A hub is a 15-to-30-year commitment justified by a 3-year forecast. Every seasoned network planner knows how this movie ends: the forecast is obsolete before the concrete cures, and the network spends a decade working around a facility located optimally for a world that no longer exists. The failure is not forecasting quality. No forecast survives a hub's lifespan. The failure is asking for a single optimal location at all.
The stochastic formulation asks a better question: across the distribution of demand futures, growth scenarios, trade-pattern shifts, cost trajectories, which configuration performs acceptably in nearly all of them, excellently in the likely ones, and catastrophically in none? The answer is frequently not the site that wins the base case. It is the site, or the pair of sites, whose performance is flattest across futures: slightly worse than optimal in the center of the distribution, dramatically better in the tails.
Staging as an option
Uncertainty also reprices staging. Building 60 percent of capacity now with secured expansion rights, at modestly higher unit cost, is an option purchase: pay a premium, keep the right to scale into the futures where scale pays. Deterministic analysis rejects the premium as inefficiency. Option-aware analysis frequently finds it is the cheapest insurance the network will ever buy. The same logic applies to leased-versus-owned, single-versus-split, and automated-versus-flexible fit-out. Every one of these is an option decision wearing an engineering costume.
The best hub for one future is the wrong hub for the average of futures.
Pick a site, then flip the demand scenario.
| Site | Cost A | Cost B | Cost C | Expected |
|---|---|---|---|---|
| S1 | 15.0 | 30.7 | 22.5 | 22.7 |
| S2 | 15.3 | 21.7 | 17.6 | 18.2 |
| S3 | 18.3 | 17.5 | 16.7 | 17.5 |
| S4 | 24.8 | 14.9 | 19.2 | 19.6 |
| S5 | 32.8 | 16.2 | 24.2 | 24.4 |
Optimizing for scenario A buys regret in scenario B. The expected-cost hub is nobody's favorite and everybody's insurance.
The reversibility test
A useful discipline for any facility decision: rank the alternatives not only by expected cost but by cost of being wrong. How expensive is exit, resize, or repurpose in the futures where this choice ages badly? Configurations that score well on reversibility deserve a thumb on the scale, because the one certainty about a 25-year asset is that the world it was designed for will not be the world it retires in.