Background
The vast majority of articles on this website that discuss service level tend to focus on fill rate, as this is the most popular service level measurement method in business. However, the majority of early work on inventory optimization and multi echelon planning that began in the late 1950s, and now drives the best of breed service parts planning software applications, was in fact designed around backorders. This is because the research was primarily paid for by the Air Force and carried out by the RAND Corporation, and the focus was squarely on solving the problem of managing military service parts networks. Therefore, it is interesting to compare and contrast two quotations from research papers that focused on minimizing backorders. This first is from Craig Sherbrooke and his METRIC (an acronym for Multi-echelon Technique for Recoverable Item Control) paper written in 1966. This is how Sherbrooke explains his use of backorders over fill rates in in his paper.
Fill rate — defined as the fraction of demands that are immediately fulfilled by supply when the requisitions are received — concentrates nearly all stock at the bases. The result is that when a non fill occurs, the backorder lasts a very long time. Similarly fill rates behaves improperly in allocating investment at a base when the item repair times are substantially different. Consider two items with identical characteristics except that one is base-reparable in a short time, and the other is depot reparable with a much longer repair time. Assume that our investment constraint allows us to purchase only one unit of stock. In that case, the fill rate criterion will select the first item, and the backorder criterion the second.
The fill rate possesses an additional defect. A fill is normally defined as the satisfaction of a demand when placed. But if we allow a time interval T to elapse, such as a couple of days, on the grounds that some delay is acceptable, the policy begins to look substantially different. As longer delays are explored, the policy begins to resemble the minimization of expected backorders.
In summary, the backorder criterion seems to be the most reasonable. The penalty should depend on the length of the backorder and the number of backorders; linearity is the simplest assumption. This is the criterion function most often employed in inventory models. - Craig Sherbrooke
Sherbrooke explains that he considers backorders superior for his purposes due to the following:
- Fill rates tend to concentrate stock at the bases (bases in Sherbrooke’s papers would correlate to DCs in industry-speak, with the depot being the regional DC or (RDC))
- Fill rates measure the satisfaction only at the point of initial delay, and do not measure how late a fulfillment actually occurs.
Therefore Sherbrooke designed an algorithm as part of METRIC a penalty which multiplies the length of the backorder by the number of backorders.
Leanard Laforteza states a similar reasoning in his paper for selecting backorders as a measurement for his paper designing a multi echelon system for supplying Marine military deployments.
Fill rate is the percentage of demands that can be met at the time they are placed, while backorders are the number of unfilled demands that exist at a point in time. In commercial retail, if customer demand cannot be satisfied, a customer either goes away or returns at a later time when the item is restocked. the first case can be classified as lost sales while the second case creates a backorder on the supplier or manufacturer. In military applications, especially in most critical equipment, any demand that is not met is backordered. The backorder is outstanding until a resupply for the item is received, or a failed item is fixed and made available for issue. These two principle measures of item performance – fill rate and backorders – are related, but very different. Commercial retailers are more interested in fill rate than in backorders because fill rate measures customer satisfaction at the time each demand is placed. Not only is fill rate easy to calculate, but it also helps retialers form a picture of how well they are meeting customer demand. Experience may tell them that a 90% fill rate on an item is not acceptable and will create customer complaints. On the other hand, backorders are not easy to compute as fill rate. Unlike commercial retail business, the military is not concerned with lost sales. The military measures performance not in terms of sales, but in terms of equipment availability.
In terms of supply support measurement, we recommend tracking backorders. Although fill rate tends to have clearer meaning to commercial suppliers, the rate does not have the same meaning in militar applications. Using the concept of backorders, a unit can determine the status of its supply support not just when the order is placed, but up to the time the item was received. - Leanard D. Laforteza
Here Laforteza does a good job explaining why backorders are more relevant for military application than fill rates. However, as the greater market for MEIO applications is civilian, vendors added fill rates and fill rates are not the dominant method of MEIO implementation. MCA Solutions, a service parts planning vendor with a substantial military client base can measure service level by fill rate or by availability (i.e. the uptime of equipment). However, while it does not measure fill rate by backorder as do Sherbrook’s METRIC or Laforteza’s approach, MCA allows for the flexible setting of backordering for different locations. MCA allows for the following settings:
- All locations to be backorderable
- Only the root locations to be backorderable
- No locations to be backorderable (which is the default).
MCA describes its management of backorders in the following way:
A Location is called backorderable if the unmet demand at that Location gets backordered at that Location and waits until the inventory is available at that Location. A Location is not backorderable (also referred to as lost-sales) if the unmet demand is passed to another Location or outside the supply chain. In backorderable models, preference is given to destinations that do not have enough inventory position to meet their child Location needs. - MCA Solutions
Conclusion
The service level measurement must fit the application. The early MEIO research papers were centered around military application, and thus used backorders, and backorder which is often the number of backorders times the average backorder duration serves as a common service level measure. However, civilian applications require fill rate as the service level measure.
References
“METRIC: A Multi-echelon Technique for Recoverable Item Control,” C.C. Sherbrooke, RAND Corporation, 1966
“Inventory Optimization of Class IX Supply Blocks for Deploying in U.S. Marine Corps Combat Service Support Elements,” Leanard D. Laforteza, Naval Postgraduate School Monterey, California, June 1997
Principles of Operation, MCA Solutions, 2007
{ 0 comments }
