Tighten what you promised the shipper. Every stop down to 96 h is a separate solve, each returned OPTIMAL at gap 0.0. Below 96 h a lane only turns to air when road cannot hold the date, and air transit there is derived rather than solved. Road transit respects hours of service — 11 driving hours, then a 10-hour reset — so the slowest road run here is —, not a straight-through drive.
Where each mode earns its place
Nothing strands anywhere on this slider. Amber is a lane past its rail cutoff, with the container going back to road, which still delivers — CHI–ATL is 689 road miles, comfortable inside 96 hours. Down to 96 h road alone holds every promise, and it costs the truck-only baseline to the cent, which is why the saving there is zero. Air only starts earning below that: at several times road’s cost for the same freight it buys nothing where road already makes the date, and everything where road cannot.
Solved row
Network cost / week—
Saving vs truck-only—
Containers on intermodal—
Intermodal share—
Routes forced to road—
Undelivered0 at every stop
Solver statusOPTIMAL · 0.0
Lanes flown · derived, not solved
Too long for road, so they fly—
Short enough to stay on the ground—
Longest lane road can still hold—
Network cost / weeknot quoted
Saving vs truck-onlynot quoted
These are the lanes no road run can hold at this promise, so they fly. Their transit is arithmetic rather than a solve, derived from published block speed and handling time. Cost is withheld because the air model and this one count different units, and converting between them without a re-run would be a made-up number.
Baseline
Truck-only on the same lanes: $11,341,898.65/wk. Savings shown above are measured against this number, not against a list price.
Not shown
27 of the 60 solved lanes are hidden, for two different reasons. Five BEA areas (010, 020, 160, 182, 185) are absent from the shipped code table, so they cannot be named; a sixth, 011 Harrisburg, is named there but has no coordinate in the terminal table used for placement. The map carries 30.1% of the instance's containers. Lane-level rail/truck is reconstructed by cost rank to match each solve's measured rail share; the share, cost and status are measured, the per-lane attribution is not.