Hotel Comp Policy Model | Technical Evidence

Policy comparison uncertainty

Paired case-bootstrap intervals, probabilistic guardrail checks, cost uncertainty, and policy-selection stability.

Evidence boundary: Operational records and outcomes are synthetic. Public context establishes property fit and guest-facing value, not internal policy, costs, or performance.

This report tests whether the simulated policy decision remains credible when the case mix is resampled and uncertain cost/fit assumptions are varied. It does not validate guest outcomes, estimate causal effects, or project Proper Hotels savings.

Generated Shadow-Validation Decision

Approve a four-week, minimum-50-case shadow validation of Guardrailed recovery as the leading candidate. Under the declared synthetic case mix and policy assumptions, it cleared the guest-protection, data-quality, escalation, and operational guardrails with the lowest modeled cost among eligible policies.

Shared-World Assumption Stress

PolicyAssumption-stress pass rateSelection frequencyModeled cost P05P50P95
Synthetic discretionary baseline0.0%0.0%$19,109$19,109$19,109
Tiered standardization0.0%0.0%$22,884$25,914$29,379
Guardrailed recovery99.6%99.6%$27,342$30,467$33,944
Recovery first100.0%0.0%$41,818$51,114$60,481
Intelligent generosity100.0%0.4%$33,796$40,114$46,517

The assumption stress test applies one coherent set of recovery-weight, fit, occupancy, and gesture-cost assumptions to every policy in each draw. It then recalculates policy metrics, reapplies all shadow-validation guardrails, and reruns the selection rule.

Paired Case-Bootstrap Intervals

PolicySafe recovery path95% intervalHigh-risk under-recovery95% intervalManager review95% interval
Synthetic discretionary baseline44.7%38.6%-50.8%67.6%58.7%-76.1%UnknownUnknown
Tiered standardization91.7%88.9%-94.2%21.5%15.4%-28.1%27.2%23.0%-31.4%
Guardrailed recovery100.0%100.0%-100.0%0.0%0.0%-0.0%27.4%23.3%-31.6%
Recovery first100.0%100.0%-100.0%0.0%0.0%-0.0%63.3%58.6%-67.7%
Intelligent generosity100.0%100.0%-100.0%0.0%0.0%-0.0%57.7%53.0%-62.3%

Case IDs are resampled once per bootstrap draw and applied to every policy. This paired design preserves case-level comparability and quantifies sampling uncertainty without inventing additional hotel outcomes.

Interpretation