Sei sulla pagina 1di 38

Supply Chain Analysis Tools

MS&E 262
Supply Chain Management

Im indebted to Dr. Mark Wilkinson, Senior Industrial Engineer and Intel


Technologist, Logic Technology Development, Intel Corporation for initial
organization of this PPT deck.

Motivation

Inventory/Service Trade-off Curve


High

Inventory
Low
Good

Poor

Service
Analytical tools help lower the curve!
Hausman and Wilkinson

Outline
Analytical Tools from MS&E 260/261
EOQ

Newsvendor
Lot Size Reorder (Q,R) Model
Periodic Review (T,S) Model

Supply Chain Improvement Tools

Hausman and Wilkinson

Standard Inventory Models


Case I

Case II

Case III

Demand: Known
Supply: Known

Demand: Unknown
Supply: Known

Demand: Unknown
Supply: Unknown

Ordering vs. Holding

Ordering vs. Holding

Underage/Overage

Underage/Overage

Trade-off:

Ordering Cost
Holding Cost
Penalty Costs:
None

Hausman and Wilkinson

Standard Inventory Models

Economic Order Quantity (EOQ)

Case I

Deterministic; determine order quantity

Newsvendor

Case II

Captures demand variability; determine one-time-shot quantity

Lot Size-Reorder Model

Case II/III

Includes variability in an on-going setting


Determine order quantity (Q) when an trigger level (R) is met

Periodic Review Model

Case II/III

Includes variability in an on-going setting


Determine order-up-to level (S) when ordering every T time units
Hausman and Wilkinson

Analytical Tools

Economic Order Quantity Model

How much to order/produce?


Fixed order cost of $ K
Inventory holding cost of $ h = $ Ic
Shortages prohibited
Deterministic (constant) demand rate per year, D

Inventory Level

Slope = -D

Q
T
(T = Q/D)

Time, t
Hausman and Wilkinson

Analytical Tools

EOQ Model - Derivation

G(Q) = cD + KD/Q + IcQ/2


Purchasing

Holding

Setup

Annual
Holding +
Setup Cost
Total = G(Q)-cD
Holding = IcQ/2
Setup = KD/Q

Q*

2 KD
Q
Ic
*

Hausman and Wilkinson

Analytical Tools

EOQ Model Sensitivities/Shortcomings

40 20 02 rule
40 % error in an input parameter results in 20 % error in
Q
The result is a 2 % increase in the costs, G(Q)
Cost function is relatively insensitive to errors in Q

Shortcomings of EOQ Model?


Zero lead time (easily extended to fixed lead time)
Infinite production rate (finite production rate, P, if P > D)

No shortages allowed (easily extendable)


Constant, deterministic demand rate

Hausman and Wilkinson

Analytical Tools

Newsvendor Model
Motivation:
At the start of each day, a newspaper vendor must decide
on the number of papers to purchase and sell. Daily sales
cannot be predicted exactly, and are represented by a
random variable, D.

Relevant Costs:
Co = unit cost of overage (not enough demand)
Cu = unit cost of underage (too much demand)

Example: Fashion (Ralph Lauren, Ann Taylor)


Retail: Stanford Shopping Center, Web Sites
Overage: Gilroy Outlet Mall
Hausman and Wilkinson

Analytical Tools

Newsvendor Model (cont.)

cu
F (Q )
cu co
*

Shortcomings of Newsvendor Model?


No consideration of positive lead times

One shot model (can be extended to multiple periods)


No setup cost for placing orders included

Hausman and Wilkinson

10

Analytical Tools

What is the Expected Profit?


