Most logistics optimization trades cost against service inside a comfortable range: a late shipment annoys a customer, a stockout delays a sale. Healthcare removes the comfort. A surgical kit that misses the case, a critical medicine that stocks out, a diagnostic sample that expires in transit: the downside is measured in patient outcomes, and no rational objective function treats those failures as just another cost coefficient to minimize.
The formal consequence is that healthcare networks should be optimized under chance constraints rather than pure cost minimization: hold the probability of failure below a hard threshold, per product criticality tier, and minimize cost subject to that. This is a different mathematics with different answers. It spends money on redundancy, dual sourcing, and forward stocking that a cost minimizer would strip out, and it spends selectively, sized to each product's actual failure distribution rather than to blanket safety-stock rules.
Criticality tiering does the heavy lifting
The practical craft is in the tiering. Not everything in a hospital's flow is life-critical, and treating everything as if it were is how healthcare supply chains end up simultaneously overstocked and exposed: capital frozen in low-risk items while the genuinely critical ones ride on single points of failure nobody has mapped. A defensible tiering, clinical consequence of failure, substitutability, demand volatility, supply concentration, lets the network hold 99.9 percent availability where it must and market-standard economics where it may.
Each nine costs more than the last.
Raise the promise and watch the cost curve bend.
In healthcare logistics the failure budget is measured in patients, which is why the curve is worth paying.
Variability is the enemy, visibility is the weapon
Healthcare demand contains structure most industries would envy, elective schedules, seasonal epidemiology, protocol-driven consumption, yet many health systems still plan on simple reorder points. Connecting the demand signal to the network model, and the network model to fitted lead-time distributions across a fragile multi-tier supply base, is where resilience actually comes from. The institutions doing this are discovering that patient-grade reliability and sane working capital are not opposites. They are what the same model produces when it is finally allowed to see the whole system.