Safety Inventory

-- Chapter 11

Managing Uncertainty in a Supply What is safety inventory?

The average inventory remaining when the replenishment lot
Chain: Safety Inventory arrives

Why need safety inventory?

To manage
demand uncertainty
Supply uncertainty

Raising the level of safety inventory

increases product availability
increases inventory holding costs

11 - 1 11 - 2

(A) With variable demand (B) With variable lead time

Inve nto ry
Inve ntory
20 0 Ave rage
200 Ave rage c yc le
cycle inve nto ry
inve ntory

{
10 0

{
100

Ave ra g e

{
Ave ra g e in ve n to r y
10 12 20 30 40

{
in ve n to ry (1 4 0 ) S a fe ty
(1 5 0 ) S a fe ty 8 10 20 30 40 s toc k
s tock (40 ) Day s
(50) Days

The lead time can vary by plus or minus two days

The average inventory would be 100 units when demand (20/d) and
lead time (10day) were constant
If demand was actually 25/d, inventory would be depleted by 8 (200/25).
And there would be a stockout of 50 units.
A safety stock of 50 units would prevent a stockout due to variation in demand
11 -
When the order arrived 2 days late, on day 12, the firm would experience
stockouts for 2 days (40 units)
If the management believed that shipments would never arrive more than
two days late, a safety stock of 40 units would ensure a stockout would not
occur. This would require holding an average inventory of 140 units
3 11 - 4
(C) With variable demand and lead time Safety Inventory
Inve nto ry

20 0 Two key questions

Ave ra ge
c yc le What is the appropriate level of safety inventory to carry?
inve nto ry
What actions can be taken to improve product availability while

{{
10 0
reducing safety inventory
Ave ra g e
in ve n to r y
8 10 12 20 30 40
A key to success
(2 0 0 ) S a fe ty
s to c k Day s To decrease the level of safety inventory carried without hurting
(10 0) the level of product availability
- Successful stories : Dell (comparing with Compaq), Wal-Mart, 7-11
Japan
The demand is above forecast by the maximum, 25 units and the incoming
order arrives two days late Factors influencing safety stocks
The result is a stockout period of four days at 25 units per day Forecast error (demand uncertainty)
If management wanted to protect against the maximum variability in both Lead time (supply uncertainty)
demand and lead time, the firm would need a safety stock of 100 units.
This policy (no stockout) would result in an average inventory of 200 units Exposure to stockout
Service level required (product availability)
11 - 5 11 - 6

Measuring uncertainties and product availability 2. 3 Basic Inventory Replenishment Polices

Measures of demand uncertainty
Average demand per period, D 1) Continuous review systems (Q, ROP)
Standard deviation of demand per period, D
Order fixed quantity (Q) when total inventory drops below
Coefficient of variation, cv
Reorder Point (ROP)
the ratio of the standard deviation to mean, cv = /
Fixed order size and varied order interval

Measures of product availability 2) Periodic review systems (T, OUL)

Product fill rate, fr Order at fixed time intervals (T) to raise total inventory to
The fraction of product demand that is satisfied from production in inventory Order up to Level (OUL)
Order fill rate Varied order size and fixed order interval
The fraction of orders that are filled from available inventory

Cycle service level, CSL

3) The combination of 1) and 2) (ROP, OUL)
The fraction of replenishment cycles (the interval between two successive fixed reorder point and fixed maximum inventory level
replenishment deliveries) that end with all the customer demand being met
Equals to the probability of not having a stockout in a replenishment cycle

11 - 7 11 - 8
 Continuous (Q)
Review Evaluation of Replenishment Policies
Policy:

Given a replenishment policy, determine

(ROP)
Safety inventory (see Example 11.1)
Cycle service level (see Example 11.2)
Fill rate (see Example 11.3)
Periodic
Review Given desired product availability, design
Policy: (OUL)
replenishment policy
Given desired Cycle Service level, determine safety
inventory (see Example 11.4)
Given desired Fill Rate, determine required safety
inventory (see Example 11.5)
T 11 - 9 11 - 10

3. Continuous Review Policy: Example 11.1-2 - Determine safety inventory and cycle
Determining Safety Inventory and Cycle Service Level service level of a given policy

L: Lead time for Given the replenishment policy

D:
replenishment
Average demand per unit
D L
= DL D = 2,500 /week;

D:
time
Standard deviation of
L
= L D
D = 500
Average lead time: L = 2 weeks;

Given continuous review policy: (Q, POR) = (10000, 6000)

