What patterns work for which property profiles.
Anonymised cross-customer aggregates updated nightly. Every signal is computed at the network level — no individual customer's data is visible to any other customer. Methodology, threshold, and qualifying-tenant count are published here so you can audit the numbers before relying on them.
By property profile
Acceptance + override + net-autopilot rates grouped by the property's profile code (Independent, Urban Branded, Resort, Limited-Service, etc.). Each row averages across all qualifying tenants in that profile.
| Profile | Properties | Mean acceptance % | Mean override % | Net-autopilot % |
|---|---|---|---|---|
| No profile signals yet — first dimension to populate as tenants accumulate. | ||||
By demand segment
Per-segment acceptance + override patterns across the 8-segment forecast decomposition (Corporate, Leisure, Group, OTA, Direct, Government, Negotiated, Other). Shows where the engine produces high-trust recs vs where operators routinely override.
| Segment | Properties | Mean acceptance % | Mean override % |
|---|---|---|---|
| No segment signals yet — populates once the per-recommendation applied_segment column is backfilled. | |||
Override-reason clusters
Operators give a structured reason on every manual override (14-code taxonomy from Phase 3.0.1). The mean override delta — how far the operator moved the rate from the recommendation — tells us where the engine systematically under- or over-shoots, by reason.
| Override reason | Observations | Mean rate delta % |
|---|---|---|
| No override clusters yet — populates after a tenant accumulates ≥10 overrides of a given reason. | ||
Methodology
Cross-tenant aggregations recompute nightly via the same overnight pipeline that regenerates per-property recommendations. For each dimension, we compute a per-tenant statistic over the lookback window (default 90 days), then average across all qualifying tenants. Published only at sufficient sample size, and aggregated so no individual customer can be identified from the signal.