Expected Demand = Expected Sales + Expected Lost Sales
Expected Lost Sales = L(z) s when demand follows normal distribution
L(z) is the standardized loss function for a Normal(0,1) distribution
Excel: L(z) = NORMDIST(z,0,1,False) z(1-NORMSDIST(z)) *

Order Quantity (Q) = Expected Sales + Expected Overage


Expected Overage = max (0, Q Expected Sales) no back orders
Expected Profit

* Nahmias 6th ed, p. 278

= (p - c) Expected Sales (c - s) Expected Overage


= (p - c) (m L(z) s) (c - s) (Q (m - L(z) s))

Hausman and Wilkinson

11

Lot Size Reorder Point (Q, R) Model

We need to decide two things:


How much to order each time we place an
order (Q)?
At which inventory position (R) do we order?
Inventory
Position
R

t
t= Lead Time

Time
Safety Stock
Hausman and Wilkinson

12

Safety Stock
Since demand (and possibly supply) is uncertain, the
system builds in safety stock to protect against uncertainty
during the reorder lead time.
Safety Stock = Insurance associated with uncertainty

Safety stock in the (Q,R) model is a function of:

Demand distribution (standard deviation of demand)


Desired availability (e.g., 99% probability demand satisfied)
Suppliers lead time duration
Suppliers availability and delivery randomness (Case III)

Hausman and Wilkinson

13

Analytical Tools

(Q,R) Model Notation:

Average demand rate/year l


Setup cost
K
Variable cost
c
Holding cost
h=Ic
Order quantity
Q
Reorder point
R
Lead time
t
Safety stock
s

Hausman and Wilkinson

14

Analytical Tools

(Q,R) Model Unit Shortage Cost p

The expected total annual cost (excluding lc) is

Kl
Q
l

G(Q, R)
Ic R lt p ( x R) f ( x)dx
Q
2

Q R

Setup

Holding

Shortage

Defining n(R) = E [# of units short in a cycle], we obtain

2l ( K pn( R ))
Q
Ic
QIc
F ( R) 1
pl
Hausman and Wilkinson

15

Analytical Tools

(Q,R) Model Unit Shortage Cost p (cont.)

For normally distributed


demand, define

L( z ) (t z ) (t )dt

where z = (R-mL)/sL.

Hence

n(R) = sL L(z)

Use the approximate solution:


Q = EOQ
Obtain z from tables for L(z)
R = mL + zsL
mL: mean lead time demand; sL: standard deviation of lead time demand
Hausman and Wilkinson

16

Analytical Tools

(Q,R) Model Service Level Approaches

Type 1:

Bad

a = Prob(no stockout in lead time)

Type 2:

Better

b = Proportion of demand met from on-hand stock


Recall n(R) = E[# of units short in cycle]
n ( R ) s L L( z )

1 b
Q
Q

L( z )

(1 b )Q

sL
Hausman and Wilkinson

17

Periodic Review (T, S) Model

Review every T time units


Order such that system inventory position
reaches S units at every review
Response Time = T + t
Inventory

Lead Time

Time

t
t

Hausman and Wilkinson

18

Analytical Tools

(T,S) Model (cont.)


T = EOQ/l

S = mt+T + z st+T
Where
mt+T = mean demand over t+T periods
st+T = standard deviation of demand over t+T periods
z
see (Q,R) model

Hence

Safety Stock = z st+T


Hausman and Wilkinson

19

Analytical Tools

Additional Tools/Methods

ABC Analysis
Exchange Curves
Forecasting
Scheduling
Aggregate Planning
Capacity Expansion
Optimization (LP, IP, MIP, DP)

Hausman and Wilkinson

20

Supply Chain Improvements

Uniform vs. Non-uniform Service Levels


Risk Pooling/Consolidation
Multi-Echelon Analysis
Postponement
Leadtime Reduction
Review Period Reduction
Variable Lead-time

Hausman and Wilkinson

21

Supply Chain Improvements

Example 1: Risk Pooling/Consolidation

What is meant by Risk Pooling?

Example
Laser Printer Supply Chain

Hausman and Wilkinson

22

Supply Chain Improvements

Laser Printer: Finished Goods Logistics


Penang, Malaysia

Long Beach
CA, USA
Memphis
TN, USA

Represents a DC location for distributor D1


http://www.ups.com/maps

Hausman and Wilkinson

23

Supply Chain Improvements

UPS Ground Map for Memphis, TN

http://www.ups.com/maps

Hausman and Wilkinson

24

Supply Chain Improvements

Laser Printers Distributor Network


Assume the following distributor network:
5 Independent Distributors (D1, D2, D3, D4, D5)
Each distributor operates 8 DCs across the US

Who are the distributors customers?

