Multiechelon Inventory Optimization White Paper en Us
Multiechelon Inventory Optimization White Paper en Us
Multi-Echelon
Inventory Optimization
Overview
Contents
Overview 1
Managing Inventory in
Single-Echelon Networks
2
Managing Inventory in
Multi-Echelon Networks
Background
Challenges
10
Solution
11
Benefits
12
Conclusion 12
The network carries excess inventory in the form of redundant safety stock.
End customer service failures occur even when adequate inventory exists
in the network.
This paper will examine alternative approaches for tackling the thorny problem
of managing inventory in a multi-echelon network and will present a method
for minimizing inventory across echelons while simultaneously meeting all of
your customer service goals.
The complexities of managing inventory increase significantly for a
multi-echelon distribution network with multiple tiers of locations (i.e., a
network comprising a central warehouse and downstream customer-facing
locations). All locations are under the internal control of a single enterprise.
Instead of simply replenishing the warehouse or the DCs that sit between your
supplier and your end customers, as in the single-echelon situation, you also
need to contend with the problems of replenishing another distribution point
between your supplier and your DCs. For the purposes of this paper, we will
refer to this additional distribution point as an RDC. The objective of multiechelon inventory management is to deliver the desired end customer service
levels at minimum network inventory, with the inventory divided among the
various echelons.
Note: We use the terms RDC and DC to distinguish the entities in the
separate echelons. For a manufacturer or wholesaler, the terms hub
and spoke could be equally applicable. For a retailer, the terms central
warehouse and store might be relevant. For clarity, we will focus on the
interaction between two echelons. The logic is easily extended to more than
two echelons through recursion. In addition, we will use the term enterprise
to refer to the distribution entity as a single, cohesive organization.
Managing Inventory in Single-Echelon Networks
Before delving into the difficulties of managing the inventories in more than
one echelon, let us review the single-echelon problem. In this environment, the
distribution network is Supplier to DC to Customers.
Inventory
Description
Demand
Demand Variation
Lead Time
Expected time delay between ordering and having new product available
to fulfill demand
Frequency with which the DC checks its inventory position to see if a new order
is needed
The DCs time supply objective, which depends on the economic trade offs
between carrying inventory, handling, transportation and purchase cost.
Inventory Position
The DCs available stock, taking into account the on-hand inventory, on-order
quantities, back orders and committed stock
Table 1 describes the inventory drivers for a SKU that is located at a DC.
In this single-echelon situation, the lead times are between the DC and its
external supplier. The enterprises order supply strategies depend on its
internal cost factorssuch as those associated with handling and carrying
inventoryand the external suppliers ordering constraints and bracket
discounts. For this reason, the replenishment quantities depend on a
combination of internal and external factors.
Managing Inventory in Multi-Echelon Networks
Now consider the same product in a multi-echelon network that includes
an RDC between the suppliers and the DCs. The same inventory drivers
described in the preceding section apply for the SKU at the RDC location.
However, some significant issues emerge:
What is the proper measure of demand to the RDC, and how should this
demand be forecasted?
ow does the trend toward larger orders from the RDC to the supplier
H
affect the order supply strategy for the RDC SKU?
hat is the optimal service level goal between the RDC and its
W
customers, which are the DCs?
ow do you factor the individual DCs inventory positions into the RDC
H
replenishment decisions?
hen faced with a limited supply situation at the RDC, how should you
W
allocate product down to the DCs?
Because the RDC also stocks inventory for the SKU, the replenishment decision
at each DC also must address some new questions because of its relationship
with an internal supplier:
ow will the ordering constraints imposed by the RDC (not the supplier)
H
influence the DCs order supply strategy?
o achieve the targeted service level commitment with its end customers,
T
should the DC use the same service level goal when the RDC is available as
a backup source for end customers?
o the external supplier lead time and lead time variation still play a role in
D
the DCs replenishment strategy?
Figure 1 illustrates how the inventory drivers are linked in the two echelons.
Circular nodes represent locations that are under the control of the one
enterprise. The node labeled DC stands for all the DCs that stock the
same product. The inventory drivers, denoted in red, are controllable by the
enterprise. That is, the replenishment review frequency, the order supply
strategy and the service level goal are replenishment control variable that a
decision maker can set to influence the amounts of inventory to be carried
and the service levels offered to downstream customers. The approaches for
setting these control variables in the single-echelon case are well known, but
how should you set the control variables in the multi-echelon case?
