Skip to main content
All CollectionsKnowledge
Understanding Confidence Scores: Elevating Answer Relevance
Understanding Confidence Scores: Elevating Answer Relevance
Written by Athul Rajeev
Updated over a week ago

At Docket, our mission is to deliver accurate and highly relevant answers tailored to your queries. In this guide, we'll delve into Confidence Scores and their pivotal role in ensuring answer relevance.

Understanding Confidence Scores:

Confidence Scores assess the certainty and reliability of answers provided by Docket, ensuring users receive reliable information for their queries.

Categorization of Confidence Levels:

Confidence Scores are segmented into three levels:

  • Low Confidence: Indicates relatively low reliability or certainty, prompting users to verify information from additional sources.

  • Medium Confidence: Represents a moderate level of certainty, suggesting the answer is likely accurate but may benefit from further validation.

  • High Confidence: Signifies a high level of reliability and certainty, assuring users of accurate and well-supported information.

Determining Confidence Scores:

Confidence Scores rely on various data sources, including:

  • Sales Enablement platforms like Highspot and Seismic

  • URLs

  • Documents, PPTs, and PDFs

  • RFPs

  • Call recordings from platforms like Gong and Chorus

  • Slack messages

  • Articles

  • Tickets from platforms like Intercom and Zendesk

Significance of Data Source Weightage:

Data source weightage guides Docket in delivering accurate and relevant answers. Prioritizing reliable sources ensures users receive trustworthy and tailored information.

Allocation of Weightage:

Weightage allocation considers factors such as reliability, comprehensiveness, and relevance. Sources known for their dependability, credibility, and depth of information receive higher priority.

Impact on Answer Relevance:

Incorporating data source weightage significantly enhances answer relevance. Prioritizing reliable sources ensures users receive precise and pertinent answers aligned with their queries, minimizing irrelevant or outdated information.

Future Directions:

Our journey towards precision-driven information retrieval continues. We aim to incorporate factors such as recency of information, user feedback, and involvement of subject matter experts to further refine our approach.


Confidence Scores are a valuable tool in our quest to provide accurate and relevant answers. By offering insight into information reliability, Confidence Scores empower users to make decisions confidently.

Did this answer your question?