An ecommerce search engine that’s highly accurate and easy for your customers to use is a critical success factor for any retail company selling products online. Improving ecommerce search can have a significant business impact on the bottom line. But to gain this benefit requires careful planning, leveraging the proper tools and maintaining an ongoing commitment to best practices.

What is ecommerce site search and how does it work

Simply put, ecommerce site search is the process by which visitors to your site find what they want. In the early days of ecommerce, this simply meant matching keywords typed into a search box with indexed product names. Modern ecommerce site search, often termed “intelligent search” or “smart search,” goes far beyond simply matching keywords to indexed product names. It corrects spelling errors, handles synonyms, and uses AI and natural language processing (NLP) to infer shoppers’ intent. 

Today’s top search engine algorithms not only identify the right products from the consumer’s point of view. They also let companies influence the ranking of results based on internal factors (profit margins, inventory levels, etc.) as well as consumer data such as purchasing history to provide personalization.

Business impact of ecommerce site search 

Getting search right can have a very big impact on your bottom line. Consumers who use search are 2.4 times more likely to buy, and whether they buy are not is clearly influenced by the quality of the search experience. Desktop conversion rates are 2x higher for retailers with advanced search vs. basic search capabilities. Given these numbers, intelligent search clearly makes sense. But how should you get started?

Challenges to search implementation

Like any major project, implementing smart search is not without challenges. Here are several of the most significant:

• Data management: Smart search depends on AI, and AI depends on reliable data. This means that you need to have processes in place to make sure customer and product data are current. 

• Integration: Your search system will surely need to integrate with at least one or two existing internal applications, and perhaps more. 

• Customization: Search engines offer varying degrees of customization and you need to be clear about what you need in this area.

• Event tracking: Events like purchases, add-to-cart, ratings, and more are vital for training AI search, personalization, and recommendations.

• IT help: Whether it’s “data wrangling” to get data into usable, compatible formats or integrating applications, you’ll need IT resources with the requisite skills.

Once you’ve gotten past the initial set-up, there’s a second layer of challenges: making the most of your search investment. In this blog, we’re going to share 15 best practices and examples for increasing conversions. We’ve analyzed scores of our own customer search accounts and scoured the web to find great examples of search UX that works.  While there is no one-size-fits-all approach, there are general tenets for on-site search success.

15 best practices for ecommerce search (with examples)

1. Make search more visible

 

Both Make-up Forever and Bicycle Warehouse have search on their site and many categories and collections to guide visitors on their buying journey. But, the search bar visibility is significantly greater for Bicycle Warehouse. 

What difference does visibility make? It can have a great impact on search and site performance. 

When one company made their on-site search more visible…

• Site searches climbed by 439%
• Visitors stayed on the site 110% longer when they used search
• Page views increased by 45%
• Goal completions increased by 242% 

Enlarging a search bar is a simple site update, but can have an outsized influence on site performance.

2. Use typo tolerance to avoid losing customers

An invisible and often overlooked UI element to ecommerce search is how to handle typos and misspellings. Somewhere between 20-30% of search queries can contain a misspelling! You don’t want to lose customers because they’ve misspelled queries only to get a  “no results” page in response.

There are a few different ways to manage search typos and misspellings. A good site search tool will give you options, including a way to add “did you mean?” results in search result pages and/or autocomplete with typo tolerance. 

Using typo tolerance of some flavor is a best practice to avoid losing customers who think you don’t carry “bakcpacks” or “addidas.” 

3. Smart merchandising — promote relevant content at the right time

Shoppers are visual. In-store merchandising — whether it’s in an aisle or at the register — is an essential practice to encourage shoppers to add one more item to their carts. 

Online retailers have it tougher. You need to balance getting shoppers to the thing they’re looking for first. Then you need to determine when and how to present other products, and which related products to display.

Here we can see that Chewy.com makes it easy to find related items — which could be some combination of best sellers, higher margins, or personalized results.

They also offer similar results after you add the item to your cart. 

It’s smart to offer related items in both locations to capture merchandising opportunities wherever the visitor chooses to go, and can be designed to be unobtrusive to not interfere with the buying experience. There are also many solutions available to help optimize for related items including our own Algolia Recommend.