Today, enterprises typically use one of two approaches:
1. Apply the single-echelon approach to each echelon in the network.
2. Use a distribution requirements planning (DRP) approach or a variation.
SingleEchelon
MultiEchelon
Supplier
Supplier
Lead Time
Lead Time
RDC
Inventory Drivers
Lead Time
DC
Inventory Drivers
Demand
Customer
DC
Inventory Drivers
Demand
Customer
Supplier
Lead Time
RDC
Inventory Drivers
RDC Replenishment
Optimization
Lead Time
DC
Inventory Drivers
DC Replenishment
Optimization
Demand
Customer
Figure 2: Sequential Approach
L ack of visibility down the demand chain. When the RDC replenishes itself,
it is oblivious to customers beyond those of the DCs. In addition, the RDC
has no visibility into the DCs inventory balances or demands.
emand distortion from the bullwhip effect. Because the RDC and the DCs
D
develop independent demand forecasts (based on their immediate customers
demands), the bullwhip effect causes increased demand variation between the
RDC and the DCs. This results in unnecessary inventory at the RDC.
otal network costs remain unevaluated. When the RDCs or the DCs
T
replenishment strategy changes (i.e., by altering one of the controllable
inventory drivers), the cost implications of the new strategy on the other
echelon are not considered. The focus is strictly on the impact on the
echelon at hand.
The DRP approach has several major shortcomings. The No. 1 weakness is
the deterministic perspective vis--vis pass-up demands and lead times. An
immediate consequence of this is the subjective way in which the RDC safety
stock is usually determined. Because the requirements passed up to the RDC
include no uncertainty, there is no rigorous method for determining safety
stock. This is why enterprises that use this replenishment method generally
use rules of thumb for the RDC safety stock; this unscientific approach leads
to excess inventory. It is not surprising that safety stock determination is
somewhat looseDRP has its roots in manufacturing, where production and
transportation costs are of greater concern than inventory costs. Like the
sequential approach, DRP fails to exploit visibility up the demand chain and
lacks a network view of inventory optimization.
In particular, there is no linkage between the safety stocks in the two echelons.
Therefore, any attempt to optimally balance inventory between the two
echelons is impracticable.
Assessing the Sequential and DRP Approaches
Both the sequential and the DRP approach result in excess inventory without
necessarily improving service levels to the end customers. Although each
echelon might be able to achieve reasonable resultsfrom its own myopic
viewpoint of the total problemthe results are not necessarily the optimal
solution for the total network. That is, the total inventory of the RDC and at the
DCs is not minimized in the pursuit of end customer service objectives.
ccount for all lead times and lead time variations. In each echelon, the
A
replenishment decisions account for lead times and lead time variations of
all upstream suppliers, not just the immediate suppliers.
onitor and manage the bullwhip effect. The enterprise measures the demand
M
distortion and determines the root causes for possible corrective actions.
nable visibility up and down the demand chain. Each echelon takes
E
advantage of visibility into the other echelons inventory positionswhat is
on hand, on order, committed and back ordered. At the DCs, this negates
any need for shortage gaming. At the RDC, visibility into DC inventories
improves projections of demand requirements.
Key Areas
Sequential Approach
DRP Approach
Optimization
Objective
Demand
Forecasting
Pass-up demands or
projected orders with no
measures of their variables
Lead Times
Bullwhip Effects
Ignored
Ignored
Network Visibility
Immediate downstream
customers demands and
immediate upstream suppliers
lead times - myopic view of the
network
Order
Ignored
Synchronization
Between Echelons
Differentiated
Not possible
Customer Service
Not possible
Not possible
ffer differentiated service levels. The RDC can provide different service
O
levels (for the same product) to different DCs. A multi-echelon approach
makes this possible, because the enterprise controls how and when a
product enters and leaves the RDC.
Supplier
Lead Time
RDC
Inventory Drivers
Lead Time
Network
Replenishment
Optimization
DC
Inventory Drivers
Demand
Customer
Figure 3 illustrates how all of the inventory drivers in both echelons are
factored into the DC and RDC replenishment decisions to truly optimize the
networks inventory.
