The importance of Site Search Relevance and Boost's algorithm to fine-tune it

Posted by Harvey Nguyen on

The importance of Site Search Relevance and Boost's algorithm to fine-tune it

It is no coincidence that businesses are increasingly investing in site search. The benefits of a relevant and efficient search engine are undeniable.

On the customer side, a successful site search will return accurate results quickly, saving customers’ time and leading to the final steps of a purchasing process. 

Up to 30% of visitors to a website use the site search box, and those who use site search are presumably more likely to buy something compared to those who just use standard website navigation, because they already have the intention of what to buy. 

On the business side, site search is a huge source of shoppers' behavioral insights, giving the company valuable analytics data. This information, if used successfully, can increase the conversion rate and customer retention. 

Let’s have a look at why site search relevance is important to your business and how Boost’s algorithm helps you to increase the efficiency of your site’s functionality.

What is Site Search Relevance?

Site search relevance refers to the way the results match what customers are looking for when they are searching on your website. Can your users search for exact results easily and quickly? Does your site search provide what your user needs? If your customers can find what they are looking for, you have created a successful search engine.

When it comes to evaluating the effectiveness of your site search, relevance is a vital factor. The concept of site search relevance is to show the results that are closest to the searchers' desire. There are two metrics which measure the relevance of your site search: recall and precision. Focusing on both metrics can help you maximize conversion rates. 

  • Recall implies the proportion of relevant search results the search engine returns among the total relevant results. For example, 20 products are qualified to be displayed on the search results page if customers run a query “top hat", but only 11 are shown. The recall rate we have is 11 out of 20, or 55%.
  • Precision means the proportion of only relevant search results the search engine returns among the total search results retrieved. For example, of 11 results displayed for the search term “top hat”, 8 products are top hats and 3 are classic hats. Then the precision rate is 8 out of 11, or 72.7%. The inefficient Natural Language Processing (NLP) capabilities may be one of the main reasons for low precision rate. 

These two metrics seem similar at first, but we hope these above examples will help you imagine how different they are. The most ideal case is when both recall and precision rates are 100%, but of course, it will not happen. In fact, precision is often at odds with recall, and vice versa. If you want to improve recall, then the search requirements will be loose, the search engine will return more erroneous results, thus decreasing precision. 

On the flip side, if precision is high, then the search requirements will be tighter, your search results will eliminate well-matched ones, thus lowering recall. 

For this reason, we suggest making improvements quantitatively by trying running searches on your site manually, so that you can measure the values and get a general metric for precision and recall. 

If you do not want to optimize your search function manually, machine learning will help you by optimizing itself automatically based on real-time usage data. Optimizing search results with machine learning can help increase conversion rates because it refines search result relevance, thus improving recall and precision rates.

The goal of e-commerce businesses is to provide what customers need, and improve their retention. One of the essential steps in achieving is to optimize customers’ interactions with your site search so they can find exactly what they want. 

Why is Search Relevance important to your website?

The search function on your website is often the first interaction between a potential customer and your online business. Therefore, you need to satisfy your customers right from the strong site search.

Making a successful and effective e-commerce site search depends a lot on the relevance. When your customers search for products on your e-store, it is essential to show results most related to what’s in their minds. If they get unsuitable results, they will be dissatisfied and less likely to come back to your website. Without relevance, search is just a mindless tool that gives back results containing search keywords.

Research shows that conversion rates through site search can reach 50% higher than average. If consumers have a bad experience with site search, like not getting the results as they want and wasting too much time waiting for the results to display, they will switch to other competitors who offer a better experience.

Here is another statistic: on average, over 96% of visitors will leave your website without purchasing anything. If you want to raise this super low rate, start with improving your site search relevance.

Boost’s algorithm to increase Search Relevance

Boost’s search engine analyses the information and returns the products, collections, pages, and blog posts that best match search terms that users enter. The search results mainly depend on the Field area and the Minimum percentage a search term must match.

Field area

Boost’s search engine looks for matched products in the fields of:

  • Product Title
  • Product Type
  • Vendor
  • Tag
  • Product options
  • Product SKU
  • Product Description

By default, the Description is turned off. You can disable a specific search field or change the priorities of the fields to your preferred priority order.

field area site search relevance

By default, the criteria of product attributes, including Product Title, Product Type, Vendor, Tag, Product options, SKU, and Barcode are searchable, while Description is turned off. You can turn on/off this default setting and change the priority order of these attributes as you want.

Minimum percentage a search term must match 

The default and recommended minimum percentage a search term must match your product data to make search results appear is 100% (for 2-word search terms), and 85% (for search terms that contain 3 words and above).

