Safety Stock Is the Wrong Kind of Insurance
Safety stock has enjoyed a good reputation for far too long. The phrase itself is excellent branding. It sounds prudent, numerical, managerial. Add a bell curve, a service-level target, and a neat formula, and the method passes for science. Yet it is a poor guide for capital allocation and a poor guide for resilience. In many companies, the units bought in the name of safety are where the losses begin.
I laid out the longer argument in Introduction to Supply Chain, especially Chapter 3 on safety stocks, Chapter 4 on the economics of inventory, and chapter 8 on resilience, delayed commitments, and insurance. The short version is simpler. A company does not become safer because uncertainty has been turned into pallets. It becomes safer when it has chosen the cheapest and most effective hedge against the loss it actually faces.
Unsafe stocks
The difficulty begins with a categorical mistake. Safety stock treats each SKU as if it alone mattered. Extra units are computed item by item, while the balance sheet is shared by all items at once. Cash committed to one slow mover is cash unavailable for another product, a price move, a supplier improvement, or simply the option to wait. Once the problem is seen as capital allocation, the idea of an “optimal” stock buffer for one SKU in isolation becomes very hard to defend.
Then comes the service level, a proxy that has been promoted into a command. It pretends to settle the question without pricing it. A 98% target can be reckless in one business and wasteful in another. For an aircraft grounded for lack of a part, it is laughably low. For a fashion item near season end, it can be indulgent. The percentage does not know the difference. One day late, one unit short, thirty units short, or a complete miss are flattened into a dashboard convention too crude to carry the economics of the situation.
The statistical story is no better. Demand is lumpy, intermittent, and prone to abrupt shifts. Lead times are rarely well-behaved bell curves; they are often stable until they are not, with customs delays, supplier stockouts, or transport failures producing the tails that matter. The usual repair tells the story: companies inflate the safety stock by hand, or they ask the software for a higher service level than they truly want, simply to compensate for the model’s blindness. A method that calls itself safe and survives only by systematic correction has already failed its own promise.
I call them unsafe stocks for this reason. The warehouse does not contain noble “working stock” on one side and sacrificial “safety stock” on the other. It contains units. Some will be sold at full price, some at a markdown, some after costly delay, some never. A hundred units expiring tomorrow do not have the same economic meaning as a hundred units expiring next year. Any method that treats them as equivalent has wandered away from commerce and into clerical fiction.
When the better buffer is not physical
Once the problem is stated correctly, inventory becomes only one hedge among several. Sometimes it is the right one. Often it is clumsy. A tariff risk may be better handled by developing a second supplier than by filling a warehouse. For fashion or perishables, preserving generic stock and delaying final commitment can protect margin far better than buying finished goods early. When the exposure is a temporary interruption of earnings, a contractual or financial hedge may absorb the loss more cheaply than months of speculative inventory.
When I speak of insurance here, I do not mean only a policy from an underwriter, though that too can have a place. I mean every expenditure whose purpose is to preserve future profit under adverse conditions: backup suppliers, reserved capacity, favorable termination clauses, postponement mechanisms, and, in some cases, literal business interruption cover. The common feature is easy to state. The company pays a visible premium today in order to avoid a larger loss tomorrow. That is cleaner than pretending that every uncertainty should be neutralized by extra inventory sitting on a shelf.
A financial hedge has limits, and there is no benefit in denying them. No insurance policy can place a missing part on the shelf this afternoon. For line-down manufacturing, aircraft-on-ground events, or other cases where the service itself is immediate physical availability, stock can be indispensable. Inventory does not need to disappear. It needs to win an explicit competition against the alternatives.
Why the old reflex survives
Mainstream academia has not walked away from the old reflex. Recent reviews still treat safety stock dimensioning, positioning, and management as a central research program. The newer resilience literature broadens the menu to stockpiling, multi-sourcing, capacity reservation, and flexible supply contracts. That is an improvement. Yet the instinct remains physical. A smaller stream at the operations-finance interface studies interruption insurance together with inventory and preparedness. It is closer to the real question, but it remains peripheral to the teaching canon.
The large software vendors mirror the same hierarchy. SAP and Oracle still document inventory planning in terms of service targets and safety stock calculations. Kinaxis describes inventory optimization as a balance between service targets and inventory levels, with single- and multi-echelon techniques centered on safety stock. o9 adds postponement, shelf-life, network rebalancing, and risk sensing to its MEIO toolkit, while Blue Yonder presents inventory optimization as choosing the amount of stock required to meet customer service goals while minimizing capital tied up in inventory. The software is more polished than the old reorder-point spreadsheet. The mental picture is largely unchanged.
This persistence explains the cult of spreadsheet overrides. Planners know, even when the official model does not, that the suggested order quantity may be absurd once perishability, markdown risk, supplier quirks, truck constraints, or political shocks enter the picture. When a planning system needs an army of clerks to correct its outputs line by line, the company has already paid for the refutation. The spreadsheet is less a complement to the model than its daily refutation.
A sensible supply chain does not ask how much safety stock feels comfortable. It asks which loss threatens the business, which hedge addresses that loss most cheaply, and how much flexibility is worth buying. Sometimes the answer is physical inventory. Quite often it is cash kept uncommitted, capacity kept available, commitments delayed, backup suppliers kept ready, or insurance bought explicitly. Once the problem is framed that way, safety stock loses its mystique. It becomes what it has always been: one possible hedge, often overpriced, often poorly targeted, and very often mistaken for safety simply because the name flatters the buyer.
