Troubleshoot Query Suggestions (QS) models

This article describes how to resolve a Coveo Machine Learning (Coveo ML) Query Suggestion (QS) model that shows no learned suggestions or appears empty.

QS model shows no suggestions or appears empty

Symptoms

Cause

A QS model learns only from queries that were performed and followed by a click event on a search result. The model needs enough of this search activity to produce suggestions, approximately 10,000 search events. Low-traffic and non-production search hubs accumulate qualifying queries slowly, so a new or low-traffic model can show few or no suggestions until more usage data accumulates over the model’s data period.

Solution

  1. Confirm that the model has had time to gather data across its data period.

  2. From the Reports (platform-ca | platform-eu | platform-au) page, review query suggestion candidates to check whether the search interface has enough qualifying queries before you enable the model. See Reviewing query suggestion candidates for details.

  3. In a test or low-traffic environment, pre-populate suggestions by configuring a Default Queries file through the Advanced Model Configurations API. See Pre-populating query suggestions for details.