Professional Documents
Culture Documents
Time
Actual Sales
Old Forecast
New Forecast “BIAS”
Qty
Safety Stock to
cover forecast error
New forecast
To understand the use of a “standard Deviation” we need to understand the “Normal Distribution Curve”………
Test of understanding….
What percentage of demands
will be between the mean and
plus 1 Std Deviation?
95% of all demands will be within 2 Std Devs. of the mean demand.
99% of all demands will be within 3 Std Devs. of the mean demand.
If a value was known for 1 Std Deviation in terms of “pieces”, then the percentage of demand
that would be covered by a given stock level could be predicted. (ie; predicting service level)
How can the normal distribution curve help?
It has to be assumed that the “Mean Demand” is the forecast. This is so because
the forecast has already been found to have no “Bias”, and the errors are “normally”
(equally) distributed above and below the forecasts.
If “mean demand” is equal to the forecast, and we know that, for example, 34% of all
“demands” will be one standard deviation above the forecast and the rest will be
below the forecast, we also know the following……..
• 50% of demands will be below the forecast….and we have these covered by
the “stock to cover forecast”.
• A further 34% of demands will be within one “standard deviation” of the
forecast.
• With the stock we have to cover the forecast, plus one “standard deviation”
worth of stock, we will cover 84% (50% + 34%) of demands. (ie; we will
achieve an 84% “customer service” level.
What we DON’T know is: How many pieces make up one “standard deviation”?
From the diagram we have derived that we must set one component of stock at the
forecast level (this covers 50% of demands that are likely to occur)
• Measure the forecast error over selected past periods and determine the
“MAD” (average of errors regardless of sign)
• Choose how many standard deviations we need to achieve the desired customer service level. The chart
below may help….
67% 0.5 Std Deviations
84% 1 Std Deviations
95% 1.5 Std Deviations
97.5% 2 Std Deviations
The result is the safety stock component. Add this to the stock to cover the forecast
and you have the total stock required for the desired service level.
Likely exam questions on forecast error measurement and demand probability……
• With a stock level set at 1 Std deviation above the mean, what percentage of demands are
likely to be satisfied?
• What is the probability of demand being within plus or minus 3 Std deviations of the
forecast when the forecast error is normally distributed about the mean?
• True or False? “Safety stocks must be increased if excessive bias exists in the forecast”
Answers;
1. 84% (50% plus half of 68%)
2. 99%
3. False (Safety stock cannot be set for a biased forecast)
4. 440 (The total stock level would be 1940)
5. c.