CAF design-area weighting scheme (AB#6797)

The problem

Every caf.*.yaml rule file previously carried weight: 1.0. Get-Score computes the CAF framework score as a weighted average of each area's score, weighted by weight (src/assess/engine/Get-Score.ps1, AB#5087). A flat weight of 1.0 across all eight CAF landing zone design areas means Scout's CAF: Azure Landing Zone roll-up treats the Governance design area (10 verified recommendations) as carrying exactly as much weight in the overall CAF score as the Network topology and connectivity design area (123 verified recommendations, ~155 counting numbered task steps in pages mid-rewrite). Microsoft does not publish a numeric weighting for the landing zone design areas — the design area review process is a sequential checklist, not a scored rubric — but the volume of guidance Microsoft ships per area is itself a signal of how much of the overall "is this landing zone ready" question that area represents. An assessment that scores Governance and Network as equally weighty when Microsoft's own guidance is 12x larger for one of them produces a CAF score that does not track how much landing-zone risk each area actually carries.

The scheme

Weight each design area proportionally to its verified recommendation count, normalized so the mean weight across the eight areas is ~1.0 (so a roll-up computed before this change and one computed after differ only in how much each area's score moves the needle, not in scale).

weight(area) = recommendation_count(area) / mean(recommendation_count across all 8 areas)

Recommendation counts are the verified figures from pmo/audits/AZURE-SCOUT-AUDIT.md §8 Table 2, which were produced by fetching all 43 Microsoft Learn pages across the eight design areas on 2026-07-30 and counting top-level bullets under each page's ## Design recommendations heading (or the area-specific equivalent, e.g. "Management group recommendations"). That table also documents where the prior audit's figures were wrong (Security was double-counted against network-security pages it only links to; Management and Governance conflated "considerations" bullets with "recommendations" bullets) — this scheme uses the corrected numbers, not the prior ones.

Design area	Rule file	Verified recommendations	Mean (365/8 = 45.625)	Weight
Azure billing and Microsoft Entra tenant	caf.billing.yaml	42	45.625	0.92
Identity and access management	caf.identity.yaml	65	45.625	1.42
Resource organization	caf.resourceorg.yaml	35	45.625	0.77
Network topology and connectivity	caf.network.yaml	123 (formal)	45.625	2.70
Security	caf.security.yaml	45	45.625	0.99
Management	caf.management.yaml	15	45.625	0.33
Governance	caf.governance.yaml	10	45.625	0.22
Platform automation and DevOps	caf.platformauto.yaml	30	45.625	0.66

weight lives in each rule file's YAML header next to a one-line citation back to this document, so a future rule-count change (new rules added, or Microsoft republishing the design-area pages) has a clear place to recompute from.

What this deliberately does NOT do

• It does not claim to be Microsoft's weighting. Microsoft ships no landing-zone-review scoring weight; this is Scout's own derived scheme, built from a number Microsoft does publish (recommendation count per area) rather than invented from nothing. The audit and the citation comment in each rule file make this distinction explicit so a reader does not mistake Scout's weighting for an official Microsoft score.
• It does not change per-area scoring. Each design-area assessment (see the registry entries added under manifests/assessments.psd1 for AB#6797) scores only its own rules at 100% weight internally — the weighting only changes how the eight area scores combine into the single CAF framework number inside the CAF: Azure Landing Zone roll-up and inside Get-Score's Frameworks output.
• It is not a coverage weight. Scout's own rule-count-per-area (6-7 rules per file today) is intentionally NOT used as the weighting basis — that would reward files that happen to be under-built with a lower influence on the score, which is the opposite of what an honest "how much does this area matter" weighting should do. The weighting tracks Microsoft's guidance volume, not Scout's current implementation depth.

Reconciliation

Because Get-Score's framework score is computed from the area scores it weights (not scored independently), the CAF: Azure Landing Zone roll-up's CAF framework score is guaranteed, by construction, to equal the weighted average of the eight CAF: <design area> assessments' own framework scores when run against the same collect.json. tests/Assessment.Restructure.Tests.ps1 pins this with a non-vacuous test (break the weight, watch the reconciliation assertion fail, restore).