1
demand per period
RL : Mean demand during lead
ss = F S
(CSL) L

L:
time
Standard deviation of
ROP = D L
+ ss ss =
Average Inventory = cycle inventory + ss =
demand during lead time
CSL: Cycle service level
CSL = F ( ROP, D ,
L L
)
Average Flow Time =
=Prob(demand during The demand during lead time, DL =
lead time ROP) The standard deviation of the demand in lead time, L =
ss: Safety inventory Average Inventory = Q/2 + ss Cycle service level, CSL = F(ROP, DL, L) =
ROP: Reorder point
= NORMDIST(ROP, DL, L, 1) =
11 - 11 11 - 12
 Continuous Review Policy:
Evaluating Fill Rate Example 11.3: - Determine Fill Rate of a given policy

Following Example 11.1-2, given

fr : Proportion of customer ESC
fr = 1 DL=5,000, L=707, and (Q, ROP) = (1000, 6000)
demand satisfied from Q
stock ss = ROP-DL = 1000
ESC: the expected ss
ESC = ss{1 F S } ESC = -ss{1-NORMDIST(ss/L, 0, 1, 1)} +
shortage per cycle
L L NORMDIST(ss/L, 0, 1, 0)
= -1,000{1-NORMDIST(1,000/707, 0, 1, 1)} +
ss 707 NORMDIST(1,000/707, 0, 1, 0)
+ L f S = 25.13
L
ESC = -ss{1-NORMDIST(ss/L, 0, 1, 1)} fr = (Q - ESC)/Q = (10,000 - 25.13)/10,000 = 0.9975.
In Excel:
+ L NORMDIST(ss/ L, 0, 1, 0)
Refer to the textbook page 305 for Excel computation.
11 - 13 11 - 14

Example 11.4: - Determine safety inventory and the policy

satisfying desired CSL

Key Point: Given

D = 2,500/week;
D = 500
Both filling rate and cycle service level increase as L = 2 weeks;
the safety inventory is increased. Q = 10,000;
CSL = P(Demand during lead time ROP) = F(RL+ss, DL, D) = 0.90
For the same safety inventory, an increase in lot
size increases the fill rate but not the cycle service DL =
level . L =
ss = FS-1(CSL)L =
ROP =
The replenishment policy: (Q, ROP)=
11 - 15 11 - 16
 Example 11.5: - Evaluating Required Safety Inventory Given F ill R a te S a fe ty I n v e n to r y
Desired Fill Rate 9 7 .5 % 67
If desired fill rate is fr = 0.975, how much safety inventory 9 8 .0 % 183
should be held? 9 8 .5 % 321
9 9 .0 % 499
Step1: ESC = (1 - fr)Q = 250
9 9 .5 % 767
Step 2: Solve (by using GOALSEEK tool in Excel or trial-and-error
method)
ss ss (2% ) ( 11 times )
ESC = 250 = ss 1 F S + L f S
L L
ss
250 = ss 1 NORMSDIST + L NORMDIST
ss
,1,1,0
Key Points:
L L
The required safety inventory grows rapidly with
an increase in the desired product availability
Safety inventory: ss=67
See P308 of the textbook for Excel computation.
11 - 17 11 - 18

Key Managerial Levers

4. Periodic Review Policy
-- to reduce the level of safety inventory required in a way
that does not affect product availability
L: Average lead time
1. Reduce the supplier lead time, L T: Reorder (review) interval
If lead time decreases by a factor of k, the required safety D: Average demand per unit time D T +L
= (T + L ) D
inventory decreases by a factor of k D: Standard deviation of demand per
As the reduction of L requires a significant effort from the supplier whiles unit time T +L
= D
L +T

1
the reduction in ss occurs at the retailer. It is important for retailer to L+T : Standard deviation of demand ss = F S
(CSL ) T +L
share some of the resulting benefits during L+T periods
CSL: Cycle service level
OUL = D T +L
+ ss
Reduce underlying uncertainty of demand, D
2.
= P(demand during L+T OUL) = NORMSINV (CSL ) T +L
If D is reduced by a factor of k, the required safety inventory OUL: Order up to level Q = DT = DT
also decreases by a factor k. ss: Safety inventory
Better market intelligence Q: Average lot size
Better forecast
Better information sharing along the supply chain
11 - 19 11 - 20
Example 11.10: Periodic Review Policy Comparison of the Two Policies
(Lego at a Wal-Mart store)
Comparing Example 11.4 and 11.10
Given Given the same CSL=0.90,
D = 2,500/week; ss = 906 boxes when using continuous review policy
D = 500 ss = 16570 boxes when using periodic review policy
L = 2 weeks;
T = 4 weeks;
CSL = 0.90
Key Points
With a continuous review policy, the safety inventory is used to cover for
What is the required safety inventory ? demand uncertainty over the lead time. With a periodic review policy, the
safety inventory is used to cover for demand uncertainty over the lead
time and the review interval, T+L, in which higher uncertainty must be
Mean demand during T+L periods, DT+L = accounted
Periodic review replenishment policies require more safety inventory than
Standard deviation of demand during T+L periods, T+L = continuous review policies for the the same lead time and level of product
availability
The required safety inventory, ss = Fs-1(CSL) T+L = Periodic review is simple to implement and does not require continuous
track on the inventory
The order up to level, OUL =
Companies may partition their products based on their value. High-value
The Periodic Review Policy: (T, OUL) = products are managed using continuous review policies and low-value
11 - 21 products managed using periodic review policies 11 - 22

