{
  "grades": {
    "A": "Stated directly in RFI 36C10D26Q0163 (VA/VBA, 2026-07-16) — the primary source.",
    "D": "Modeling assumption, published as such. Not sourced to any document; chosen for calibration or normalization and open to challenge."
  },
  "parameters": [
    {"id": "automation_rate", "name": "Baseline end-to-end automation rate", "value": "≈60%", "grade": "A",
     "basis": "\"an overall automation rate of approximately 60% across all benefit chapters\""},
    {"id": "auto_days", "name": "Automated claim resolution time", "value": "≈1 day", "grade": "A",
     "basis": "\"Automated claims are resolved in approximately one day across all chapters\""},
    {"id": "manual_days", "name": "Manual adjudication time", "value": "9.5–18.7 days (by chapter)", "grade": "A",
     "basis": "\"compared to 9.5–18.7 days for manually adjudicated claims, depending on chapter\""},
    {"id": "auto_quality", "name": "Automation quality rate", "value": "95%+", "grade": "A",
     "basis": "\"overall automation quality rate of 95+%\"; modeled as 5% of automated claims offramping to the manual queue"},
    {"id": "annual_benefits", "name": "Annual benefits processed", "value": "$12B+", "grade": "A",
     "basis": "\"processes more than $12 billion in education benefits each year\" (context only; not a model input)"},
    {"id": "chapters", "name": "Benefit chapters administered", "value": "30, 33, 35, 1606", "grade": "A",
     "basis": "Title 38 chapters named in the RFI background; the model is chapter-agnostic because the chapter mix is not published"},
    {"id": "platform_inventory", "name": "Platform inventory", "value": "700+ repos, 50+ DBs, 76 EC2, 590+ artifacts, 30+ S3, 90+ licenses", "grade": "A",
     "basis": "RFI current-systems configuration (context for the migration questions; not a model input)"},
    {"id": "baseline_arrivals", "name": "Baseline arrivals", "value": "100 claims/week (normalized)", "grade": "D",
     "basis": "Normalization unit. The RFI publishes no absolute volumes, so the model publishes none; every backlog number is in these units"},
    {"id": "touch_days", "name": "Manual hands-on/in-process time", "value": "lognormal, median 13.0d, σ 0.45", "grade": "D",
     "basis": "Calibrated so the baseline scenario's manual median (14.1d) sits inside the RFI's stated 9.5–18.7d band; queue wait supplies the remainder"},
    {"id": "utilization", "name": "Baseline examiner utilization", "value": "90% of manual inflow", "grade": "D",
     "basis": "Staffing assumption defining capacity = 1.0; the capacity slider scales it 60–160%"},
    {"id": "surge_window", "name": "Enrollment-surge window", "value": "9 weeks (weeks 5–13 of the run)", "grade": "D",
     "basis": "Fall-enrollment seasonality is directionally well known in education claims; its magnitude here is a slider (×1.0–×3.0), not an estimate"},
    {"id": "arrival_process", "name": "Stochastic forms", "value": "Poisson arrivals, binomial automation split, lognormal touch time", "grade": "D",
     "basis": "Standard queueing choices; 200 Monte Carlo runs per scenario, fixed seed 36101263 for reproducibility"}
  ]
}