How locations are ranked
Model profiles are data. Every use model is defined in the model profile
library (demand signals, supply and competition signals, land floor, radii, weights, deal
required unknowns) — iVerify ranks locations and explains fit from those profiles;
development models, component recipes, and cost live in modular.nsgia.com. Truck parking and
IOS are fully ranked. The pilot models rank from held data at their profile radii and floors
— travel plaza scores the distance to the nearest branded plaza (competition inverts: a
Love's or TA inside the primary band scores against a site unless the corridor carries high
freight flow, where co-located plazas are the norm), fleet/3PL and port/rail and petrochem
score carrier depth, DC clusters, and typed anchors, and distribution scores 2020 Census
population plus DC clusters as the labor-shed stand-in. Every pilot score is labeled, its bands
are fixed, and the model card states what each still needs.
Candidates are mainline-AADT stations on the market's NHFN freight routes — TxDOT
mainline AADT for Texas markets; the FHWA HPMS 2022 NHS AADT database (NHFN-flagged
segments) for all other markets — (clustered to corridor points) plus freight generators
(ports, rail, terminals). For each candidate the engine precomputes ring measures at
1/2/3/5/8/15/30 road-miles (road = straight-line × 1.3), then scores each model from the
bands you set. Land gate: a candidate with no parcel ≥ 20 acres in the
primary band is set aside.
Demand (LD): corridor = 0.45·N(ln 1+truck_aadt) + 0.35·max(N infra, N carriers) + 0.20·min(…);
generator = 0.60·max + 0.40·min. Supply gap (SG) =
N(ln 1 + demand/max(spaces,10)) where demand is truck flow for parking and carrier
power units for IOS, and spaces = public NTAD spaces + private truck stops at brand-typical
capacity estimates (Love's 85 · Pilot Flying J 75 · TA/Petro 100 · QuikTrip 15 · other 25).
For parking, SG additionally folds in a branded-stop-desert signal —
SG_parking = 0.65·N(flow gap) + 0.35·N(distance to nearest branded stop, capped 50 mi)
— because a stretch of high truck flow with no branded stop nearby is the strongest
unmet overnight-parking signal (a secured yard is a different product and can coexist near a
branded stop, so this is a demand signal, not a competition penalty).
Cost (COST) = 100 − N(ln 1 + land $/acre), where $/acre is the
band median of per-parcel market estimates: assessed values (cast from county records;
$0/exempt parcels excluded, not read as missing), with ag-band parcels (<$15K/ac assessed) in
validated corridor counties (Chambers, Jefferson) uplifted ×4.75 to market rough-order — the same
rule the broker site tool prints. Southeast counties (FL/NC/GA) carry assessor market/just
value with no uplift. Corridor position (CP) =
+5·min(valid,2) − 15·min(direct,2) − 10·[saturated], clamped [−30,+10].
Mainline visibility (VIS). vis_mi = road-distance (straight-line × 1.3)
from the candidate to the nearest qualifying freight mainline: an NPMRDS-on-NHS segment
carrying at/above the national-median combination-truck AADT (measured, not a road-class proxy).
VIS scores 100 at vis_mi ≤ 0.40 mi, decaying linearly to 0 at
vis_mi ≥ 4.50 mi. Both thresholds are derived from the branded-stop set:
0.40 mi is the 75th percentile and 4.50 mi the 95th percentile of stop-to-mainline
distance across 2,339 stops. That confirms the MAASTO "within sight of a major highway" siting
anchor. Travel plaza carries VIS as a weighted component and suppresses any candidate beyond
4.50 mi ("not visible from a freight mainline"). Parking applies it as an additive penalty
(0 to −12) rather than a killer; a secured yard slightly off the exit still works.
IOS is unaffected.
Model weights. Truck parking = 0.44·LD + 0.44·SG + 0.12·COST + CP + VISPEN.
IOS/laydown = 0.43·LD + 0.29·SG + 0.28·COST + CP.
Travel plaza = 0.42·LD + 0.23·SG + 0.20·VIS + 0.15·COST. Where land value is not published for a
county, COST is omitted, the remaining weights are renormalized, and confidence is capped. Scores clamp to 0–100.
Stated assumptions. Several constants in the formulas above are current
house assumptions, not yet derived from data or anchored to an external standard. They are:
the parking supply-gap blend 0.65 flow / 0.35 branded-gap; the branded-gap
distance cap of 50 road-mi; the spaces floor of 10; the screen
bands (parking 3 / 8 mi primary/secondary, IOS 5 / 15 mi); the
demand (LD) blends (corridor 0.45 / 0.35 / 0.20, generator 0.60 / 0.40);
and the top-level model weight vectors (parking 0.44 / 0.44 / 0.12,
IOS 0.43 / 0.29 / 0.28, travel plaza 0.42 / 0.23 / 0.20 / 0.15).
Each is pinned by the MTS3 gate, carries a review date of 2026-07, and is scheduled for
derivation or external anchor.
Flood is shown as data, not scored. At screen scale flood does not separate
one location from another — it is a parcel-level question, resolved in the site report. Each
location still carries its FEMA SFHA exposure for context where a local flood layer is loaded.
This screen does not know lease/parcel rates, ownership and assemblage, utilities
to the pad, entitlement and zoning, or truck-route legality — those are site-report questions.