Availability, safety stock and minimum order quantity — a measured reading of supply resilience

The observation. A stock-out is never measured by its cost of goods alone. It is paid in lost sales, in customers who take their business elsewhere, and in eroded trust. Yet supply resilience is too often treated as a logistics-cost topic rather than as a discipline of risk engineering. The founding work of supply-chain research — from the bullwhip effect described by Lee, Padmanabhan and Whang in 1997 to the classical safety-stock models — all says the same thing : availability is not luck, it is the result of explicit assumptions about demand, lead time and variability.

Safety stock is not a cushion, it is a calculation

Safety stock covers the joint uncertainty of two quantities : demand during the replenishment lead time, and the lead time itself. The classical formula links the target service level (expressed through a safety factor drawn from the normal distribution), the standard deviation of demand and that of the lead time. The counter-intuitive lesson is that lead-time variability often weighs more heavily than demand variability : a supplier who delivers sometimes in two weeks, sometimes in six, forces a far costlier cushion than a slow but reliable one. Regularity can beat speed.

Service level, in turn, is not linear. Moving from 95 % to 99 % fill rate does not cost 4 % more inventory but a great deal more : the tail of the distribution is expensive. This is why APICS — now ASCM — recommends differentiating service targets by item class (ABC analysis) rather than applying a uniform rate that overstocks slow movers and understocks critical references.

The bullwhip effect : variability amplifies up the chain

The paper by Lee, Padmanabhan and Whang, published in Management Science in 1997, formalised a long-observed phenomenon : a modest swing in end demand amplifies at each upstream link. Four main causes are identified — order batching, rationing games under shortage, promotional price fluctuations, and above all forecast updating by each actor in isolation. Each link over-reacts to the signal it receives and passes on an even noisier one.

The operational consequence is sharp : the more independent links a chain has that do not share true demand information, the higher the total inventory tied up for a given service level. Information sharing — not the stacking of local cushions — is the first lever of resilience.

Minimum order quantity : the hidden economics of the lot

Minimum order quantity (MOQ) and economic order quantity (EOQ, the Wilson formula, 1913) pull in opposite directions. EOQ trades off the cost of placing an order against the cost of holding stock, suggesting an optimal lot size. MOQ, imposed by the supplier or by production constraints, often forces ordering more than the theoretical optimum. The gap between the two shows up as dormant stock, obsolescence risk and tied-up cash.

When the same reference is sourced from several entities — factories, intermediaries or distinct legal entities — each source brings its own MOQ. Summed together, these constraints can inflate total inventory well beyond what consolidated sourcing would require. This is the multi-sourcing paradox : it reduces the risk of single-supplier failure, but can raise working-capital needs if the lots are not coordinated.

Single, dual, multi-sourcing : a risk trade-off

The literature on supplier-risk management — abundant since the 2020 pandemic and the semiconductor shortages — converges on one principle : diversifying sources reduces the risk of correlated disruption, at the price of greater coordination complexity. Dual sourcing is often presented as the pragmatic optimum : a competitive primary source and a qualified secondary one ready to ramp up. But diversification only has value if the sources are genuinely decorrelated — two suppliers depending on the same raw material or the same port region diversify nothing.

Fragmenting sourcing across distinct legal entities adds a layer : each entity holds its own inventory, its own lead times and its own forecasts. Without shared visibility, the bullwhip effect replays inside the group itself. The remedy is not forced centralisation but transparency of demand and stock data end to end — what research calls end-to-end visibility.

What it requires

Building a resilient chain is neither overstocking out of caution nor chasing zero inventory. It is making the assumptions explicit : which service level per item class, which lead-time variability per supplier, which imposed MOQ and which gap to the economic optimum. It also means accepting that demand data is to be shared rather than guessed link by link. Resilience is a property of the whole system, not of an isolated warehouse.

Where we stand

Montandor operates a multi-entity supply base, which places us squarely in front of these trade-offs. Our work consists of making lead times, minimum lots and service levels legible, so that the ordering decision rests on named assumptions rather than habit. We publish no internal figures here : what is shared is the method, not the order book.

“You do not judge a supply chain on the good days. You judge it on the day of the stock-out — and on that day it is the assumptions set months earlier that speak. A serious house writes its assumptions down before it needs them.”
— Wouter Meijboom, CEO, Montandor Andorra.

Sources

• Lee, H. L., Padmanabhan, V., Whang, S. — Information Distortion in a Supply Chain : The Bullwhip Effect, Management Science, vol. 43, 1997.
• Harris, F. W. — economic order quantity model (EOQ / Wilson formula), 1913.
• ASCM / APICS — Dictionary and reference corpus on safety stock, ABC analysis and service levels.
• Silver, E. A., Pyke, D. F., Peterson, R. — Inventory Management and Production Planning and Scheduling (classic reference on safety stock and lead-time variability).
• McKinsey & Company / Bain & Company — analyses of supply-chain resilience and dual sourcing post-2020.
• Eurostat — industrial statistics and inventory indicators in EU manufacturing.

Research led by Sanne Bakker (Supply Chain & Demand Forecasting), in collaboration with Margaux Lefèvre (CTO), Mihail Ivanov (Integration Engineer) and Lev Marchuk (Data Scientist). Published 13 May 2026 by the Montandor team.