Modeling Out-Of-Stock Conditions
by James Hull, Business Process Manager, Staples and Hani Noshi, Strategy Practice, Accenture
Historically, due to the many factors and variables present, projects resulting in modest improvements early in the Supply Chain have not been viewed as having a quantifiable impact at the retail store level. Using Monte Carlo simulation, a team models and presents the ranges of improvements for Store Level Gross Margin that can be expected from a change in ordering practices.
One of the key drivers of customer perceived value in a Retail environment is whether the retailer is in stock for the item that the customer is looking to purchase. Measuring In Stock position is relatively easy using the Inventory Management System gives a good indication of In Stock, and this can be augmented with store associate and Mystery Shop audits to identify discrepancies between system in stock and true in stock positions.
However, identifying Out of Stock conditions (when store on hand drops to zero) is only the first step. A particular Out of Stock condition can arise from any of a number of process failures throughout the Supply Chain. Item sales at a particular store may oversell forecast and drain down safety stock. Store level shrink may cause discrepancies between system inventory position and actual in stock, delaying store replenishment. Store level execution can cause an item to be in the store but not on the sales floor. Or, Supply Chain failures may result in an item not being available for replenishment. Measuring and analyzing all of these possible root causes, across a network of thousands of stores, across thousands of different items and millions of transactions can be challenging to a project team to say the least.
The Challenge
Improving performance is good, but what does it mean for our customers? One of these Project Teams was chartered at a national retailer to improve a metric called 'Retail Insufficients'. This retailer has approximately 1500 stores, serviced from a handful of Distribution Centers. The inventory replenishment system is setup to supply the stores on a Pull basis, and identifies each individual item by a Store Keeping Unit, or SKU number. The stores are set on a pick schedule at the Distribution Centers to be delivered anywhere from one to five days each week. Each night, the system polls store level inventory for those stores that will be shipped the following day, and identifies which stores need which skus. The system then creates a 'Store Order' for any skus to be picked and shipped the following day.
Historically, due to the many factors and variables present, projects resulting in modest improvements early in the Supply Chain have not been viewed as having a quantifiable impact at the retail store level. Using Monte Carlo simulation, a team models and presents the ranges of improvements for Store Level Gross Margin that can be expected from a change in ordering practices.
One of the key drivers of customer perceived value in a Retail environment is whether the retailer is in stock for the item that the customer is looking to purchase. Measuring In Stock position is relatively easy using the Inventory Management System gives a good indication of In Stock, and this can be augmented with store associate and Mystery Shop audits to identify discrepancies between system in stock and true in stock positions.
However, identifying Out of Stock conditions (when store on hand drops to zero) is only the first step. A particular Out of Stock condition can arise from any of a number of process failures throughout the Supply Chain. Item sales at a particular store may oversell forecast and drain down safety stock. Store level shrink may cause discrepancies between system inventory position and actual in stock, delaying store replenishment. Store level execution can cause an item to be in the store but not on the sales floor. Or, Supply Chain failures may result in an item not being available for replenishment. Measuring and analyzing all of these possible root causes, across a network of thousands of stores, across thousands of different items and millions of transactions can be challenging to a project team to say the least.
The Challenge
Improving performance is good, but what does it mean for our customers? One of these Project Teams was chartered at a national retailer to improve a metric called 'Retail Insufficients'. This retailer has approximately 1500 stores, serviced from a handful of Distribution Centers. The inventory replenishment system is setup to supply the stores on a Pull basis, and identifies each individual item by a Store Keeping Unit, or SKU number. The stores are set on a pick schedule at the Distribution Centers to be delivered anywhere from one to five days each week. Each night, the system polls store level inventory for those stores that will be shipped the following day, and identifies which stores need which skus. The system then creates a 'Store Order' for any skus to be picked and shipped the following day.
Typically, these store orders are fulfilled at the Distribution Center, and shipped to the store on the next delivery. Sometimes, though, there is not sufficient inventory at the Distribution Center to fulfill the store order. When this occurs, it is measured as a Retail Insufficient. These Insufficients are then further classified by cause, one of which is an Order Failure Insufficient, in which the sku is not available due to a failure in the Ordering Process.
Management recognized that this may be a key factor in Retail Out of Stock, and chartered a team to analyze and reduce the occurrence of Retail Insufficients due to Order Failures using the Lean Six Sigma DMAIC methodology.
Management recognized that this may be a key factor in Retail Out of Stock, and chartered a team to analyze and reduce the occurrence of Retail Insufficients due to Order Failures using the Lean Six Sigma DMAIC methodology.
One of the first challenges to the team was to quantify the potential benefits of reducing the insufficient rate. Historically, the business owners had managed to this metric and worked to reduce it, but there had not been a direct linkage made between this metric and overall business performance. In general, it is known that reducing insufficients is a good thing, but what is the optimal level of insufficients that should be expected to balance the value of being in stock for the customer, and reducing the on hand inventory at the store and DC?
To truly understand the customer level impact of a Retail Insufficient, the team needs to understand:
-How often do insufficients occur across skus, categories, market segments, etc?
-If an insufficient occurs, how often does a particular store run out of stock as a result?
- If a particular store runs out of a particular sku, will there be demand for this sku during the out of stock condition?
-If an insufficient occurs, how often does a particular store run out of stock as a result?
- If a particular store runs out of a particular sku, will there be demand for this sku during the out of stock condition?
- If there is demand, will the customer step up or down to a substitution sku, or will he or she simply walk away without making a purchase?
- If the customer makes a substitution purchase, what is the margin impact of this substitution compared to if the original sku had been in stock?
To better understand this chain of inter-related questions, the team turned to a Monte Carlo simulation using the Crystal Ball software.
The team first created a mathematical representation of the benefits that would be realized by reducing out of stock conditions. When an out of stock condition results in a lost sale, the company will lose the gross margin related to that sale. Thus, the economic benefit of reducing out of stock will be shown through a reduction in lost gross margin. The project benefit can be calculated as shown below, where the subscript represents the insufficients performance prior to the project, and is the insufficients performance expected after the project completion.
Note that in some cases, the GM of a substitution may in fact be higher than the original sku, and thus an Out of Stock condition could result in an economic benefit to the company.
The team then designed a data collection plan to gather representative samples to develop distributions for each variable across the spread of categories. A category is simply a set of related skus, such as Candy or Clothing. By analyzing sales, out of stock and substitution data the team was able to both build distributions for, and develop correlations between, the variables that drive the above equations. The team then built a Crystal Ball spreadsheet that models the current state lost gross margin and the future state lost gross margin, with the project benefit being the delta between the two.
James Hull is a business process manager at Staples with 18 years of experience in continuous improvement across the military, manufacturing and service sectors. He can be contacted at [email protected]. Hani Noshi is a senior consultant and Lean Six Sigma Master Black Belt in Accenture's Strategy Practice, Operational Excellence Capability. Based in Greenville, SC, he can be reached at [email protected]