4. Incorporate natural language processing (NLP) for highest quality results

If a user types in “red basketball Nikes size 14” on your shoe store’s search bar, the search engine needs to parse this information to return high quality results. The use of natural language processing (NLP) for ecommerce websites can impact your brand perception, user experience, and conversion rates.

I ran a search for “red basketball Nike size 14” on FootLocker.com. At first glance, it appears their site isn’t parsing the query.

Then I restarted my visit from scratch. This time I didn’t use the search bar. Instead, I browsed their site by clicking on men’s basketball shoes and sorting by filter. There were actually a ton of results! I clicked into each of these results and some with size 14s were in stock. In other words, these products should have shown up on my first search.

If a visitor had only used search, they may have abandoned the site without realizing there are more products available.

In contrast, Zappos does a very good job with a similar search. I added “shoe” to the search phrase because Zappos sells other clothing items. 

A complex search like “red basketball Nikes size 14” is becoming more common. Studies show that one-third of all searches contain 4+ or more keywords. 

In the context of site search, natural language processing (NLP) analyzes queries to infer structure and meaning. Structure, in this case, is referring to information that is highly defined, for example, a category or a number. It can also represent relationships between things. Common examples include sizes, colors, places, names, times, entities, and intent, but there are many more. 

Investing in a search solution that supports NLP is a best practice for ecommerce providers to help ensure visitors get relevant search results to find the right product quickly. 

5. Use filters and facets to help customers zero in on relevant options

Moosejaw should win an award for their long list of filters and facets. It reminds me of a scene in the movie Airplane — the lists seem to go on forever. More is certainly better than few or none. 

Let’s explain some terms briefly. When it comes to search, there are actually two kinds of filters:

• Facets: These are dynamic lists that change depending on the search. In this case, the search is for “jacket” and the facets shown are different from the ones you’d find for other searches like “boots” or other items.
• Filters: Filters typically don’t change; it’s a static list regardless of what product you search for.

More on the difference between facets and filters here.

If you have a large store with hundreds or thousands of items, it’s impractical to design facets and filters for each product category. The trick with filters is to generate them dynamically based on product metadata at index time. 

For example, Algolia customers can simply wrap any HTML element in special filter tags to tell the app to store the element’s attributes. Those attributes can then be used to create filters and facets which are searchable in queries.

Another option for facets is to use query categorization, which can classify information to generate filters and facets automatically. 

Faceted search is invaluable for helping customers narrow down results to find exactly what they’re looking for. The same facets and filters you create for site search can also work for your static or dynamic catalog or collections pages. 

6. Integrate personalization for more relevant results and higher conversion rates

Personalization is about improving search relevance to boost conversions. As known users search your site, the search function should deliver personalized results — specific products that a user is more likely to purchase.

The use of personalization can tremendously improve user experience and conversion rates. BigCommerce found that personalization can lower bounce rates by as much as 20-30%. (Typical retail site bounce rates are around 30-55%). 

In a McKinsey study, 100% of the top quartile retailers said personalization is a top 5 priority for them, but only 15% have implemented it across channels. 

Amazon is famous for building personalization on their platform. As HBR reported:

“The company has effectively created a new model for what digital marketing should actually look like: a one-on-one relationship with the customer, informed by data collection, optimized with machine learning, and nurtured with other forms of AI.”

The HBR author went on to ask: “How does an offline player…  hope to catch up?”

Amazon has more than a thousand engineers working on the problem! Whether you’re an offline or online player, the task seems daunting. 

The good thing is that with Algolia, or other third-party personalization solutions, you do not need an army of search engineers anymore. 

Whatever solution you pick, data is key. The more data you collect — while respecting data privacy laws — the better. Once collected, data is used with each query to change the results to be more relevant to the individual. This can be as simple as a clothing size preference, gender, location, or any other individual characteristic that can influence their results and behavior.

Typically the first stage of personalization is sending the information through with the queries to be recorded with other analytics. The performance impact can then be analyzed offline to determine if this is useful, or you could even A/B test results with and without personalization.

Often queries can be layered with personal information to augment and influence queries to produce highly personalized results.

7. Design your U/X to include mobile search for maximum reach

In Q1 2019, smartphones accounted for 65% of retail site visits worldwide. Mobile-optimized search is challenging, however. Responsive design isn’t a simple matter of resizing a few objects — it’s taking into account the entire buyer’s journey. 

