Glossary

Glossary of terms used in the Maven dashboard.

Term	Description
Conversation	A full message thread including messages from end users, human agents, Maven responses, or Maven suggested responses.
User Interactions	The number of individual times an end user interfaces with Maven via a message within a conversation.
Count	The number of conversations as defined above
Inserts	The number of Conversations where at least one Maven suggested response has been inserted.
Coverage	The percentage of conversations Maven has assessed it could resolve
Handle Time	The length of time the conversation lasted. (time of the last message - time of the first message)
Response Time	The length of time it took for the first user message to get a response. (time of the first agent message - time of the first message)
Customer Sentiment	The user sentiment during a conversation as determined by Maven. Either positive, negative or neutral.
Maven Quality	How well Maven was able to answer questions within a conversation. A conversation can be considered Good, Needs improvement or Ineligible. See below for more details.
Category	Each conversation has exactly one category, determined by Maven.
Tag	Conversations can have multiple tags. Tags are supplied by external integrations. (e.g. a Maven App like Zendesk or Freshdesk)

Quality

Maven Quality is a measurement of Maven’s confidence in its own answer. There are three categories of quality:

Good: Maven’s self-assessment is that the answer was both correct and of high quality
Needs Improvement: Maven could not confidently answer the question due to a lack of knowledge or because the end user gave the interaction a thumbs down.
Ineligible/Pending: Ineligible conversations are those where the user cancels the conversation mid-answer or a question was considered an attack attempt and was blocked.

How Maven Quality is Calculated

Maven Quality takes the following signals into account:

Automatic self grading: Maven has a self-grading mechanism that determines how confident it is based on how much relevant information it has in knowledge and actions
Automatic grading against human responses: where human agent responses are available they can be compared to Maven’s own response
Human feedback: user provided thumbs up and down feedback serves is a strong signal of quality