Research question and scope
Published September 9, 2026.
This study asks how often sampled support decisions cite or reflect a policy version that was not applicable to the underlying event. It covers selected policy families, channels, and decision dates. It does not determine legal compliance or the fairness of the policy itself.
Methodology
Create a version ledger with effective dates, superseded dates, scope, jurisdiction, and approved exceptions. Sample completed cases around policy transitions and stable periods. Two reviewers independently identify the relevant event date, applicable version, guidance accessed, decision, and any authorized exception. Preserve uncertainty where the case record cannot support a conclusion.
Measures and analysis
Report confirmed version mismatches, indeterminate cases, stale-link exposure, unsupported exceptions, and customer recontacts. Separate errors caused by knowledge publication, search ranking, macros, training, and individual selection. Compare transition windows with stable periods descriptively and publish denominators.
Limitations and inference limits
Case notes may not reveal which source an agent used. Policies can overlap or contain legitimate judgment. Observed association between stale guidance and a changed outcome does not prove causation. Results cannot be generalized beyond the audited policy families and dates.