Interactions
An Interaction represents a single exchange between a user and your AI application, typically consisting of one user message and one AI response. It is the smallest unit of data stored in Nebuly and the foundation for all analytics. Every interaction sent to Nebuly is automatically enriched with AI-generated insights, allowing you to understand not only what users said, but also what they were trying to achieve, how they felt, and whether the interaction was successful. Read more about interactions here. Interactions are ideal when investigating individual events, understanding failures, or reviewing specific conversations.Conversations
A Conversation groups together one or more interactions that belong to the same session or thread. While interactions provide a detailed view of individual exchanges, conversations give you the broader context of the entire user journey. Nebuly aggregates interaction data to compute conversation-level metrics and insights, making it easier to understand whether a user ultimately achieved their goal. Read more about conversations here. Conversation-level analysis is particularly useful when measuring task completion, identifying friction across multiple turns, or understanding the overall quality of an AI experience.Users
A User represents a person interacting with your AI product. Users are identified by a stable user ID provided during data ingestion, allowing Nebuly to connect all of their conversations over time. Looking at data from the user perspective makes it possible to move beyond individual conversations and understand long-term behavior, adoption, retention, and engagement. Read more about users here. User-level analysis is particularly valuable for understanding how different customer segments interact with your AI product and how their behavior evolves over time.User groups
User groups are collections of users who share common characteristics or behaviors. They allow you to analyze and compare specific populations instead of individual users. Groups can be created using user attributes, custom tags, or behavioral criteria, making it easy to answer questions such as:- How do enterprise customers use the assistant compared to free users?
- Are employees from different departments asking different questions?
- Which customer segments experience the highest error rates?