Key Metrics - Documentation

Nebuly automatically extracts a set of metrics from every conversation between your users and your AI agent. Here is an overview of all the metrics available in the platform.

Metric	Definition
Topic	The high-level theme of a conversation. Topics are the broadest unit of classification and group together interactions around the same subject.
Action / Task	The specific task a user is trying to accomplish within a topic. Actions sit one level below topics and provide more granular detail on user intent.
Query	A concise summary of a single interaction between a user and the AI. Queries are the most granular unit and represent what the user asked or expressed in a specific exchange.
Sentiment	The overall tone of a conversation, classified across six levels: Very Positive, Positive, Neutral, Negative, Very Negative, and Mixed.
Emotion	The emotional state of the user, detected across 29 possible emotions for a more nuanced understanding of how users feel during interactions.
Conversation Length	The number of back-and-forth exchanges between a user and the AI agent in a single conversation.
Duration	The total time a user spends in a conversation with the AI agent.
Activation Rate	The percentage of eligible users who have actively used the AI agent in a given period.
Error Rate	The percentage of interactions where the user did not find a satisfactory answer, either because the AI failed to respond correctly or because the user expressed frustration with the response.
Business Risk	Interactions that need your team’s attention, spanning safety and compliance issues (PII exposure, toxic content, policy violations) and cases where users express complaints or dissatisfaction directed at your company (not the AI itself).
AI Fluency Index	A measure of how effectively users are able to interact with the AI agent — capturing the quality of their prompting and their ability to get value from the AI.
ROI	The estimated return on investment generated by the AI agent, calculated automatically based on standard assumptions.
Hours Saved	The estimated number of hours saved by automating interactions through the AI agent, calculated automatically based on standard assumptions.