πŸ“Š Clow Studio β€” ARM Analytics

Adoption Β· Retention Β· Monetization for the AI agents ecosystem
β€” 30d ● READY dashboard β†’ admin β†’
Range

πŸ“₯ Adoption funnel last window

Every drop is a diagnosable pillar. Fuse by the weakest link, never the mean.

πŸ‘₯ User growth cumulative

Total registered users over time, plus the moving 7-day new-user rate.

πŸ€– Per-bot adoption active users

Where the ecosystem actually lives. Dark bars = zero-user bots β€” candidates for consolidation or killing.

πŸšͺ Acquisition channels first-touch

Where new users entered the ecosystem. Every unlabeled channel is a self-report gap.

πŸ” Weekly cohort retention % of cohort still active week N

A flat row is the win condition. A cliff after W1 means onboarding is a leak, not a funnel.
β‰₯50% 30–50% 15–30% 5–15% <5%

πŸ“ˆ DAU Β· WAU Β· MAU

Daily / weekly / monthly actives. Stickiness = DAU/MAU, shown below.

🩸 Churn & stickiness DAU / MAU

Churn is the inverse of retention. Stickiness <10% is a product nobody comes back to.

πŸ’° MRR & paid conversion

Monthly recurring revenue, with paid-conversion rate overlaid. Both pay rent.

🎯 Monetization funnel active β†’ paid

The conversion ladder. Each rung has its own failure mode β€” the weakest link is the one to fix.

🏷️ Tier distribution

πŸ’³ Payment channels

Mix is a resilience signal: a single provider is a single point of failure.

πŸ’Ά Revenue by bot

Revenue concentration β€” the Pareto tell.

πŸ“Ž Source Discipline one rule, four surfaces

Say where it came from, or do not say it. This is the only property a competitor can't copy by adding features β€” because it's a posture, not a feature. The four surfaces below each measure how well a bot output can be traced to a named input. The composite score is fused by the weakest surface (min, not mean) β€” the G18 discipline applied to its own metrics.
Signal source: audit log tags ([E]/[C]/[A]/[K], OBSERVED/INFERRED/UNKNOWN, /fuse pillar names, /paper traceability pass). Live wiring to /admin/data is next cycle β€” until then this panel shows the model, not the number.

πŸ›‘οΈ Audit Trail EU AI Act Β· 4-pillar coverage

Enterprise governance baseline. The four columns below mirror the four audit pillars named in GOVERNANCE_EU_AI_ACT.md. Backing data: audit/audit_client_views.sql (seven v_client_* views). Sidecar: audit/docker-compose.audit.yml β†’ Datasette on :8001.
⇣ agent-bom.json (signed) πŸ’³ BILLING.md πŸ›‘οΈ Audit pack spec
Shortcut to the raw views (admin): http://<host>:8001/clow/v_client_evoskillopt_inventory Β· …/v_client_qgate_results Β· …/v_client_breaker_events Β· …/v_client_audit_daily. Live wiring to /admin/data is next cycle β€” this panel shows the model and the daily roll-up shape.

🚦 Signals & recommendations

Composite answer fused by the weakest pillar β€” the 40 in a 95/40/85 split dominates.

πŸ† Leaderboard top activation

Activation = share of reached users who made it to 2nd session within 7 days.