or instant search functionality to your search bar further enhances the search experience and can help visitors enter better queries. Added bonus: it helps people eliminate typos, which can further speed up their searching.
The autocomplete user experience — also known as query suggestions, type ahead, or autosuggest — gives users query suggestions as they type in the search bar, using analytics (top searches) and context to predict what they’re looking for.
Autocomplete can steer a visitor to the right category or product page faster and signal that you offer what they want. While autocomplete doesn’t improve the ability of a search engine to handle different types of queries, it can be massively beneficial for delivering useful results, eliminating irrelevant results faster, and facilitating excellent overall site usability.
Alternatively, rather than receiving query suggestions in real time, instant search refreshes the actual search results page in real time as the user types their query. This is known as instant search or search as you type. Image search results appear almost magically, before a shopper has finished typing their query:
When it comes to handling different types of search queries, ecommerce platforms can be hamstrung by challenges. While you obviously can’t read the minds of your visitors, a good search engine powered by the right artificial intelligence can seem positively clairvoyant in figuring out their search terms and meeting their needs.
Whether you have a few hundred web pages or millions of SKUs, Algolia has your ecommerce search covered with cutting-edge features and speed, plus a new hybrid API. Sign up for a free trial or personalized product demo and see how we can help you expertly handle your visitors’ search queries and get them quickly through the virtual check-out counter.