Strategies to reduce safety inventory without Impact of Supply Uncertainty

affecting product availability
Manage the component inventory for the Dell assembly
plant at Austin
A reduction of supply uncertainty (Example 11.6) Demand uncertainty: demand during L is Normally distributed
Aggregation reduces demand uncertainty and thus the Supply uncertainty: lead time is Normally distributed
required safety inventory (Example 11.7)
Information centralization (Wal-Mart case) D: Average demand per period
Specialization (Example 11.8)
D: Standard deviation of demand per period
Product substitution
Component commonality (Ex.11.9)
L: Average lead time
Postponement (Dell case) sL: Standard deviation of lead time
D L
= DL
(All using continuous review policies)

2 2 2

11 - 23
L
= L D
+ Ds L
11 - 24
 Example 11.6: - Impact of Supply Uncertainty on
Safety Inventory

Given
Key point:
D = 2,500/day; D = 500, L = 7 days; sL=7 days,
Q = 10,000; CSL = 0.90

DL = DL A reduction in supply uncertainty

L = can help reduce safety inventory
ss = Fs-1(CSL) L = required without hurting product
availability
When sL = 0, ss(units) = 1,695 ss(days) = 0.68
= 1, = 3,625 = 1.45
= 2 = 6,628 = 2.65
= 3 = 9,760 = 3.90
= 4 = 12,927 = 5.17
= 5 = 16,109 = 6.14
= 6 = 19,298 = 7.72 11 - 25 11 - 26

Impact of Inventory Aggregation Accurate Response: The Impact of

on Safety Inventory Inventory Pooling
Which of the two systems provides a higher level of service for a given level
of safety stock?
key questions System A (Decentralized) System B (Centralized)

How aggregation affects safety inventory ?

How supply chain can exploit inventory aggregation to ( D , )
i i
reduce the level of safety inventory required without
hurting product availability ? D ,
C C
( D
)

k k
Aggregate
D; cov( i, j )
C C 2

Demand: D =
i =1
i D
= var( D c ) =
i =1
i
+2
i j

11 - 27 11 - 28
When demand across all regions is independent Example 11.7 A luxury car dealership at Chicago has
cov( i , j ) = ij i j = 0, var( D c ) = i 2 , DC = var( D C ) = i2 4 retail outlets
i i Given at each outlet, Disaggregate Aggregated
RC i Weekly demand mean, D = 25 cars ss ss
i standard deviation, D = 5
0 36.24 18.12
L = 2 weeks
When the demands being aggregated are independent, Targeted CSL = 0.9 0.2 36.24 22.92
the s.d. of aggregated demand is less than the sum 0.4 36.24 26.88
of the s.d. of individual demands Disaggregated option: (assuming ij = 0) 0.6 36.24 30.32
0.8 36.24 33.41
When individual demand is perfectly positively L = 2 =D 1.4142x5 = 7.07
1.0 36.24 36.24
correlated ss = Fs-1(CSL) L = Fs-1(0.9)x7.07 = 9.06 (cars/per outlet)
Total ss required = 4 x 9.06 = 36.24 cars
ij = 1, var( D C ) = ( DC ) 2 = i 2 + 2 cov( i , j ) = i 2 + 2 i j = ( i ) 2
i i j i i j i Aggregated option: (with a central outlet)
RC = i
i DC = 4 x 25 = 100 DC = = 10
When the demands being aggregated are perfectly D
L = L C
D = 2 10 = 14 .14
positively correlated, the s.d. of aggregated demand is the ss = Fs-1(CSL) LC = NORMSINV(0.9)x14.14 = 18.12
sum of the s.d. of individual demands
11 - 29 11 - 30

Impact of Correlation on Benefit From Key points

Aggregation
Aggregation reduces demand uncertainty and thus the required safety
0.6 inventory as long as the demand being aggregated is not perfectly
positively correlated (Examples: Dell, Gateway )
0.5 The square root law:
0.4 If the number of independent stocking locations decreases by a factor of
n, the average inventory is expected to decrease by a factor of
0.3 Benefit n
In general, demand being aggregated is unlikely to be perfectly
independent
0.2
The physical aggregation of inventories in one locations may not be
0.1 optimal due to (Examples: Gap, McMaster-Carr, Amazon.com )
Increased response time to customer order
0
Increased transportation cost to customer
0 0.2 0.4 0.6 0.8 1 There are clear benefits to aggregating safety inventory

