# Validation report

## Overall assessment: Ready to share with explicit caveats

The calculation package is internally consistent, source-traceable and reproducible. It is ready to support a transparent public article as a **council-performance model**, not an official or objective ranking of the best places to live.

## Calculation spot-checks

- Verified hashes for 11 frozen source artifacts.
- Canonical joins preserve 32 unique councils and the selected data is one row per council-indicator-period.
- Recomputed every headline score from the five published pillars; maximum rounding error 7.5e-07.
- Spot-checked higher, lower, zero-target and 100%-target ranking directions against raw source values.
- Reconstructed the waste value-for-money family from four indicator scores; maximum error 4.9e-07.
- Robustness ranks and probabilities are internally coherent across 10,000 simulations.
- The central scenario reconciles exactly, and East Renfrewshire remains first in all ten explicit scenarios.
- Resident-verdict table contains five complete 2024 questions and reconciles to its published average.
- Standalone landlord and place-context tables preserve their eligible populations and expected suppression.
- Confirmed 53 distinct headline LGBF measures, four additional landlord measures, and no context-only leakage.

## Required caveats

- The Local Government Benchmarking Framework is a diagnostic framework, not an official league table.
- Most measures are 2024-25; complete adult-care and roads measures are 2025-26; satisfaction estimates can be multi-year rolling survey figures.
- Percentile scores show relative position among 32 councils, not a pass/fail standard or the absolute size of a gap.
- Adult care and school outcomes are shared with the NHS, households and national systems; the score does not establish causation.
- Costs are paired with outcomes but still reflect geography, need, service mix and accounting choices.
- Missing/suppressed cells receive a neutral 50 and lower coverage; modelled source values are flagged.
- Place conditions, Crisis Grant acceptance, budget gaps, accountability-record counts and council-landlord metrics are not folded into the headline score.
- The 10,000-run probabilities measure stability across plausible modelling choices, not the probability a council is objectively best.

## Remaining editorial judgement

Before publication, the article headline, standfirst and chart annotations must retain those distinctions. Small survey differences should not be written as statistically significant unless confidence intervals support that claim.