Standard Dimensions
The full list of Dimensions Dimension Labs has already written, at session and section level, and how to get them run on your data.
Standard Dimensions are a set of general-purpose Dimensions that Dimension Labs has already written. They work on most kinds of text and are most useful on conversations, where they cover both the whole conversation and the parts within it. In the platform they are called System Prompts, and they are listed on the Prompts tab alongside your own prompts. The session-level standard Dimensions are stored in the system_session_prompt table.
Availability
Section titled “Availability”Standard Dimensions are available to every workspace, with nothing to configure. They run every day only when Dimension Labs turns on daily enrichment for your workspace.
- To apply them to conversations you already have, or to a file upload, run a backfill from the Backfill tab.
- To have them run on new conversations every day, email [email protected] and ask for daily enrichment.
See When enrichment runs for the full schedule.
Two levels of analysis
Section titled “Two levels of analysis”| Level | What it reads | Good for |
|---|---|---|
| Session analysis | The whole transcript or document | Reasons, categories, summaries, outcomes and predicted feedback for each conversation |
| Section analysis | Each section within a transcript or document | How a conversation moves from topic to topic, repeats, and where automation could take over |
Use both together: session analysis for the trends, section analysis for what happens inside the conversations behind them.
Dimensions marked (conversational) below are designed for dialogue, such as chat, call and bot transcripts. On reviews, posts and survey answers they may have little to read.
Output types
Section titled “Output types”| Output type | Meaning |
|---|---|
| Dynamic label | Text the model writes from what the conversation says, with no fixed list |
| Fixed category | A value from a fixed list |
| Score | A whole number on a fixed scale |
| Boolean | true or false |
These are the same output types used for custom prompts.
Session analysis Dimensions
Section titled “Session analysis Dimensions”| Dimension | What it gives you | Output |
|---|---|---|
| Reason | A concise label for the main topic or intent of the full session transcript | Dynamic label |
| Category | A business-level category for the reason or main driver of the session | Dynamic label |
| Effort | A score from 1 to 5 predicting how hard it was for the user to complete their task, 5 being the most effort (conversational) | Score |
| Helpfulness | A score from 1 to 5 predicting how helpful the agent assisting the user was, 5 being the most helpful (conversational) | Score |
| Predicted Feedback | Simulated user feedback summarizing their satisfaction with the request or experience | Dynamic label |
| Feedback Cluster | Predicted feedback grouped by similarity | Dynamic label |
| Predicted Solution | A proposed improvement or fix that would make the experience better next time | Dynamic label |
| Solution Cluster | Predicted solutions grouped by similar improvement themes | Dynamic label |
| Agent Sentiment | The overall tone or emotion of the outgoing side (your agent or bot) | Dynamic label |
| Agent Sentiment Cluster | Agent sentiment grouped into positive, neutral and negative | Fixed category |
| User Sentiment | The overall tone or emotion of the incoming side (the customer) | Dynamic label |
| User Sentiment Cluster | User sentiment grouped into positive, neutral and negative | Fixed category |
| Outcome | The predicted end state of the interaction: complete, abandoned or escalated (conversational) | Fixed category |
| Predicted Rating | A score from 1 to 10 predicting the user’s satisfaction rating | Score |
| Predicted Product | The product or service mentioned, if any | Dynamic label |
| Recommendation | A recommendation the user made to improve a product or service | Dynamic label |
| Recommendation Cluster | User recommendations grouped by similarity | Dynamic label |
| Predicted Success Score | A score from 1 to 100 estimating the percentage of the request that was handled successfully (conversational) | Score |
| Summary | A short but detailed summary of the full session transcript | Dynamic label |
| Summary Cluster | Session summaries grouped by similarity | Dynamic label |
Predicted Rating is the general-purpose satisfaction score. Predicted scores explains what it measures, what it doesn’t, and the more specific scores you can add.
Section analysis Dimensions
Section titled “Section analysis Dimensions”| Dimension | What it gives you | Output |
|---|---|---|
| Reason | A concise label for the main topic or intent of a section of the transcript | Dynamic label |
| Category | Section-level reasons grouped by similar underlying themes | Dynamic label |
| User Sentiment | The tone or emotion of the customer’s messages in a section (conversational) | Dynamic label |
| User Sentiment Cluster | Section-level user sentiment grouped into positive, neutral and negative (conversational) | Fixed category |
| Agent Sentiment | The tone or emotion of the agent’s or bot’s messages in a section (conversational) | Dynamic label |
| Agent Sentiment Cluster | Section-level agent sentiment grouped into positive, neutral and negative (conversational) | Fixed category |
| User Reason | The main topic or intent the customer expressed in a section (conversational) | Dynamic label |
| User Explanation | A short explanation of the User Reason label for a section (conversational) | Dynamic label |
| Agent Reason | The main topic or intent the agent or bot expressed in a section (conversational) | Dynamic label |
| Agent Explanation | A short explanation of the Agent Reason label for a section (conversational) | Dynamic label |
| Predicted Intent | A suggested automation intent inferred from a section, for example one a bot could handle (conversational) | Dynamic label |
| Product | The product or service mentioned in a section, if any | Dynamic label |
| Summary | A concise summary of the content and purpose of a section | Dynamic label |
| Repeat | True if the section repeats or comes back to an earlier topic in the transcript | Boolean |
| Specificity Score | A score from 1 to 5 for how detailed and actionable the content of a section is | Score |
“Incoming” and “outgoing” here mean the same as the incoming field in your data: incoming is the customer or end user, outgoing is your agent, bot or company. See Conversation format.
Using your own categories
Section titled “Using your own categories”Category is written by the model from what it reads. If you need your own taxonomy, product lines or journey stages, write a custom prompt with a fixed category list. Standard and custom Dimensions can run on the same data, and the agent can cross them: your category against the standard Predicted Rating, for example.