1. Dataset Name

companion_funnel

2. Description

A stage-level funnel dataset that tracks each AI Companion session's progress through four milestones: Scheduled to Join, Joined, Recorded, and Feedback Generated. Each row represents one companion session at one stage it reached, including the stage timestamp, the next stage reached, and the time spent at the current stage. A session only appears at a stage once all prior stages have been reached. This dataset powers the AI Companion Funnel visualization, showing where sessions progress, stall, or drop off.

3. Source Tables

• customer_views.companion_session (sole source table — all four stages sourced from different column conditions on the same table)

4. Grain

One row per companion_session_id per funnel stage reached. A session that progressed through all four stages will have four rows. The LEAD/LAG window functions provide next-stage context on each row so the full transition path is visible without self-joining.

5. Dashboards Supported

• AI Companion Dashboard

6. Metrics Supported

7. Known Exclusions / Filters Applied

• Stage 1 (Scheduled to Join): latest_bot_status IS NOT NULL AND latest_bot_status != 'deleted' AND companion_scheduled_to_join_at IS NOT NULL — sessions with a deleted bot status are excluded entirely.
• Stage 2 (Joined): was_companion_dispatched_to_join = TRUE.
• Stage 3 (Recorded): has_companion_recorded = TRUE.
• Stage 4 (Feedback Generated): status = 'feedback_generated'.
• Sequential gate rule: a row is only included in the final output when stage_group_index = 1 (the first stage) OR prev_stage_group_index = stage_group_index - 1, ensuring strict step-by-step progression — a session cannot appear at Recorded without first appearing at Joined.
• scheduled_at is computed as MAX(stage_ts WHERE event_name = 'companion_scheduled') windowed per companion_session_id, so every stage row for a given session carries the same scheduled timestamp.

8. Schema