Evidence class. Observed is a publisher value; derived is deterministic arithmetic over observed values (e.g. sum member-county GDP endpoints, then calculate CAGR — county growth rates are never averaged); modeled is an explicit estimate; proxy is directional evidence that does not satisfy the literal criterion and never increases direct coverage; missing means no supported computation exists yet. None of these labels are a forecast-confidence score.

Direct coverage & weighting. Requested criteria are weighted equally (100.0% each) because no explicit weights were supplied. Coverage is the share of that weight backed by direct observed/derived evidence — proxy and missing criteria stay in the denominator and never inflate it.

Decision interval. The lower bound sums only direct-criterion contributions; the upper bound adds every proxy or missing criterion's full weight, since it could theoretically be satisfied. Overlapping candidate intervals mean the ranking between them is unresolved — Atlas will not force an ordinal precision the data doesn't support.

GDP growth above the national average. Metro nominal GDP CAGR (endpoint sum of member counties, or the published CMA series for Canada) compared only against the same-country national aggregate over the identical period. US and Canadian levels are never compared to each other directly — each is scored pass/fail against its own country's benchmark.

Increasing building permits. Residential permit-unit CAGR from complete annual endpoints (2021-2025 for US metros and Canadian CMAs); increasing means CAGR greater than zero. Known limitation: Connecticut metros currently show this criterion as unavailable — Census Bureau BPS permit data uses legacy county FIPS codes in 2021 and the state's new planning-region FIPS codes in 2025, so the two endpoints don't join to the same geography under the current delineation. This is a real data-vintage gap, not a general unavailability of Connecticut permit data.

Industrial land values below the regional median. Requires comparable metro industrial land transaction or assessment values against a defined regional peer median. No free, maintained source of this data exists; Atlas has not approved a proxy for it, so it stays missing rather than being estimated.

Expanding freight volumes. Requires a metro-attributed freight-volume time series. Static transport access, distance to infrastructure, or a single year of port tonnage does not establish growth and is never substituted for it.

Fewer than 15 civil construction contractors over $50 million revenue. Requires civil-sector classification, company-level annual revenue, and operating-market geography — public-company HQ counts or establishment totals do not satisfy it, since most civil contractors are private and outside current company coverage.