Build search

In this section, we show how practical site search means better viability for returned products, which allows customers to quickly find the things they want. Once your products are inside the index and the Coveo commerce catalog has been created, you can surface these products in a search interface. Applying several recommended configurations to boost your commerce relevance will translate into higher conversion rates, as customers who find what they search for are more likely to make a purchase, and learning from search terms is critical for merchandising and recommendations.

Integrate a search page

Integrating a Coveo-powered search interface into your website allows users to retrieve and view results from your index. The leading practice is to use the Coveo Atomic library, which provides a simple way to implement a full-featured search interface that includes machine-learning-ready usage analytics (see Integrate a search page into your ecommerce solution and Query your commerce catalog content).

Depending on your use case, you may want to consider using the Coveo Headless library.[1]

1. When using Atomic or Headless isn’t an option, developers can achieve the same results using the APIs, but with significantly more work.

Leverage variants and availabilities

For catalogs with product variants, or availability constrains, you’ll want to use the Coveo catalog for variant attribute filtering. This part of the process considers proper facet association, how to properly query your commerce catalog content and product filters for a specific channel. Once all the above factors have been properly managed, the search interface attached to your query pipeline will only return products, but variant attributes will be available to create additional facets (see Leverage variants and availabilities).

Manage relevance

Your first step is to get familiar with query pipelines. We recommend that you create a dedicated pipeline for your search page using conditions that will apply to your specific use case. You should avoid modifying the default query pipeline, as it will serve as an empty state, which is useful for debugging purposes. Apply default Coveo Machine Learning (Coveo ML) models to the query pipeline, such as Automatic Relevance Tuning (ART), Query Suggestions (QS), and Dynamic Navigation Experience (DNE) (see Coveo ML).

Once you’ve understood the basics of setting up a query pipeline, here are some additional configurations to help your commerce relevance.

Personalize experiences

To offer the best search experience we’ve leveraged the idea of product embeddings and vectors based on cosine similarities. The ecommerce-focused ML model that will benefit from this is Predictive Query Suggestions (PQS).

External search engine optimization

For a high-level overview of best practices to ensure that your content can be properly indexed by external search engines (SEs) refer to our External search engine optimization documentation.