Here’s a terrific example: Untuckit has done a good job of building a mobile-friendly site with an easy-to-find search button right at the top of the screen. They offer both a convenient drop-down menu of common filters and a button for additional filters. As I typed, the search bar displayed auto-suggested results. 

8. Recognize synonyms to include all relevant results

Are your customers searching for a t-shirt or tee shirt? If your search box only recognizes one spelling, you’re going to miss opportunities. Using synonyms to create a dictionary of terms is a best practice for healthy site search. 

Related terms should work alike, too. For example, blouse, top, shirt, or jersey will contain overlapping items, as will the terms striped shirt, striped top, or breton. In this case, “top” is a category that would include a blouse or breton, but it can also be used as a search term itself. 

Today, searches like these are relying on good titles, product descriptions, and metadata. Soon ecommerce search engines will offer vector-based search which uses a mathematical approach to understanding words and their meanings to deliver more relevant results. 

9. Instant search with product thumbnails makes the U/X faster and richer

HannaAnderson.com offers a masterclass in instant search. Here’s what’s happening in the screenshot above:

• Results and thumbnails instantly update as you type
• A “did you mean” spell checking feature is displayed
• Related collections (e.g., the Easter Shop) are displayed
• Item prices are listed

For shoppers, instant search like this is a winner because it’s so visual. Like autosuggest (or autocomplete), instant search shows results as you type, but also includes thumbnails. Also displaying price is a great way to help shoppers find what they want fast.

By the way: The HannaAnderson search works brilliantly on mobile devices, too (I checked).

For some reason, most online stores either haven’t invested the time in instant search — they still require people to hit “enter” on their keyboards after typing a query — or only offer limited result displays. Our ecommerce site search product and many others now offer some flavor of instant search out-of-the-box so any company can build a rich experience in minutes. 

10. Machine learning improves relevance over time

It’s well known that Google improves web search results based on search activity. If you click a result, visit a site, and then immediately bounce back to the search results, Google understands that the result you had just clicked on was a poor match. Over time it adjusts and refines the algorithm to boost better results to the top of the page. 

New site search engines have similar capabilities these days to connect searches to  ecommerce conversion rates. They can distinguish a good result from a bad one based on clicks and conversions (shopping cart activity, purchases, checkout, trade-ins, or whatever you consider success) to deliver better results that improve your conversion rate.

For example, “Face masks” meant something completely different in 2019 versus 2023. Prior to COVID-19, a buyer who was searching for a face mask was probably looking for cosmetic skincare products. That all changed with the pandemic. A search platform with reinforcement learning would automatically begin to adjust results based on conversions. 

As a best practice, you should be able to adjust the inputs, too. You may want to tweak search results towards higher margin items or layer personalization on top of the results. 

11. Expand consumer possibilities with visual search

Visual search, sometimes dubbed visual shopping, is becoming increasingly important for marketers when designing their on-site search and developing SEO (search engine optimization) tactics. 

Over half the human brain is dedicated to vision and visual processing, and for most buyers images are more important than text. 

Visual results should appear for both instant search results and on your results page. There are other use cases as well for visual search, including augmented reality (AR). Even Snapchat has gotten into the game with visual search optimized for AR. 

12. Use A/B testing to list high-conversion results first

In a classic website A/B test, you might split traffic to determine which variation — a blue or green call-to-action button — converts better. You can do the same with site search results by either testing the search result design or the search data itself. 

Here’s just a few examples of things you could test with A/B testing:

• Personalized vs regular results
• Best sellers vs top rated items
• New vs discounted items
• Different product categories 
• Top sellers in the last 7 days vs all time

13. ADA and WCAG usability

It’s reasonable to assume that a significant number of your website visitors — as many as 1 in 5 — will be people with disabilities. This can be anything from visual impairments like color blindness to cognitive impairment such as dyslexia. 

In one study, as many as 71% of people with disabilities bounced from a site they were shopping on due to poor experience. 

For such disabilities, aspects of your website like color contrast, animations, and transitions need careful consideration, and appropriate alternatives should be made available. Site search, whether it’s your search results landing page or typo tolerance for misspelled words, is no different.

