Understanding the Coveo Relevance Maturity Model
The Coveo Relevance Maturity Model™ (CRMM) can help you visualize your journey towards predictive relevance:
Stage 0: Siloed and Federated Search
Siloed and federated search do not apply to a Coveo search solution. The Coveo engine uses end-users data to pertinently rank search results.
Siloed search solutions only offer single source search, which constrains users to access each silo of information separately to achieve their search tasks. In addition, a siloed search does not predict what users are searching for. Therefore, users have to explicitly request what they need to obtain quality results.
Federated search solutions allow users to access many sources of data to find the information they need. Even if this kind of solution interrogates several data sources, it fails to rank the search results according to relevance. This can be frustrating when facing a vast information ecosystem since relevant items can appear at the bottom of the search result list.
Stage 1: Secured and Unified Ranking
Like federated search solutions, search results in Coveo are consolidated from many sources. However, Coveo unifies and ranks these search results according to the sources’ access controls and follows security implementations (see Coveo Cloud V2 Management of Security Identities and Item Permissions).
Unified ranking ensures the most relevant results surface at the top by applying a ranking process that filters irrelevant items (see Index Ranking Phases).
Stage 2: Content Navigation
This implementation permits building tabs and facets with which your end users interact to narrow down the initial search results. Unified ranking combined with faceted navigation allows users to quickly take advantage of comprehensive results and easily find the answers they are seeking.
Stage 3: Tunable Relevance
At this stage, you are already providing information to your users based on their context. You are also gathering insights on the way they behave with your search solution (see Examining Usage Analytics Data).
Based on the data you analyzed from your users’ behavior, you will likely discover areas that may need improvement in your search solution. To overcome this lack of relevance, you can use the Query Pipelines page of the Coveo Cloud administration console, from which you can manually optimize your search results relevance and your search experience.
This powerful tool allows you to modify your end-user queries without having to use any code. From the Coveo Cloud administration console, you can create rules to improve the relevance of your search solution such as boosting specific items and creating synonyms (see Setting up Your Query Pipeline).
Stage 4: Contextual Relevance
At this stage, you are more than simply responsive to your end-user experience. In fact, Coveo Machine Learning (Coveo ML) now helps you to proactively act on your users’ queries by suggesting items in the result list based on the similarity of context.
In other words, the manual tuning you made to improve your search experience at Stage 3 is further refined by receiving additional contextual inputs at the point of the search query (see Coveo Machine Learning Features - Automatic Relevance Tuning (ART) Feature).
Stage 5: Contextual Suggestions
Contextual suggestions bring Coveo Machine Learning (Coveo ML) to another step. When reaching this point, a model can proactively provide information to users based on the understanding of what they are currently trying to achieve.
The main difference between the contextual relevance of Stage 4 and the contextual suggestions of Stage 5 is the predictive aspect. In other words, the capacity to evaluate probable future outcomes based on an understanding of the end users and their context.
The model now uses the search behavior of other users to predict the intent of a specific user by providing them with immediate relevant Query Suggestions (QS) (see Coveo Machine Learning Features - Query Suggestions Feature and How Does Intelligent Term Detection (ITD) Work?).
Stage 6: Self-Learning Predictive Recommendations
This is the stage where you get the most out of Coveo ML and maximize your positive business impact through a relevant search solution. At this point, you deliver an effortless search experience by providing your end users with key elements they need at a given moment, without them having to send further queries.
Coveo ML uses search navigation data to anticipate what a specific user will need next to accomplish their tasks by providing Event Recommendations (ER) (see Coveo Machine Learning Features - Event Recommendations Feature).
The next article in this guide outlines the steps to conduct a successful search project (see Search Project Overview).