Introducing The Coveo Dynamic Navigation Experience Model

 

Coveo’s Dynamic Navigation Experience (DNE) model is a machine learning model that leverages usage analytics to interpret end user intent. It dynamically controls the user interface to adapt to the situation. Jason Mlyniec, Director for Customer Onboarding at Coveo, explains the four different components of the DNE model which include:

• Facet order on the page
• Ordering of facet values within each facet
• Ability to auto-select facets based on the query
• Usage analytics data that powers the model's learning

 

By leveraging user behavior data like facet selections and successful queries, the DNE model can dynamically adapt the search interface. This helps guide users to the most relevant content more efficiently, reducing customer effort and creating a more intuitive search experience.

 

How Does The DNE Model Work?

 

The DNE model works by learning from user behavior data collected through Coveo Usage Analytics. As Mlyniec demonstrates, when a user selects a facet filter to narrow down their search results, it's appended to the query expression in Coveo's Search API call. The model then analyzes this data alongside the query along with additional context like the search hub or tab, and the user profile. There are four aspects to the DNE model that makes it work. These include:

• Learning from user behavior data (queries, facet selections, successful searches, etc.)
• Analyzing query expressions and user context to understand intent
• Dynamically adapting the search interface based on this understanding of user intent
• Reordering facets and facet values to prioritize the most relevant options

 

Mlyniec walks attendees through a few prerequisites for implementing this model which are:

• Coveo Usage Analytics must be enabled to collect the necessary user behavior data
• The Coveo JavaScript Search Framework must be version 2.6063 or newer
• The search interface must include the Coveo Facet Component and Facet Manager Component
• A Coveo license that includes the DNE model functionality is needed

 

The DNE model leverages ML and usage analytics to continuously learn and adapt, which is what creates a more intuitive and efficient search experience for users.

 

Live Demo: How to Build The DNE Model

 

Mlyniec walks attendees through a live demo of how to build the DNE model, showcasing how easy it is to enable the model in the Coveo platform. As noted in parts 1 and 2 of our ML series, administrators will start building the model by navigating to the Machine Learning section in the Coveo Administration Console. To create a new DNE model, click the "Add Model" button and select "Dynamic Navigation Experience" from the list of available models. Next, configure the model settings:

1. Set the learning interval to determine how frequently the model refreshes and how much historical data it considers.
2. Choose the data sources the model should learn from, such as specific search hubs or tabs.
3. Decide whether to enable the "Facet Auto-Select" feature, which automatically selects the most relevant facets based on the user's query.
4. Create the model and associate it with a query pipeline.

 

Mlyniec notes that refresh rates will vary depending on your needs. For frequently changing content, like on an ecommerce site during peak seasons, a daily refresh rate ensures the most trending content surfaces. For more static content, a longer data cycle can be used for consistency. You’ll also need Coveo JavaScript Framework components like the Dynamic Facet and Dynamic Facet Manager in the search interface to leverage the DNE model effectively.

 

Putting Coveo’s ML Learning Models Together

 

Lipika Brahma, Customer Success Architect at Coveo, closes this three-part learning series by tying it all together. She showcases how different models and model combinations can be used to optimize search experiences for different contexts and audiences. She goes on to provide some tangibility to model combinations by presenting three specific use cases.

 

Use Case #1: ML Models for Your Customer Community

 

For a customer community, the primary goal is ensuring customer satisfaction, improving NPS, and reducing customer effort. Brahma explains how a combination of models can help achieve these goals.

 

Query Suggestions learns from successful searches. ART boosts relevant results. DNE adapts the search interface based on user behavior. Recommendations help customers find related content, while Smart Snippets give searchers quick answers directly in the search results.

 

Use Case #2: ML Models for Your Workplace

 

In a workplace environment, the focus is helping employees become proficient and giving them access to the resources they need to succeed. Brahma shows how Query Suggestions, ART, DNE, and Content Recommendations work together to guide employees to relevant information quickly.

 

They also learn from employee interactions, making the experience more relevant and individualized over time. The models work in tandem to boost the most useful content, which reduces the time employees spend searching for information (and also reduces frustration).

 

Use Case #3: ML Models for Your Ecommerce Website

 

For ecommerce websites, the main objectives are increasing conversions, average order value, and revenue. For this use case, Query Suggestions, ART, DNE, and Product Recommendations are appropriate models. Query Suggestions and ART put relevant products at the top of search results, while DNE adapts the search interface to guide users to the right products. Product Recommendations display complementary items based on shopper context.

 

Coveos' powerful ML models, when combined, help organizations create search experiences that are intelligent, intuitive, and customized to specific use cases. Our models work together seamlessly to understand user intent, produce the most relevant content, and adapt to user behavior.

 

Learn how to create an adaptive, user-centric search interface with Coveo's DNE model in this final installment of our three-part Mastering Machine Learning series. The final part of the series also demonstrates how Coveo's ML models can work together to drive a cohesive, intelligent, and efficient search experience across different use cases.