Outputs that play like a playlist.

A normal dashboard hands you a wall of charts and leaves the thinking to you. The Player hands you a conclusion — and shows its work. Every card carries four things: the binding constraint, the evidence, a safety status (was the data safe to show, or too thin / too identifying?), and a recommended action.

Because the answers are precomputed, they're fast (no model spins up when you open a card), consistent (the same question gives the same answer for everyone), and auditable (you can trace each card back to the responses behind it).

This card is titled "High capability is not converting to quota performance in Enterprise Sales East." The dimension chip says Support — the diagnosed binding constraint. Note what it is not blaming: Capability sits at 82 pts, near the top. These people can do the job.

The trend line is the Support score over ten weeks: 78 → 58, a steady slide, not noise. Quota Attainment is 71%, with a −24 pt conversion gap — the distance between the performance the team's capability predicts and what's actually landing. Key drivers name the leaks — Deal desk responsiveness −22 pts, Manager escalation support −18 pts.

Safety status: Passed Min-N — 24 responses against a 7-response minimum, with 0 cuts suppressed. The number cleared the bar to be shown. Recommended action: run a 10-day support bottleneck review with Sales Ops, Deal Desk, and the regional manager.

The Player covers more than trends. A distribution card shows Customer Success splitting into two cohorts — a 50–60 pt group and an 80–90 pt group with a hollow middle. A comparison card puts East beside West across all four CAMS dimensions, where the Support gap (30 pts) dwarfs the Capability gap (4 pts). A heatmap shows all five sales teams at once — coldest cell East·Support, hottest West·Support.

The Library is your full catalog; a curated playlist like "Your team's binding constraint" sequences four cards into a narrative — the problem, comparative proof, a replication target, and the org-wide heatmap.

The cards are not a language model improvising on demand. They're composed from a psychometric substrate — protected-feedback responses, scored against measures drawn from peer-reviewed research — and only published after passing the Min-N safety gate that suppresses anything based on too few responses or too identifying. The model never invents a number; it reads the substrate, the same source every card cites.

The Insight Player and Library surfaces are built and rendering on seeded demo data today, replaying the range of pre-computed output types (trend, distribution, comparison, heatmap). When real customer data loads through the measurement pipeline, the same player plays back finished insights with no interface change.