Companion Feedback metrics
1. Dataset Name
companion_feedback_metrics_dataset
2. Description
A session-level dataset that exposes companion session records for feedback generation volume, coverage, and time-savings analysis. Each row represents one companion session, with status indicating its lifecycle outcome. Sessions that reached feedback_generated status are the core population for feedback output metrics. The dataset also carries scheduling context (companion_scheduled_to_join_at), join/recording flags, and conference platform information. This dataset powers companion feedback generation counts, coverage percentages, and the estimated interviewer time saved metric.
3. Source Tables
- customer_views.companion_session (sole source table — no joins)
4. Grain
One row per companion session (id). Each session represents a single AI Companion engagement for a specific interview slot, candidate, and position.
5. Dashboards Supported
- AI Companion Dashboard
6. Metrics Supported
| Metric | Supported Breakdowns |
| Companion Feedback Generated | No drilldowns |
| Companion Feedback Generated % | No drilldowns |
| Estimated Time Saved by Companion | No drilldowns |
7. Known Exclusions / Filters Applied
- Companion Feedback Generated and Estimated Time Saved use status = 'feedback_generated' as their condition. Sessions at earlier stages or that did not reach feedback generation contribute zero.
- Companion Feedback Generated % uses COUNT(*) as the denominator — all sessions regardless of status or type are included in the base, making this a coverage metric across the full session population.
- Estimated Time Saved applies a fixed 10-minute assumption per feedback_generated session — this is an estimation, not a measured value, and is displayed with a '~' prefix to reflect its approximate nature.
- All three metrics filter by companion_scheduled_to_join_at for date-range scoping.
8. Schema
| Column Name | Data Type | Notes |
| id | bigint | Primary key. Unique companion session identifier; sourced from companion_session |
| uce_id | bigint | Foreign key to the user calendar event; sourced from companion_session |
| slot_id | varchar | Identifier of the interview slot within the calendar event; sourced from companion_session |
| position_id | bigint | Position associated with the session; sourced from companion_session |
| profile_id | bigint | Candidate's profile identifier; sourced from companion_session |
| status | varchar | Session lifecycle status; sourced from companion_session. Key value: 'feedback_generated' — used as numerator condition for feedback generation metrics. |
| latest_bot_status | varchar | Most recent bot status; sourced from companion_session. Used in the funnel dataset to exclude deleted sessions. |
| companion_scheduled_to_join_at | datetime | Timestamp when the companion was scheduled to join the interview; sourced from companion_session. Primary date field for time-based filtering. |
| was_companion_dispatched_to_join | bigint | Flag (0/1); TRUE when the companion was dispatched to join the call; sourced from companion_session |
| has_companion_recorded | bigint | Flag (0/1); TRUE when the companion successfully recorded the session; sourced from companion_session |
| conference_platform | varchar | Video conferencing platform used for the interview; sourced from companion_session |
| interview_language | varchar | Language of the interview; sourced from companion_session |