A disruption is a period in which supply from one or more sources is late, short or absent for longer than the normal variation allows. Sheffi's account of how firms weather disruptions, and Simchi-Levi's time-to-recover and time-to-survive measures, are the right frame: the question is not whether a disruption will happen, but how long the business can serve its customers when it does, and how quickly it can restore normal supply.

Before: know your exposure

Map every SKU to its sources and every source to the nodes it depends on: supplier plant, port of loading, carrier, transhipment hub, border, port of discharge, receiving warehouse. Two suppliers that share a node share a failure. Count the effective number of independent sources per SKU, weighted by how much capacity each can actually provide, and rank SKUs by the product of exposure, margin and customer commitment. This is the list that deserves protection; most of the range does not.

During: a sequence, not a panic

1. Reposition before buying. Stock already inside the network is the cheapest supply there is. Before an emergency purchase, check whether a transfer from a site with excess can cover the shortage without creating a worse one there. Transfer lead time, receiving capacity and shelf life all bound what is feasible.

2. Allocate by commitment. When stock is short, serve the customers with the highest priority and contractual targets first, and say so. A recovery plan that raises a strategic customer's service by lowering a spot customer's is legitimate; one that quietly does the reverse is not.

3. Use approved alternatives, priced honestly. An alternate supplier, an alternate route, a split order, expediting or an approved substitute each costs something. The right comparison is that cost against the expected shortage avoided plus the reduction in tail exposure. Moving part of an order to a more expensive supplier is worth it when the shortage it prevents is worth more, and not otherwise.

4. Do not fight the last war. One late shipment is not a regime change. Protection should rise with evidence: observed lateness, observed partial deliveries, confirmed route signals. A single abnormal week should move a probability, not the whole plan.

After: let protection fall

The most common mistake after a disruption is to keep the protection forever. Once the supplier's observed lead times return to normal and the route signal clears, the extra stock is working capital doing nothing. A plan that cannot come back to lean is not resilient, it is just expensive.

What to measure

Service achieved against target, by customer class; the share of misses that were avoidable (a feasible action existed) versus unavoidable (nothing feasible could have met the target); inventory in periods of demand; the number of emergency actions; and the time from the first missed period to full service. Track these separately from cost savings, because a plan can save money and fail every one of them.