Multi-echelon Inventory Optimization
Many companies are tasked with managing their inventory in a multi-echelon distribution network, which introduces additional levels of complexity over the traditional single-echelon environment. In a single-echelon environment, there is typically only a single stocking point between supplier and customer, as shown below:
Supplier -> Warehouse -> Customer
A multi-echelon environment is differentiated by at least one, and maybe more, stocking points between supplier and end customer, as shown below:
Supplier -> Regional DC/Warehouse -> DC/Warehouse -> Customer
In the multi-echelon environment, stock transfer management and communications must be not only outward-facing, to the supplier and end customer, but inward-facing between a regional distribution center (RDC) or central warehouse and a local DC or warehouse.
This additional layer of management introduces a level of complexity not efficiently managed by traditional distribution requirements planning (DRP) systems, or by systems that consider each location as an isolated ‘island’ of inventory with their own stock and service level parameters.
Multi-echelon inventory optimization (MEIO) seeks to move beyond the limitations of single echelon inventory management by considering multiple points in the supply chain as a whole when performing planning and replenishment runs. MEIO is still in the early stages of adoption for many companies, due to its complexity and the requirement for looking at data in a new way. MEIO’s adoption is mainly being pioneered by distribution-intensive companies with possibly thousands of local, national and even international nodes in the overall network.
At this stage of maturity, MEIO is most often implemented in a blended model in which parts of the overall supply chain are managed as a group (multi-echelon planning), while many other parts of the system are still using the traditional single echelon model. Due to its complexity, MEIO most often requires a phased, transitional approach in moving selected parts of the business over as a group, instead of betting the business on a total/global cutover.
In many respects, the overall themes and issues faced by distribution companies when implementing MEIO resemble those faced by manufacturing companies when implementing lean manufacturing and similar techniques. At the highest level, those techniques required a paradigm shift in no longer looking at work centers or production lines as isolated entities, but linked parts of a greater whole. A key learning was that separately optimizing production rates, queues, etc. for each single entity led to the sub-optimization of the overall system.
Elements of Distribution Inventory Management
While the specific importance may vary, in a standard distribution environment standard variables that need to be managed for any specific location often include:
- Response to demand (make to stock or make to order)
- Variability of demand
- Standard lead times
- Variability of lead times
- Number of standard and alternate supply sources
- Number of customers and individual share of total sales
- Product substitution policies
- Service level policies
- Inventory carrying costs
- Transportation network variability
In the single echelon environment, analysis for each of these variables results in a defined policy for a specific material, or material/vendor and material/customer combination. That policy results in the creation of parameters for lead times, safety stock levels, forecast model and other elements used by the system in suggesting new stock orders and adjustments to existing orders.
So what's different in a multi-echelon environment? Each of these variables may still exist at a specific location; the key differentiation is that they need to be managed as a whole for multiple location.
For MEIO to enable a supply chain to increase on-time delivery performance without also increasing stock levels to an undesirable level, visibility up and down the chain is a key requirement. The RDC needs to not only consider the demand from its immediate customer (the DC), but also the end customer receiving the product. Likewise, supplier information is required not only for the RDC but the DC.
In setting system parameters, several key areas need to be addressed as a whole instead of making isolated decisions by location. Some of these areas include:
Inventory stock levels: Where will you hold stock- at the RDC, the DC, or both? What is the lead time and variability between the RDC and DC, and between the DC and end customer? If you set separate safety stock levels for both entities, the combined total is almost guaranteed to drive overall levels above company targets.
Demand management: Where will you forecast demand- the RDC, DC or both? Will demand be triggered from the RDC to DC only based on a confirmed stock transfer, or do lead times require a forecast? If you forecast at both locations for the same overall end demand (transfer from RDC to DC and subsequent shipment to end customer), even world-class forecasting systems introduce a level of error that will suggest more safety stock than company targets may accept.
Demand prioritization: Does the RDC supply both DC’s and end customers? If so, does the end customer get the priority? Does one DC have priority over another? Can the RDC account for specific end customer priorities (DC to customer) when the system makes replenishment suggestions to each DC? How is product allocated in a shortage situation- fair share, order date, customer priority, or other (considering both DC and end customer)?
Push vs. pull management: How are stock transfers reviewed and initiated in the network- by the supplying location or the receiving location? If stock is pushed, not pulled, who is responsible for service levels, stockouts, etc. at the receiving location?
As with any other large-scale supply chain initiative, MEIO is not a magic bullet that will solve all replenishment issues up and down the chain. What it does provide is a framework and methodical approach to a holistic view of the chain. When safety stock levels are set, the best levels consider the past history and future total effect of the parameters on multiple locations, not just one. The same applies for forecasting and other policies. MEIO may result in a shift in job responsibilities and performance measurements, which again need to be considered by management as a whole instead of in isolation.