Average demand does not need safety stock. If you sold exactly twenty units every day and the supplier took exactly fourteen days, you would order 280 units at the right moment and never hold a single spare. Safety stock exists only because neither of those numbers holds still.
safety stock = Z × standard deviation of daily demand × √lead time
With a daily standard deviation of 5 units, a 14 day lead time and a 95 per cent service level, the safety stock calculator gives 31 units and a Z-score of 1.65. The average daily demand does not appear anywhere in that formula.
The square root is the interesting part
Lead time enters under a square root, which means a longer lead time hurts less than intuition suggests. Quadrupling it from 14 days to 56 takes safety stock from 31 units to 62: four times the wait for twice the buffer.
Demand variability enters linearly. Doubling the standard deviation from 5 to 10 also takes safety stock from 31 to 62. So those two levers, one quadrupled and one doubled, land on precisely the same number.
That is worth sitting with, because it is the opposite of how these decisions usually get argued. Switching to a slower supplier is treated as a serious inventory decision and a product with erratic demand is treated as a merchandising problem. In the arithmetic, taming the demand variability is worth four times as much per unit of effort as shortening the lead time.
The reason is that errors over independent days partly cancel. Over 56 days the good days and the bad days offset each other more completely than over 14, so the total uncertainty grows more slowly than the window does.
Service level is bought in increments
The Z-score comes from how often you are willing to run out. At 90 per cent it is 1.28 and safety stock is 24 units. At 95 per cent it is 1.65 and 31 units. At 97.5 per cent it is 1.96 and 37 units, and at 99 per cent it is 2.33 and 44 units.
Nine percentage points of service level, from 90 to 99, costs 83 per cent more safety stock. The curve steepens all the way up, and pushing toward 99.9 per cent costs more again for a reduction in stockouts that most businesses cannot even measure. Choosing a service level is a commercial decision about the cost of a stockout against the cost of capital sitting on a shelf, and it is the only genuinely subjective input on the page.
Safety stock is one term of the reorder point
What you actually need is the level at which to place the order, which is the demand you expect during the wait plus the buffer for the demand you did not expect.
reorder point = average daily usage × lead time + safety stock
At 20 units a day over 14 days that is 280 units of expected usage, and the reorder point calculator adds an 80 unit buffer to give 360. Notice the split: 78 per cent of the reorder point is ordinary expected demand and only 22 per cent is protection against variance.
That proportion is the sanity check worth running. If safety stock is a large share of your reorder point, either the demand is genuinely erratic or the service level has been set by nobody in particular, and the second is more common than the first.
The Z-scores come from the normal distribution, and the z-score calculator is where that mapping lives if you want a service level the dropdown does not offer. The assumption that demand is normally distributed is the model's weakest point: real demand is often skewed by promotions and lumpy wholesale orders, and in those cases the formula understates what you need.