What data do you need in order to determine correct safety-stock levels?
Common sense says that safety stock addresses variation with unpredictable timing. (Variation with predictable timing is what a forecast is for!)
What kinds of variation affect service levels and have unpredictable timing?
Typically, there are just two:
- Random demand (or usage) variation
- Random lead-time variation
By definition, random variation has unpredictable timing. However, in a process (such as customer demand or replenishment lead time) that is reasonably stable, random variation has quantifiable magnitude. From a safety-stock perspective, this is good news, because we may now use objective statistical techniques to establish and optimize safety stock.
In a stable process, the magnitude of historical random variation provides an excellent approximation of expected future random variation. In addition, the more historical data you have, the more reliable this approximation will be.
What does this mean in practical terms?
First, let’s look at the easier of the two sources of safety-stock-related variation – demand.
Usually, this information is available
- At a granular level, often daily, which:
- Provides more data points and more reliable statistical results
- Reflects the customers’ sensitivity to what constitutes “late” – often, just one day
- At the individual-inventory-item level, since customers (and, for components, higher-level assemblies) are sensitive to individual-item lateness.
In our next blog, we will consider the second, more challenging source of safety-stock-related variation: Lead time.