11 - 31 11 - 32
 Methods to Extract Benefits of Aggregation Information Centralization
Physical Centralization
Information centralization McMaster-Carr:
Virtual aggregating inventory
Specialization Without physically
centralize all Gap:
Product substitution inventory in one Virtually aggregate inventory across all retail stores, even though the
location inventory is physically separated
Component commonality
Wal-Mart:
postponement
uses information centralization with a responsive transportation
system (to allow store managers to exchange products) to reduce the
Williams Sonoma and Laura Ashley as examples of Physical centralization. Mail order companies
with multiple warehouses as examples of Information centralization. Also Wal-Mart and The amount of safety inventory carried while providing a high level of
Gap allow for Information Centralization cutting the level of safety stock carried. product availability
Bennetton for raw material commonality. The major benefit here is that since greige goods
are to be ordered, the estimate is much more accurate since it aggregates across all colors.
Also refer to the HP Laser Jet example.

11 - 33 11 - 34

The supply chain ability in

Postponement
Component Commonality organizing the production and
distribution of products in such
a way that the customization of
these products is made as close
A very effective supply to the point when the demand is
No component commonality
known as possible
chain strategy to exploit Product A

aggregation and reduce A reactive lever for reduction

of the demand and supply
component inventories Product B
variability SC flow without postponement
Component commonality The ultimate goal is to improve
With component commonality
the flexibility capabilities of
decreases the safety Product A
the firm, and thus to improve
inventory required. The profitability by better matching
marginal benefit, however, supply with demand
Product B
decreases with increasing A powerful concept for e-
commonality Step 1 Storage 1 Step 2 Storage 2
Operation commerce channel
Buffer
SC flow without postponement
Successful examples: HP,
Motorolla, - commonality and postponement

11 - 35 11 - 36
 Postponement is the ability of a supply chain to
delay product customization or differentiation
Postponement implementation
until closer to the time the product is sold

Realignment of processes
Realignment of resources Strategy

Realignment of
infrastructure Cost
Time
Capabilities Quality
Flexibility

Infrastructure Processes Resources

Order taking Human resources

Production & Supplies
Non-customized Dying process Customized warehousing premises
Purchasing
Manufacturing
Production equipment
(customization) information & IT systems
Warehousing
11 - 37 Expedition 11 - 38

Levers to reduce variability of demand and supply

Example: The PC company Postponement

Reactive Substitution
A PC company that manufactures and sells netebook PCs directly to users levers
Specialization
One day ship is guaranteed
A new model has 25 variations (software and RAM) Centralization
Improve Match supply Reduce supply &
Same demand for each of 25 variation - a normal distribution, with an average of profitability and demand demand variability
100/day and a s.d. of 100 Reduction of the demand variability

Proactive Reduction of the delivery time

1. Without postponement:
methods
= 100/day, = 100, service level = 95% Reduction of the supply variability

Total inventory needed at beginning of the day = 25( + 1.65 ) Reduction of the supply time
= 25x(100+1.65x100)=6625unit

Estimating and Managing Safety Inventory in Practice

2. Postpone installing the software and RAM until after the order has been received
Account for the fact that supply chain demand is lumpy
while maintaining 95% of the PCs be shipped that day
Adjust inventory policies if demand is seasonal
= 100/day, = 100, service level = 95%
Total inventory needed at beginning of the day = 25 x + 1.65 25 Use simulation to test inventory policies
= 25x100+1.65x5x100=3325unit Start with a pilot
The inventory level is cut virtually in half !! Monitor service level
If each PC costs $3000, the the reduction in inventory is $9.9 million. Focus on reducing safety inventories
11 - 39 11 - 40
 The Supply Chain of HP
HPs Ink-
Ink-jet Printer:
Estimating and Managing Safety Inventory in
Practice Supplier IC Mfg

Supplier PCAT US DC Customer

Account for the fact that supply chain demand is Supplier Print Mech European Customer
lumpy Mfg FAT DC

Adjust inventory policies if demand is seasonal Supplier Far East DC Customer

Use simulation to test inventory policies

Start with a pilot
Key: IC MfgIntegrated circuit manufacturing
PCATPrinted circuit assembly and test

Monitor service level

FATFinal assembly and test
Print Mech MfgPrint mechanism manufacturing

Focus on reducing safety inventories Concepts:

Component commonality
Factories Centralization
Distributors Aggregation
DC's Postponement
11 - 41 11 - 42

11 - 43