Methodology

The strongest defense against a critic is naming your own gaps first. These are the places where the published score is least informative — read accordingly.

• Counties and independent cities are scored together. Virginia's 38 independent cities share the 51xxx FIPS namespace with its 95 counties. Per-capita denominators (MW/pop, permits/pop, estab/pop) make small cities volatile at both extremes — interpret tiny-population jurisdictions with extra care.
• Connectivity uses household broadband subscription rate (ACS S2802) — a residential consumer proxy. It does not measure data-center fiber routes, FCC BDC coverage, or actual compute capacity.
• Energy uses installed generation capacity, not capacity-factor-adjusted available headroom, PJM interconnection-queue position, or transmission constraints — which are the binding constraints in VA today.
• Water is a 7-region rollup of USGS gauges, not per-county hydrology, and the 30-day window measures recent weather — not long-term supply or drought normals.
• Housing uses one year of BPS permits. Permits issued ≠ units delivered ≠ workforce-housing affordability.
• Econ Dev blends establishment density (Census CBP) with per-capita personal income (BEA CAINC1) to avoid weighting a Dollar General the same as advanced manufacturing.
• No statistical normalization across pillars. Weights are flat; spread comes from rank-percentile within VA, not z-score against national benchmarks.
• No transmission constraint model. A 1 GW PJM queue project in Loudoun is not interchangeable with 1 GW in Wise County — the network won't accept them the same way. We surface PJM queue data, but the score does not yet penalize zones with binding constraints.
• No behind-the-meter generation. Self-generation at large customer sites is invisible to EIA's grid-side dataset.

Pillars refresh on different cadences (BLS monthly, ACS annual, USGS daily). A stale ingestion date means the underlying source hasn't been re-pulled — not that the source itself is out of date.

A reader should be able to take the public inputs for a county, run the formulas below, and arrive at the same composite score we publish (within rounding). Here is the procedure for a single county — replicate it for any FIPS code.

1. Pull the county's pillar inputs from the linked public sources: MW per capita (EIA 860 ÷ ACS pop), unemployment rate (BLS LAUS), permits per capita (Census BPS ÷ pop), region streamflow percentile (USGS), broadband rate (ACS S2802), and the CBP / BEA econdev components.
2. For each pillar, compute its rank-percentile across all 133 VA jurisdictions and apply the formula in the next panel. The result is the 0–100 pillar score. Missing inputs default to 50.
3. Multiply each pillar score by its weight (above) and sum. That's the composite. Compare to capacity_scores.overall_score for the same county at the same methodology version.
4. If your number differs, the source registry has the exact source URL and last retrieval time of every input we used. Disagreements should reconcile to one of: different ACS vintage, different LAUS month, or a missing pillar. The site never uses imputed inputs.

• High — 80+ · top quintile within Virginia.
• Moderate — 60–79 · above the statewide median.
• Limited — 35–59 · below median, watch for binding constraints.
• Constrained — under 35 · materially limiting for industrial-scale growth.
• Scores are rendered as whole numbers across the UI; the underlying values are unrounded.

• High — 5 or 6 pillars have real data for this county.
• Medium — 3 or 4 pillars present.
• Low — 2 or fewer pillars present; treat overall score as indicative only.
• Every metric carries its source name, URL, collection method, and last-checked date.