Safety Stock is a lot like Marinara sauce. Everyone guards their own recipe. No one divulges their secret ingredients. Everyone seems to like their own version best. In a blind taste test, it can be difficult to identify your own cooking. Most evaluations are qualitative, not quantitative.
But, on that last comment… There are two simple quality tests for safety stock that would be very revealing about your software’s overall performance. Yet, almost no vendors offer them in their ERP software packages.
- Does your safety stock formula deliver the requested results? It should be that simple… If you set a service level goal of 98% on a sku or category, that’s exactly what you ought to get. Yes, there will be demand and lead time fluctuations. That is EXACTLY what safety stock is supposed to address. Over time, and over a range of skus, your software should produce service level results that match your goal.
- Do you have the expected amount of safety stock inventory? Your “on hand” inventory just before you receive a shipment is the actual safety stock. It should be easy to compare that actual inventory to the computed, desired safety stock to validate your software’s performance.
These two measures would validate that your safety stock formula delivers the service levels you desire, and that you have the expected amount of safety stock inventory investment. If your software system offers these two measurements, use them!
Many inventory software packages use a “classic” safety stock computation, based on the Normal Distribution. The Normal Distribution accurately predicts the probability of random, independent events. For example, it will accurately predict the probability of a specific number of “heads” in tosses of a “fair” coin.
Some software packages use a safety stock formula that sets safety stock to achieve a percentage of stockouts, not a service level. A formula that predicts a sku will not have a stockout in 98% of replenishment cycles is NOT the same as asking for an overall 98% order fill service level.
In our auto parts distribution business, the “real world” does not exactly follow the pure Normal Distribution statistics. Not all demand “events” are truly independent and random. The demand for a left front brake caliper for a 2008 Toyota Camry is partially related to the demand for the right front brake caliper for that same car. They are not completely independent. If an item has a relatively high rate of alleged warranty claims, we often encounter an additional demand for a sku shortly after a sale, as a warranty replacement. That is not a random event. Events often bunch up. When the market is “hot”, demand is above forecast, and often, our vendors are also busier than usual and take a bit longer than normal to process the larger orders. So, higher than forecast demand and longer than forecast lead times are often in synch, and not independent. And, other “real world” aspects of customer behavior can make the need for safety stock greater than calculations based on the pure statistical distributions.
Most inventory managers think about service level in terms of an overall order fill rate, not a percentage of cycles with stockouts.
So,
www.InventoryCHAMP.com offers its own version of a safety stock formula.
The InventoryCHAMP.com formula has an empirical correction to the pure statistics. That is JFF’s “secret ingredient”. It is based on observations that most pure statistical safety stock formulas “underperform”. If you set the service goal at 97%, with most software packages you might only get 95%... You may find the safety stock levels in the InventoryCHAMP.com calculator to be slightly higher than most software package results.
At a minimum, the JFF model can be a good way to compare safety stock results to your own software package, and help you “tune” your software to achieve the results you desire.
And, finally, there is a lot that can be done to help decide what your service level goals should be. For more on this, please contact me at
petekornafel@msn.com.