Salesforce Knowledge Setting 'Suggest Related Articles on Cases' Not Returning Expected Results

This article explains how the two-step suggestion algorithm works, common causes for unexpected results, and configuration options that can improve article suggestion accuracy.

Step 1: Algorithm — Matching via Case Similarity

The first stage of the algorithm uses the Case Subject and the selected Case fields to search for similar Cases. If Knowledge articles are attached to those Cases through the CaseArticle object, those articles are returned as suggested results.

If the suggested articles appear irrelevant, add the Cases related list to your Knowledge article page layouts to review whether users are associating appropriate Cases with those articles.

If this search returns more than 10 results, the second stage of the algorithm is not executed.

Step 2: Algorithm — Direct Article Search

If the first stage does not return enough relevant Cases with attached Knowledge articles, Salesforce performs a second search using the same search terms directly against Knowledge article data.

Note: The search algorithms use caching to improve performance. If you recently changed your Knowledge configuration (such as adding searchable fields), users may not see updated suggestion results until the cache expires, which typically takes about one hour.

Workaround: Making Fields Searchable

If the selected Case fields are picklists or other non-searchable Knowledge fields, consider creating a searchable text field on the Knowledge article to store those values through automation.

For implementation details, see Make additional fields searchable.

Using Data Category Mapping as an Alternative

If the suggested articles are not sufficiently relevant, administrators can disable Suggest related articles on cases in Knowledge Settings. When disabled, Salesforce uses Data Category Mapping to determine suggested articles. If Data Category Mapping is not configured, the Case Subject is used to recommend articles.

Note: If Filter Articles with Data Category Mapping is enabled, those filters are applied after the initial pool of articles has been returned by the search algorithm.

When to Consider Einstein Article Recommendations

Configuring Case fields for article suggestions should be considered one part of an overall Knowledge search strategy rather than a complete solution.

For enhanced flexibility and more advanced recommendation capabilities, consider using Einstein Article Recommendations.

The quality of suggested articles depends heavily on how consistently users attach relevant Knowledge articles to Cases. As article associations improve over time, the quality of suggested article recommendations also improves.