Minimum percentage a search term must match search relevance

Default minimum must-match percentage for search terms

However, you can completely change our recommended minimum percentage a search term must match by using the “Advanced settings” in our app. 

For example, if you modify the minimum must-match percentage for search terms with 3 words and above at 40%, then a search term of 5 keywords only needs to match your product data to at least 2 words to get search results. 

Boost’s features to enhance your Search Relevance

We have always updated features of Boost's Product Filter & Search to optimize your shoppers' searches, so it returns products that not only match the queries, but also match the searchers' intention. The following are some recently released enhancements that you should try to leverage search relevance.

Suggestion dictionary

In the Instant Search widget, you can see a section called "Popular Suggestions", which suggests the most popular search terms of your store. The suggestion terms are built by an analysis process, which generates common terms found in product titles. However, if you want to push the sales of an item -  for example, “woman hat” - add its name to the Site Search Suggestion Dictionary and set its priority to “High”.

search suggestion site search relevance

 

When your customers search for the keyword “hat”, they will see the Suggestion section with “woman hat” as the first item and are therefore more likely to click it.

Synonyms

The Site search function on your website will be more relevant thanks to Synonyms. It works its magic when you want different search terms (usually synonyms) to bring up the same results. Now, customers can also search for keywords not included in your product data.

synonyms example site search relevance

For example, "beachwear" is not included in any of the Bikini product attributes. Customers are searching for “beachwear", but it is the synonym of "bikini". When you add "beachwear" to the Synonym list of "bikini", shoppers can find all relevant products even when they type “beachwear” in the search box.

Stop words

Stop words allow you to filter unimportant words from your users’ search terms so that they get a more relevant search result. The words added to the “Stop words” list will be filtered out during the search, helping the site search engine focus more on the important terms, delivering faster and more accurate results.

For example, when a customer types in the search box "casual outfit", it returns products that could possibly match either "casual" or "outfit". This could bring out other irrelevant products or brands in your store that contain the word “outfit”, such as “formal outfit”, “office outfit”, “trendy outfit" or “beach outfit". Therefore, excluding the word "outfit" from search returns will bring out only results related to “casual outfit”.

“Did you mean" feature

The “Did you mean'' feature in Boost's Search is developed to help shoppers discover the products even when their search terms are misspelled to a certain level. 

did you mean feature site search relevance

For example, when users want to search for “headwear" but mistakenly type “headweat", besides the notification of no results, the “Did you mean: headwear" phrase will appear to re-confirm the right search term, accompanied with product recommendations. (Source: Cherry Red Vintage)

Under the “Did you mean…?” phrase, similar suggestions to the search phrase will be shown.

When one of the keywords in a search phrase does not match the product data (searchable fields, like title, product type, vendor, tags, description, etc.), then Nothing found for {search-phrase}. Did you mean {suggestions}?: will be shown.

  • If the error margin is acceptable, users will see “No search result and Did you mean”: Sorry, nothing is found for {no-search-result-kw}. Did you mean {suggestions}?
  • If the error margin is unacceptable, users will see “No search result” only: Sorry, nothing found for {no-search-result-kw}.

This solution was improved to not show “Did you mean” when a search term does return results. For example, if you have both “wine" and “wide" in the product data, if you type “wine", then “Did you mean: wide" will appear. 

No search result suggestions

When your users search for keywords that return no products, the No search result suggestions feature will provide them with other alternatives so that they discover other results of your choice, increasing the chance of them browsing and making a purchase decision.

no search result suggestions site search relevance
When there are no results for the keyword “husky", instead of pushing them into a dead-end, the “No search results suggestion” feature helps you to recommend other keywords that people are also looking for and top trending products. You can easily choose the Popular Searches and Trending Products in the Search settings.

It is normal for websites to sometimes not display what users are searching for. Their queries may be so complex and cryptic that the search engine cannot offer exact results. The “No Results” page may push users to a dead-end, leading to no purchasing decisions. Rather than only displaying “No results” or “Sold out” pages, suggest other products and pages that are most related to your users’ queries. 

See more: What you should do with “No Search Results"

To sum up

Relevance has become an important factor that needs to be considered when improving the search function on your website. It helps to build a bold relationship between your business and customers, and satisfies their needs. An inappropriate site search engine leads to customers leaving your website due to unmet needs. Then, you will hardly see any increase in sales.

Boost’s Product Filter & Search is one of the useful tools that can help you fine-tune your site search, thanks to its optimized search relevance algorithm and unique features. If you have any customization requests, our support team is willing to make a tweak. Have a free-trial here and tell us how you feel.

To gain a deeper insight into the app setting, check out our tutorial video series and our Demo Store

Follow our Facebook, Twitter, and LinkedIn to get updated information!