Open methodology · v1.0 · Updated May 4 2026

How Livio Grid produces a review.

Every finding traces to a specific engineering reference. The methodology is published, versioned, and revisable. We tell you when we're confident — and when we're not.

The pipeline, end to end.

  1. 1 · Parse the proposal

    Three-layer parser: regex/heuristic patterns for proposal-specific phrasing, LLM-based parameter extraction (Claude Opus 4) for free-form text, and reconciliation that prefers the higher-precision source.

  2. 2 · Resolve the site

    Geocoder lookup against Open-Meteo (free, global, no key). Mandatory site-research pre-pass: ASHRAE design-day temperatures, elevation, seismic design category, FEMA flood zone, climate classification.

  3. 3 · Run first-principles engine

    Same V3 engine that ships hyperscaler-grade engineering deliverables. Climate-aware cooling architecture decision, four-heat-path cooling load split, NEC/NFPA/IBC compliance verification across 41 gates.

  4. 4 · Cross-check ISO data

    FERC OASIS interconnection-queue publications, utility integrated resource plans, current commodity-catalog lead times. Two-source cross-validation on every claim.

  5. 5 · Compute deltas

    Pure function: claim vs first-principles for every parameter. Severity classification: pass / warning / critical / opportunity. Confidence intervals on every numeric finding.

  6. 6 · Render the report

    Headline finding, deliverability score, red/green flags, recommended next steps, 12 detailed sections, 30-row claim-vs-reality table, methodology citations.

Code & standards cited.

Every finding in a review traces to one of these. If a specific NEC/NFPA/IBC section is not in this list, no review will cite it.

§ NEC 110.26(C)(2)Working space width ≥ 2,440 mm in front of switchboards/switchgear with OCPDs ≥ 1,200 A
§ NEC 110.26(A)(1)480V Cond. 2 working depth ≥ 1,067 mm
§ NEC 110.26(C)(3)Two means of egress at room ≥ 3.7 m + 1,200 A
§ NEC 450.27Outdoor transformer ≥ 3,000 mm from combustible building openings
§ NEC 645.10EPO at each principal exit door of any EDP room
§ NEC 230Service entrance conductors and switchgear
§ NEC 702Optional standby systems
§ NFPA 855 §6.3.3BESS aggregate energy per fire compartment
§ NFPA 855 §9.10.336 CFM/cabinet purge for BESS
§ NFPA 13 OH-2Pre-action sprinkler in INFRA bay (open-head double-interlock)
§ NFPA 37Genset exhaust ≥ 6,000 mm from any building opening
§ NFPA 2001 §5.5.1Listed clean-agent system sizing formula
§ NFPA 75Information technology equipment — fire protection
§ NFPA 110Emergency and standby power systems
§ IBC §707Fire-rated assembly between fire compartments
§ IBC §71690-minute fire-rated cross-door assemblies
§ IBC §1010.1.4.4Coiling shutter as non-egress opening
§ IBC §1006Number of exits and exit access doorways
§ ASHRAE 62.1Ventilation rates for INFRA AHU and occupant spaces
§ ASHRAE 90.1Building envelope and climate-zone constraints
§ IEEE 1547-2018Parallel generation interconnection (BTM augmentation)
§ UL 1008Automatic transfer switch certification
§ UL 2166Pre-engineered clean-agent system listing
§ UL 263 / ASTM E119Fire-resistive construction tests (Smart Shell ESR-779)
§ ICC-ES ESR-779Panelized component evaluation report (Livio Smart Shell)
§ FERC OASISInterconnection queue publication for ISOs (ERCOT, PJM, CAISO, MISO, etc.)

ISO data sources.

Power-timing claims are cross-checked against publicly-filed interconnection queue data from these grid operators.

ISO / RTORegionQueue source
ERCOTTexasGINR (Generator Interconnection Status Report)
PJMMid-Atlantic + MidwestNew Service Queue + RTEP
CAISOCaliforniaGenerator Interconnection Studies + Cluster Studies
MISOUpper Midwest + SouthGIQ (Generator Interconnection Queue)
NYISONew YorkClass Year Studies
ISO-NENew EnglandGenerator Interconnection Queue
SPPCentral USGeneration Interconnection Studies
AESOAlbertaProject List
IESOOntarioAvailable Capacity Reports

Commodity catalog approach.

Lead times are tracked weekly across stocked vs custom suppliers for every long-lead category. Specific supplier data is proprietary, but every category and the sources we use are listed.

  • UPS modular — Vertiv Liebert, Eaton 93PM, Schneider Galaxy VL · 8–14 wk stocked
  • MV switchgear — Schneider FlexSeT, ABB ZX, Siemens 8DJH · 6–18 wk depending on stocked vs custom
  • Diesel generators — Cummins QSK60, CAT C18, Kohler series · 6–14 wk stocked
  • Aeroderivative gas turbines — GE LM2500, Solar Mars 100, Mitsubishi MFT-8 · 24–36 wk
  • Pad-mount transformers — Howard Industries, Maddox, ERMCO · 2–10 wk stocked
  • CDU in-row — Vertiv CoolChip, Motivair, CoolIT · 8–12 wk stocked
  • Dry coolers — Güntner VARIO, Evapco EAW-VD, Modine · 5–9 wk stocked
  • Air-cooled chillers (hybrid) — Trane RTAF, Carrier 30RB, York YVAA · 8–14 wk
  • BESS LFP — CHINT, CPS, LIVOLTEK · 4–8 wk stocked
  • ATS — ASCO 7000, Cummins OTEC, Thomson · 6–12 wk

Confidence & caveats.

Confidence intervals. Every numeric finding carries a confidence indicator. When site-specific data is sparse, the engine surfaces a range and a "verify with utility" flag rather than a confident wrong answer.

This is not engineering of record. A Livio Grid review is a fast first pass for buyer-side decision support. Permit-stamped engineering requires a licensed engineer in the jurisdiction of construction. The reviewing engineer should validate parameter parsing and may apply local knowledge to refine findings.

Disagree with a finding? Tell us. Concerns are triaged within 24 hours. If a finding is materially wrong, the report is updated with a public revision history.