Normal Distribution

Statistical safety stock level can be calculated to prevent a give number of stock-outs by using 'normal' distribution to determine the deviation. Two methods generally are used to find measure the deviation - standard deviation and mean absolute deviation. Both are just the two different ways of measuring the same thing.

Figure 4.0  Normal Distribution

With the normal distribution, the forecast is assumed to be representative of the actual usage rate on average. Deviations above and below the forecast are assumed to relatively proportional on average. Conditions that are preventing stock-out are also assumed to be constant. Therefore, when a standard deviation is calculated, it is assumed to be representative if the deviation to be encountered in the future on average.

In figure 4.0, the midpoint is the forecast, which is the mean. The downward slope curve on either side of the mean indicates that the probability of having a specific actual demand greater than the forecast gets lesser and lesser as the deviation from the forecast gets larger and larger.

The width of the distribution curve varies from part to part based on the deviation each part experiences. A part with a small sigma (σ) will have a skinny distribution curve. A part will a large deviation will have a fatty distribution curve. The width of the curve is a direct indication of the safety stock quantity that you must stock to compensate for the deviations to prevent stock-out.

Sigma corresponds to a particular quantity based on average of the deviation. If there is a forecast of 50 and also have 50 in stock, you have 50% probability of satisfying demand for the period without the need for safety stock. However, if there is a safety stock quantity equal to 1 standard deviation, you have an 84% chance of able to satisfy demand for the period as shown in figure 4.0. This percentage, which is referred to as "safety stock service percent", increases as safety stock is provided for each additional increment of the standard deviation.