How to Improve Your Site Search Relevancy and Customer Experiences
We’ve all been there. You go to a website and enter a key phrase for a specific product or information and a “No Results Found” message displays. In under a second, the user has a negative impression about the website. It only takes about 50 milliseconds (or 0.05 seconds) for a visitor to form an opinion about your website that determines whether they like your site or not—and whether they’ll stay or leave.
As an example, let’s say you’re looking to purchase a two-foot ladder. You enter “24-inch ladder” into the website search bar, but the site search hasn’t created a relevant filter that precisely matches this search term. So, you try closely related variants of the item (such as “2 ft ladder”), and finally, after multiple attempts and possible key phrases, the product displays. Already, your customer experience falls short. Research shows that 88% of online consumers are less likely to return to a site after a bad experience. Users in this situation are faced with high friction by having to work too hard to find what they need and will likely leave, go to a competitor’s site, and never return.
Site Search: Essential to Your Customer Experience
According to a Gartner article based on their 2019 customer experience management survey, customer experience is the new marketing battlefront, especially when talking about the user experience on websites. (See the Gartner blog: “ Redesign Your Website to Do What Your Customers Want ”) No one disputes that product, price, and performance matter, but customer experience is now and in many cases, the basis of differentiation. According to the survey, 75% of organizations can show that customer satisfaction leads to revenue growth through increased customer retention or lifetime value.
The bottom line: site search experience is essential to the overall customer experience. In the age of Google and Amazon, a fast, easy, and relevant search experience--that is, how closely related a product or document page is to a query--is no longer optional, it is expected.
Site Search Factors to Improve Relevancy
We’ve pulled together the key factors to help you improve your site search relevancy (and customer experience) for a product search, but the same principles apply to information searches. The relevancy algorithm is the primary influencer of the search results. Once relevancy has been established, a best-in-class site search capability will offer a boost/bury metric and a popularity/learning metric to be applied to enhance it further. All of this information (relevancy + boost/bury + popularity/learning) is tabulated to provide a product score. This score is what is used to make the product more visible to the user. Here are highlights of the factors to consider. Additional information can be found here .
1. Site Search Algorithm
Let’s say you are interested in purchasing a red shirt, which you enter as your search term. The website algorithm should enable you to see results based on product name and variations on product description and category search phrases, as shown below:
Product Data (ABC123) Name: Red Shirt Category: Shirt
Product Data (BCD123) Name: Red Shirt Description: Columbia Sport Red Shirt Category: Shirt
Product Data (CDE123) Name: Shirt Color: Red
Once the search for “red shirt” is executed, the products display in the following order based on relevancy.
As you can see, the search algorithm, which is configured in the background, incorporates the following factors including matches based on:
•Exact phrase matches
•Number of times the terms appear within the fields used to establish relevance
•The proximity of the terms in the case (upper and lower casing) of phrases (for two or more terms)
•The boost of the data field (a section within the indexing of the data that explains how a field can be “boosted”)
•Boost and bury feature that can improve the order of results (for a category results page or a search results list page)
2. Boost Scenarios
Types of scenarios the boost feature can be used for:
•Boosting specific products on a navigation page
•New products within a category
3. Bury Scenarios
Types of scenarios the bury feature can be used for:
•Burying undesired products (the products are relevant but should not be displayed as high within the results)
•Burying a specific brand of products
4. Popularity Learning Search
You need the ability to collect click metrics for each website. These metrics can be used to influence the ordering of the products. As mentioned, the popularity metric is tabulated into the relevancy score of a product.
5. Visibility Rules
If there is a need to remove items from a results page, the visibility rules should allow you to do so based on a global level or a set of conditions.
6. Common usages for visibility rules
•Remove irrelevant products from search results
•Remove products without inventory
•Remove products based on a specific criterion
7. Troubleshooting the Product Score
If the red shirt you search for is not displaying where expected, you need to determine why. An ideal site search capability will indicate how to boost value. The system should be able to indicate which boost and bury value the red shirt is assigned to. This will help determine what changes should be made to get better search results. If the boost or bury rule is not listed for the product, you need to confirm whether the trigger is set up properly. Here are the factors that should be used in the query generation process to impact product score calculation and relevancy:
•Tokenization: Tokenization allows variations of a word to be searchable. For example, when a user enters a search query for “read” other matching words like “reading” and “reader” stored in the index should be considered and all matching records should also be returned within the results set.
•Query this Field: The “Is Quarriable” flag on the fields section controls whether that field is also used as a separate clause, in addition to being part of the content field as noted above. The content field should always be set up as quarriable and stemmed and stored in the index.
•Wildcarding: Wildcards or prefix queries should be enabled on certain fields like “title”, and “SKU” to allow for partial matches. If any fields have these settings enabled, then you should be able to add appropriate term clauses to the query index to view partial matches.
•Field Boost: If boosts are added to specific fields, these are accounted for when generating the query and calculating the score. For example, if “title” has a boost of 20 on the field and “description” does not, then a match against the “title” field will return a higher score than the match against the “description” field.
•Phrasing: If the keyword contains a phrase, matches against the exact phrase, and the appearance of the individual words on the phrase should be considered when determining the score of the record. For example, when a user searches for “red shirt,” the scores are calculated differently depending on whether the words “red shirt” appear together in the same order, or if they appear on the record but individually as “red” on the color field and “shirt” in the title field. For exact matches, a higher score should be applied.
•Term Frequency: The higher number of times a term appears within a record’s searchable fields, the higher the score is.
•Rare Appearance Rate: Rarer terms that match within a record’s searchable fields give a higher influence on the score.
•Number of Query Terms: The higher number of query terms that matched within a record’s searchable fields will provide a higher score than other products that matched fewer query terms.
•Numerical Sorting: Your index feature should do lexicographical sorting (or alphabetical sorting).
Time to Act!
In summary, you can see why search relevancy and customer experience are important to the overall success of your customer’s journey. Finding the information and acting on the results will provide better results and increase your business revenue.