For the search bar, some considerations include:

• Locating the search bar in a predictable spot (typically top center or top right)
• Using distinguishable colors
• Using alt-text on the search magnifying glass icon

Similarly, for the search results, some considerations include:

• Allowing users to navigate through results using their keyboard
• A status message for screen readers
• Search results with images

You can read more about suggestions on our longer post about Web Content Accessibility Guidelines (WCAG) and the Americans with Disabilities Act (ADA) digital standards and search. 

14. Extend search beyond products for customer convenience

Site search should provide a great user experience whether someone is looking for a product or an answer to a question such as your return policy. Whatever search solution you select should be capable of delivering results and, ideally, categorizing those results for easy scanning.  

15. Analyze search metrics to improve conversion rates

Your site search provider should offer built-in search analytics to help you understand trends, ineffective searches (like no results), and areas for improvement. 

Search volume is a good indicator of overall traffic and tells you how many people use the search on your site. Click-through rate helps you understand how effective your search is and whether you have the right content.

The top searches and unexpected trends show your users’ intent. This is an important list to keep an eye on. If people’s interest changes, so will the list of top searches. It’s a great indicator that you might have to add additional content to your website or stock new products in your ecommerce shop.

Reports on things such as missing content, query terminology, and number of searches per session can help you uncover opportunities for improving site search.

How to measure the impact of improved ecommerce search

If you manage an ecommerce site, you’re no doubt drowning in data. When it comes to search, some numbers are more important than others. The search reports in Google Analytics (GA)  identify what visitors are looking for after they land on your website, and what search terms they’re using. GA also measures engagement for those searches. 

Image via OptimizeSmart showing on-site search tracking via GA4

However, for search optimization, the number we’ll focus on is the percentage of your site’s total revenue that comes from site search. This is the number we’re looking to improve. In this section we’ll review the KPIs that influence it 

There are several numbers that indicate progress towards the goal of increasing search driven revenue. At the front end of the customer journey, it’s important to track the percentage of site visitors who use search. A low number can mean you need to look at your landing page design to make sure the search window is easy to find. Click-through rate (CTR) is one of he best indicators of success. The higher the CTR, the higher the relevance of the results your search engine is delivering. Conversion rate is obviously important, but it’s affected by factors beyond search such as aspects of the product itself and price point. The same is true for average order value (AOV).

As well as tracking successes like a higher AOV, there are several types of negative events that deserve ongoing attention. Search exit rate indicates the percentage of visitors that exit search without clicking on any results. Smart search minimizes this problem in numerous ways. To cite one example, it’s aware of synonyms, so that someone looking for a couch can find one even if they type in “sofa.” A high percentage of searches with no results may indicate a similar problem. Smarter search can reduce both these numbers.

Another important number that can indicate problems is searches per visit. In this case, the ideal number is one. The more times visitors have to initiate a new search to find what they want, the more frustrating the experience becomes. Smart search reduces this number by basing results on the intent of the search.  

There are several other metrics that are important in search analytics. Search volume is a good indicator of overall traffic. The top searches and unexpected trends show your users’ intent. This is an important list to keep an eye on. If people’s interest changes, so will the list of top searches. It’s a great indicator that you might have to add additional content to your website or stock new products in your ecommerce shop.

Reports on things such as missing content, query terminology, and number of searches per session can help you uncover opportunities for improving site search.

Getting started on the road to improved site search

The first step in moving to intelligent, AI-powered search is to establish goals. This means both hard numbers, such as projected revenue uplift, and soft benefits, notably a better, smoother CX. The next step is establishing which vendors can provide the tools you need to achieve your goals. The basic “must-have” capabilities you can expect from an intelligent search engine are:

• Browsing, aided by filters and facets

• Text processing, including stemming, type-ahead/autocomplete, synonym recognition and other such technologies 

• Semantic search that goes beyond recognizing words and their variants to understand meaning and intent 

• Analytics. The ability to analyze and react to data is at the heart of ecommerce success, and the appropriate tools should be incorporated into a solution, rather than “bolted on” by third parties.

• IPAs for quick, efficient integrations without extensive coding

• Personalization of results. This is particularly important, as 71 percent of consumers expect it, and it works. To cite just one example, Algolia client Huckberry realized a 9.4% increase in website revenue when it adopted personalization.

Additional capabilities may include:

• Product recommendations (including replacements for out-of-stock items as well as upsell/cross-sell)

• A/B testing of search results. This capability should be included, as it is one of the best tactics for improving results.

