Pattern recurrence by category across the fleet. Dining and Wayfinding are the highest-volume; AI Quality patterns surface knowledge gaps that feed back to the content team.

• Dining — 47 occurrences · wine-list legibility · $11K/sailing
• Wayfinding — 38 occurrences · welcome cocktail signage · +29% redemption
• Operations — 31 occurrences · tender backup · $8K excursion uplift

Five recovery stories with attribution-modelled lift. Each links a specific moment to a specific action to a measurable outcome.

• +11% bar revenue — Wine-list paper backup (23 min)
• 62% → 91% redemption — Welcome cocktail signage
• 6/10 → 9/10 NPS — Top-tier loyalty intervention ($40K LTV retained)

Fleet benchmarks across all 5 active sailings, with NPS proxy, validator pass, time-to-close, and recovery rate. Useful for spotting regional or sailing-stage variance.

• SLG-0430 Transatlantic — NPS 84, validator 99.9%, TTC 48m
• SLG-0428 Mediterranean — NPS 78, validator 99.8%, TTC 1h 18m
• SLG-0427 Caribbean — NPS 62 (low) · needs attention

Knowledge gaps detected by the AI quality layer and routed to the content/ops team for source-system fixes. 389 guests benefited from this week's closures.

• Tender berth congestion Port 2 — 4 sailings · 188 guests · port renegotiation
• Specialty Italian dress code — 4 sailings · 89 guests · CMS rewrite
• Spa couples-massage pricing — 3 sailings · 47 guests · tablet + POS

Validator pass rate, citation rate, and avg latency tracked over 7 days. No degradation detected. Sonnet generates; Opus validates relevance + precision; every response source-cited.

• Validator pass — 99.6% (range 99.3–99.6)
• Citation rate — 100% (range 99.8–100)
• Avg latency — 2.4s (range 2.4–2.6 · target ≤ 3.0)