Constructor’s Related Search dynamically generates suggestions that guide users towards other or more effective queries. It works by analyzing a user’s search query and applying different methods to generate a list of related terms that might interest the user.

Key benefits

• Suggest strategic related search terms
  • Constructor uses multiple methods to generate related search terms and ranks every one to ensure only the most effective ones appear for users.
• Enhance user experience
  • Enhances a user’s experience by providing relevant, related search terms based on their queries, helping those users find products more efficiently.
• Ensure continuous product discovery
  • Even when initial queries return low or zero results, related search terms can present alternative search opportunities and help users explore categories and products they may not have considered.
• Reduce bounce rates
  • When presented with related alternatives, users are more likely to stay and continue browsing other relevant products instead of leaving the site. It also encourages increased time on your website.

Methods used for Related Search generation

Constructor uses four unique methods of generating related search terms.

Expanding

The expanding method zooms out on a user’s query to deliver broader terms related to the query. Suggested terms might include part of the original search team, or terms that have some type of interaction history with the original search term (i.e., a user searched the original search term then rephrased their query to a new term with success, and now that new term is surfaced as a suggestion).

For example, a user searches red pepper chocolate. The expanding method would return broader search terms, such as chocolate and flavored chocolate, allowing the user to quickly navigate to a broader search if the original results do not meet their needs. Both related terms incorporate the original term, while flavored chocolate was suggested because previous users who have searched red pepper chocolate have broadened their query to that term with success.

Refining

The refining method provides more specific alternatives to a user’s original query, helping users find products that better match their preferences. Often these suggestions go deeper into a search term.

For example, a user searches chocolate. The refining method would return refined search terms, such as dark chocolate and milk chocolate, allowing the user to jump to a search that returns more refined results for that original query.

Co-occurrence suggestions

The co-occurrence suggestions method identifies terms frequently searched alongside a user’s original query, leveraging behavioral and clickstream data across similar shopping journeys to surface related search terms.

For example, a user searches red wine. The co-occurrence method would return search terms often associated with the user’s original query, such as steak and cheese, because there is a link between these terms and the user’s original query. For instance, previous users who searched red wine then searched for steak or cheese.

Top groups suggestions

The top groups suggestions method suggests popular groups or categories related to a user’s query. It leverages behavioral and clickstream data to identify connections between the original query and other categories often visited by users who previously searched that query.

For example, a user searches chocolate. The top groups method would return terms such as confectionary and candy because previous users who have searched chocolate have also bought items from these categories.

How Related Search works

Related Search analyzes a user’s search query and generates a list of related terms that might interest the user. This is done using the different generation method discussed above.

Let’s walk through the process more in-depth.

1. User enters their search query.
   • A user enters their query in a search field and executes the search
2. Constructor generates related search terms.
   • Each of Constructor’s generation methods will generate search terms related to a query based on the nature of the method (e.g., expanding, refining, etc.). For example, the expanding method will generate a list of broader terms based on searches often related to the original query.
3. Constructor will rank each method’s suggestions and identify the best terms.
   • Constructor then ranks the related search terms generate by each method using clickstream data aggregated across all users. This helps identify which terms from each method are most relevant to the original query. The final list of returned terms will include a mix of terms generated by each method.
     Constructor updates related search terms for a query once a day, incorporating changes in users' behavioral data. For this reason, related searches for the same query are identical within a day.
4. Related search terms appear on the results page for the original search term.
   • Constructor’s Search API will return related search terms as part of the API response. Your team is responsible for determining how to render these results. Learn more in our Implement Related Search documentation.

How to get started

If you are live with Search and wish to enable Related Search, please reach out to your Customer Success Manager (CSM). Related Search is typically enabled via an A/B test.

Once enabled, Constructor takes a day to generate related search terms based on historical clickstream data and previous searches. After this time, Constructor will be able to send related searches back in API responses. Learn more in our Implement Related Search documentation.