SAMPLE AEO/GEO DATASET PACKAGE
All brands, answers, classifications and source records are fictional and locally generated.
No live engine queries, real client data, current product/policy conditions, measured improvement or benchmarks.
Relabelled consistently from an existing fictional local package; IDs and embedded references changed, counts/classifications unchanged.
480 attempts; 466 valid; 14 capture failures; 183 claims (137 assigned CORRECT, 27 INCORRECT, 19 UNVERIFIABLE); 12 fictional source records.
All valid answers are recommendation-eligible. Visibility denominator: valid answers. Accuracy denominator: CORRECT+INCORRECT. Coverage denominator: all claims.
Only four catalogue sources are linked to citations/claims; the other eight are inventory-only in this sample.
Blank Source_ID remains intentional on the 19 UNVERIFIABLE claims. No original prompts/personas or dated snapshots exist.
No trend, engine comparison, combined visibility score, improvement or causal attribution may be derived.
Read SAMPLE_METRICS.json for per-sector numbers. Proposed report actions and their NOT_RUN prompts are editorial illustrations.

ACTION PLAN FILES
This package contains exactly seven files: README, RUNS, CLAIMS, SOURCES, METRICS, SAMPLE_ACTION_PLAN.json and SAMPLE_ACTION_PLAN.csv.
JSON includes12 action briefs, role owners, evidence references, sample FAQ copy, proposed NOT_RUN prompts, acceptance checks, planning estimates and30/60/90-day timing. CSV contains12 concise action rows.
ALL ACTIONS AND RATINGS ARE EDITORIAL SAMPLE ILLUSTRATIONS, not real client findings, actual product/policy conditions, completed tests, observed outcomes, ranking uplift, ROI or incurred effort. FAQ copy requires accountable owner-approved actual sources before live publication.
Source_IDs identify existing fictional evidence links; Catalogue_target_source_IDs are proposed catalogue targets, never imputed sources for unknown claims. Record IDs join the accompanying SAMPLE data files.
Locally defined HolistiCRM ordinal prioritisation rubric (not industry certification):
Score=100*[0.30*(I-1)/4+0.25*(U-1)/4+0.20*(R-1)/4+0.15*(C-1)/4+0.10*(5-E)/4].
I=impact, U=urgency, R=reach, C=evidence confidence, E=effort. Each rating1-5; one-decimal HALF_UP rounding. Effort uses the upper person-day estimate; actual remains unknown.
Evidence confidence is capped at3/5 because this dataset is fictional. Score is priority, not probability or platform visibility.
P1:score>=70,day30; P2:score>=50and<70,day60; P3:score<50,day90.
P0 overrides priority independently of score for confirmed material-condition errors. INS-01/BANK-01 are fictional exclusion/eligibility workflow examples, NOT live incidents. A real P0 needs an accountable domain reviewer, a complete captured answer and dated approved source. Sample triage targetdays1-7.
There is one undated synthetic snapshot. No trend, causal improvement or vendor/model-run comparison is supported.