Challenges
NWDCos end customers have very high service requirements. One way
NWDCo can assure that it meets these high service goals is to carry extra
inventory at the DCs. At the RDCs, inventories are carried to replenish the
DCs and to satisfy the RDCs external customers. The problem is to deploy
the optimal balance of inventory in the two echelons. NWDCo can use many
control levers to shift the inventory balance between its RDCs and DCs. How
should NWDCo set these to achieve service objectives at minimum inventory?
Can slow-moving products simply be centralized at the RDCs with little or no
product at the DCs?
The inventory management problem is further complicated because of the
high service requirements of the RDCs external customers. The RDCs must
simultaneously set aside enough inventory to satisfy external customers
demands and to provide the DCs with enough product on hand to satisfy the
DCs customers demands.
An overriding question NWDCo must address is how to come up with the right
demand forecasts at the RDCs. These forecasts are crucial to replenishment
decisions at the RDCs. Should the DC demands simply be passed up to the
RDCs for use in determining statistical forecasts?
Should the RDCs simply aggregate demand forecasts that are determined at
the DCs? Should the RDCs develop statistical forecasts based on historical
shipments to the DCs? What about historical orders from the DCs?
Other challenges exist. NWDCo needs to attain and fully exploit information
transparency in its network. This means the RDCs and the DCs must freely share
information on demand, inventory, in-transit orders and factors that impact
demand, such as promotions activity. Another problem is to synchronize the
replenishment activities between the two echelons to minimize lead times and
to balance orders between DCs and RDCs over the days of the week. Failure to
adequately tackle these challenges will lead to increased inventory and potential
service failures because of the bullwhip effect.
10
Solution
NWDCo implemented a true multi-echelon inventory management system
in phases, with each phase incorporating additional vendor lines. All
replenishment decisions now exploit network visibility. The system models
all cost implications of replenishment strategies at the DCs on the RDCs
and vice versa. There are direct linkages between inventory drivers in one
echelon and inventory levels in the other echelon. As an illustration, Figure 4
depicts how NWDCo sets one of the control variables, the RDC to DC service
level goal, to minimize the total inventory at the RDC and DCs. As expected,
the RDC inventory increases as this parameter increases. Less obvious is the
effect on DC inventories. As the expected fill rate from the RDC becomes
larger, the total inventory at the DCs decreases. The optimal setting for this
parameter is the value that minimizes the total inventory in both echelons.
Obtaining this optimal setting requires careful modeling of the interactive
effects on both the RDC and DC safety stocks. NWDCo follows this
optimization process for every RDC item.
60,000
57,500
55,000
Optimal
52,500
50,000
40,000
35,000
30,000
DC Inventory
25,000
20,000
15,000
RDC Inventory
10,000
80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97
11
99% service
99%+ service
Days Inventory
20
15
RDC
RDC
10
5
0
DC
DC
Suboptimal
Optimal
Benefits
By using a true multi-echelon approach, NWDCo makes optimal replenishment
decisions for all SKUs across every echelon. This enables the enterprise to
attain or exceed service goals while driving down inventory requirements in
the total network. Figure 5 illustrates a typical result for a class of SKUs with
average velocity. Inventory is properly balanced between the two echelons
to reduce network inventory by more than three days while maintaining or
exceeding 99% customer service goals. Reduced inventories free up working
capital for the enterprise. Besides increasing cash flow, the increased cash
helps to permanently raise earnings for NWDCo. At the same time, higher
service levels generate increased revenue and customer satisfaction.
Conclusion
A multi-echelon distribution network presents many opportunities for inventory
optimization that the enterprise must pursue to offset potential increases in
transportation, warehouse and occupancy costs. The key to achieving those
savings is to use a true multi-echelon strategy to manage inventory. It is not a
simple task to pursue such a strategy because of the multiplicities of inventory
drivers and the complexities in modeling the interactions of the drivers between
echelons. Nevertheless, the benefits are worth the effort. Taking the right
approach can yield rewards on both sides of the inventory equationbetter
customer service with less inventory. Using a true multi-echelon approach is the
ultimate win-win strategy for inventory management.
12