Merchandising with Coveo

This is for:


When we talk about search, we often hear the term searchandising, but Coveo powers the entire product discovery experience, not just search, so we refer to the actions to configure the whole experience as merchandising. We provide business users with all the control to influence and pick which products will be displayed on their website, in addition to leveraging AI models.

Understanding the following three key concepts are required to succesfully implement merchandising with Coveo:


For well-formed product discovery, your website will leverage three main interfaces: search, product listings and recommendations.

An end user’s experience with Coveo usually starts with keyword search. You’ll need to index products with Coveo to optimize the search results for the queries entered by end users in the search box. Triggering a search leads the customer to a search result page that contains facets (also known as filters), badging, and pagination or the more common infinite scroll. By default, the ranking of products returned by Coveo are sorted by relevance, meaning that it combines the rules you have established and the AI-driven scoring of the products.

Also, Coveo serves and reorders the content on all PLPs, the categories from your navigation menu, and special landing pages from campaigns or any pages that display products.

Coveo offers many recommendation algorithms for home page, cart, PLPs, product detail pages (PDP), and email marketing. From recently viewed, to more advanced statistical models such as people like you also viewed, we provide strategies that leverage product vector space to detect intents for one to one, anonymous personalization.

We also support extra functionalities like product badging based on real-time data, and inject personalized content like campaign banners and product slot overrides. Powering the entire product discovery with Coveo allows it to stay consistent and maximize the cross-pollination between experiences.

Coveo offers flexible search interface implementation solutions by supplying a library of Headless components that fit into any website. It can, however, also come packaged in an out-of-the-box (OOTB) framework called Atomic.


With product discovery in place, consider how you can control the products that are displayed. Merchandising goes beyond the hands-off approach that AI offers as it needs to account for inventory constraints, marketing campaigns, pushing collections, and sponsored brands and experimentation.

The main tools used in this instance are boosting or burying and pinned products. You can action each one at a global level for your site, or on any subset of keywords and pages, based on any attribute of the product. For example, you can boost a brand across a site, slightly favor private labels, bury products that are low in stock, and pin a product model at the top of a page. You can achieve this by creating the rules in the ranking tab in the query processing pipelines.

Other tools available to you include filter out or add products, redirect keyword searches to specific pages, and add special synonyms. For example, as an advanced user, you can combine variables in a rule to boost items in stock at the physical store that’s closest to your customer’s location.

Preparing and maintaining rules manually can only scale so much, therefore we recommend you keep them to the minimum required. Still, the grunt work of personalization will come from Coveo’s state-of-the-art AI. Learn more about AI solutions:

The following image shows how ranking can be adjusted to boost red socks to the top of the results page.

Boosting the ranking of red socks

Context information

Context information is any information captured about the user, such as where the request originates, and who is sending it. For example, a new customer, landing on a red Nike polo PDP, from a Google search on their mobile phone, in Florida during the summer.

We can deconstruct context into three groups of information: attributes of any anonymous visitors (device type, region, referrer URL), real-time data about the session (pages visited, items in the cart, previous products seen), and visitor history (account data, past purchases, number of visits, loyalty program status) that Coveo can stitch together based on collected data or that you can feed from your systems.

Next we have the where, and when. The product discovery content will obviously consider which interface requests products, from which website(s), and for which locale (language, country, currency). Coveo also supports the idea of time boxed campaigns, and grouping rules inside the same campaign bundle, which is useful to prepare a larger set of actions and apply them for defined periods.

These attributes are collected by your existing data layer, and can be leveraged to build rules and targeting conditions, but it’s also automatically optimized by Coveo’s AI models. Having healthy data is critical for accurate attribution metric calculation in the reporting of page and search performance.

Explore practical examples from a boat retailer for replacement parts, engineering, and demo websites such as the fashion store.

The following image shows how a change to custom context will change the results.

Modify the custom context to change results