Safety Stock Is a Policy Decision, Not a Formula

Every distribution ERP will happily compute safety stock for you. Demand variability, lead-time variability, a Z-score for your target service level — plug in the numbers and out comes a quantity. The math is fine. The problem is what the math quietly assumes: that every SKU deserves the same conversation.

It doesn't.

The formula answers the wrong question

The textbook calculation answers "how much buffer do I need to hit a 98% service level on this item?" The question a buyer actually faces is different: "is this item worth a 98% service level at all?"

Those are separated by things no formula sees:

• Whether a stockout loses the line or loses the customer
• Whether the item is one of six substitutes on the shelf or the only one
• Whether the vendor's stated lead time survives contact with reality
• Whether the carrying cost lands on a $4 fitting or a $4,000 valve

Treat safety stock as pure math and you get the classic distributor failure mode: deep buffers on cheap, forgiving items and thin ones on the SKUs that actually cost you accounts.

The service level target is the decision. The safety stock quantity is just its consequence.

What this looked like in practice

When we overhauled min/max levels across a seven-branch network, the leverage wasn't a better formula — it was segmentation before any formula ran. A simple matrix of demand predictability against stockout consequence sorted 60K SKUs into a handful of policies:

| Segment | Policy | | --- | --- | | Predictable, high consequence | High service target, formula-sized buffer | | Predictable, low consequence | Lean buffer, let the reorder cycle work | | Erratic, high consequence | Buffer + human review flag | | Erratic, low consequence | No buffer — order to demand |

The last row is the one people resist, and it's where most of the $4M in freed working capital came from.

The takeaway

If your safety stock review starts with the formula, it's already off the rails. Start with the policy question — which items are we willing to fail on, and how often? — and let the math do the only thing it's good at: pricing the answer.