• Visual search

One other very important point to consider is scalability.

Top intelligent search vendors 

There are quite a few vendors offering intelligent search, and start-ups will likely add to their number in the near future. Here are some established, trustworthy options worth exploring. 

Algolia offers a rich, AI-driven feature set including an unique hybrid search approach which combines the sophistication and high accuracy of vector technology with the speed of keyword search. In addition to semantic search that understands intent, Algolia offers personalization, visual search, A/B testing, strong analytics and machine learning that continuously improves search results. After set-up, non-technical people can easily modify system parameters. Designed for scalability, Algolia is an API-first SaaS solution that’s compatible with all the major cloud infrastructure providers. The company has a global customer base. 

Coveo markets itself as a search and genAI platform. Its semantic search product offers a full complement of features from which customers can choose, although implementing them is likely to require developer skills. Coveo positions its search offering as suitable for B2B as well as B2C use cases. Its focus is on large enterprises, and it has a significant learning curve. Coveo is a SaaS solution that’s compatible only with AWS. 

Bloomreach offers a variety of products beyond search, such as email marketing tools and a (separately sold) customer data platform, all powered by a centralized AI engine optimized for online marketing. The company’s full-featured semantic search product is most appropriate for enterprise-scale C2C businesses. It offers highly-rated merchandising features, but gaining optimal value from them involves a steep learning curve. Bloomreach is an SaaS company that runs on both AWS and GCP with limited autoscaling due to capacity.

Searchspring offers AI-powered search, merchandizing, personalization and recommendations focused on the needs of small to medium sized companies (SMBs). According to Gartner, Searchspring has the largest integration partner ecosystem among its peers, a significant benefit to companies with limited IT resources. Three tiers of service (and pricing) are available based on search sophistication, indexing frequency, and usage factors (number of domains, sessions, and SKUs). Searchspring is an SaaS solution that runs on AWS. 

Vertex AI (formerly Google Retail Discovery) includes search, browse and recommendation offerings that benefit from the company’s huge amount of data available from Google Shopping, which enables out-of-the-box intent detection. Products are marketed only in the form of consumable APIs, and are only suitable for large enterprises with significant IT resources. Pricing is based on actual monthly usage (API calls). 

Next steps: ecommerce site search solution 

We presented a ton of information above. It can seem daunting to try to do everything. 

At the end of the day, you want to deliver the most relevant results and best user experience for every query, and that requires a combination of the features and functionality we’ve outlined. 

The good news is that many modern ecommerce search platforms offer many, if not all, of the features mentioned above. 

FAQs

What is smart search in ecommerce?

Smart search goes beyond simple keyword matching to give shoppers more accurate results, which in turn lead to higher conversion rates. At the front end, smart search can understand the intent of a search through a complex mathematical process. It provides filters and facets to help customers zero in on the specific items they seek by price, color, compatibility, etc. Most importantly, it learns continuously from consumer actions to constantly modify rankings and improve results.

What is visual search in ecommerce?

Visual search leverages the image recognition technology often associated with facial recognition to deliver a new shopping experience that’s quickly gaining popularity, particularly among young adults. (EMARKETER reports that 30 percent of adults 34 and younger have used visual search.)

Here’s an example. You just saw some new running shoes on your Instagram feed that you might want to buy. You go to your favorite ecommerce site and instead of typing “running shoes” into the search you paste in a picture of the shoes copied from your feed. The result: a faster path to the shoes you want. For more examples of how visual search works, click here.

What is AI-powered search?

AI-powered search involves numerous technologies, all of which support smart search. Among the most important:

• Natural language processing (NLP) enables search engines to understand and respond to human language.

• Vector search converts search queries into mathematical patterns called vectors that can be used to help determine the most relevant results.

• Learning to rank algorithms automatically improve result rankings over time using event signals to grow clicks and conversions. 

For more information on how AI-powered search works, click here. 

How is AI changing ecommerce search?

AI is changing every aspect of online shopping. The ability to deeply understand queries and personalize responses leads to more accurate search results, greater customer satisfaction and higher conversion rates. AI can be found in everything from chat bots to visual search. For ecommerce companies, AI can optimize results faster than people, and for consumers it offers new ways to get more value faster from their favorite brands.