Inventory sitting unsold looks like waste, yet every serious operation holds some deliberately. The size of that buffer is calculated from uncertainty, not from how much a company expects to sell.
Buffers absorb variability, not volume
If demand and delivery times were perfectly predictable, a company could order exactly what it needed to arrive exactly when required.
Neither is predictable, so stock is held to cover the difference between what was forecast and what actually occurs.
The quantity therefore scales with how much outcomes vary around the average, which is a different quantity from the average itself.
Two products selling identical volumes can require very different buffers, because one moves at a steady rate and the other arrives in unpredictable bursts.
Lead time variability compounds the requirement
A supplier whose delivery time swings widely forces a larger buffer than one whose average is longer but consistent.
The buffer must cover the worst plausible combination of a demand spike and a late delivery occurring together.
This is why reliability is often worth more than speed in supplier selection, and why contracts specify delivery windows rather than averages.
Service level is a chosen target
Companies set a target probability of not running out during a replenishment cycle, and the buffer follows from that choice.
Raising the target toward near-certainty increases required stock disproportionately, because covering rare combinations demands far more than covering typical ones.
The target is a business decision weighing the cost of held inventory against the cost of a stockout, which differs by product and customer.
Lean operation trades buffer for exposure
Reducing inventory frees capital and warehouse space and reveals process problems that stock had been concealing.
It also removes the cushion that absorbs disruption, so a system running lean transmits an upstream interruption to customers quickly.
Firms that experienced severe disruption have generally rebalanced rather than reversed, holding larger buffers on components that are hardest to substitute.
The distinction that survived is between parts with many qualified suppliers and parts with one, since only the second kind justifies carrying months of cover.
Location changes what a buffer costs
Holding stock centrally lets one pool cover demand from many regions, since shortfalls in one area offset surpluses in another.
Holding it close to customers shortens delivery but requires each location to carry its own buffer, raising total inventory for the same service level.
Network design is therefore a direct trade between delivery speed and inventory investment, and the right point depends on what customers actually require.