Relevant metrics for a DC?

Hausman and Wilkinson

25

Supply Chain Improvements

Active Learning

Discuss opportunities for risk pooling:


For any particular distributor?
For any particular location (e.g., Memphis,
TN)?
For the original equipment mfr (OEM)?

Hausman and Wilkinson

26

Supply Chain Improvements

Laser Printers Distributor Network


Assume that:
1.

Demands at the multiple DCs are statistically


independent.
2. The means and standard deviations of demand for
the multiple product DCs are identical.
3. The leadtimes for the multiple DCs are
identical/constant.
4. The review periods at the DCs are identical.
5. The safety factors for the DCs are identical.
6. All DCs have the same inventory value.

Hausman and Wilkinson

27

Supply Chain Improvements

Laser Printers Distributor Network

Let:
si
demand
DC i;

= standard deviation of
per period at

ti

= lead time for DC i;

Ti
zi

= review period for DC i;


= safety factor for DC i;

= number of DCs in a region


(e.g., DCs in Memphis, TN).
Hausman and Wilkinson

28

Supply Chain Improvements

Laser Printers Distributor Network


Safety Stock at DCi = zis Ti t i
Total System Safety Stock =

i 1

zis Ti t i

System Safety Stock, Unpooled = nzs T t

System Safety Stock, Pooled =

n zs T t
Hausman and Wilkinson

29

Supply Chain Improvements

Laser Printers Distributor Network


Reduction Effect through Pooling:
SafetyStock,UnpooledTotal

- SafetyStock
,
PooledTotal

SafetyStock,UnpooledTotal

nzs T t - n zs T t
nzs T t

1
1
n

E.g., Reduction Effect in Memphis for n = 5: 55.3%

Hausman and Wilkinson

30

Example 2: Lead Time Reduction

Hausman and Wilkinson

31

Supply Chain Improvements

Example 3: Postponement

Concept
Delay/postpone product differentiation until as
late as possible in the production process

Blanks
Manufacturer

SF
Warehouse

Intel
Mask Facility

a
B

Ba

Ba

Bb

Bb

B
Hausman and Wilkinson

32

Before Postponement
Delivery

Engineering

Supplier-coated blanks
B

t = 4 weeks

Daei
Ba
Production

Dapi
b

Engineering

Supplier-coated blanks

t = 4 weeks

Dbei
Bb
Production

Demand for blank B in week i =


resist

user

Drui

Dbpi
Hausman and Wilkinson

33

After Postponement
Delivery

In-house
Coating

Engineering

Daei

a
Ba

Production

t = 1-2 days

Dapi
Engineering

Uncoated blanks
t = 3 weeks

Dbei

b
In-house
Coating

Bb
Production

Dbpi
Hausman and Wilkinson

34

Summary

Use known tools as needed in projects


Understand underlying assumptions
Perform sensitivity analysis on different
parameters
At worst, simulate the system!
Analytical tools can help significantly improve
supply chain performance!

Hausman and Wilkinson

35

Details:

Hausman and Wilkinson

36

Analytical Tools

Lead Time Demand Variability


Expectation of Sum = Sum of Expectations
General Variance Formula (Lead time = t periods, demand in
period i = di)
t

s d2 s d2 2 COV (d i , d j ), where d LT d i
LT

i 1

i j

i 1

Variance of Sum = Sum of Variances (for independent variables)

Example: Lead time demand*


Mean
Variance
* assuming

mL = t m
sL2 = t s 2

independence between periods

Hausman and Wilkinson

37

Analytical Tools

Random Lead Times


If lead time is random, with mean t and variance s2
And demand in time t has mean mt and variance s 2t

Then the demand during (random) lead time has

Mean

mL = tm

Variance

sL2 = m 2 s2 +t s 2

*Assuming

orders do not cross and successive lead times are independent


Hausman and Wilkinson

38

Potrebbero piacerti anche