G

“Blessed are the flexible, for they will never be bent out of shape”

Managing Operational Flexibility Under Demand Uncertainty

Dissertation Defense:

Manu Goyal

Chapter 2: Strategic Technology Choice and Capacity Investment under Demand Uncertainty.

Analytically studies the impact of competition on the adoption of product flexibility in an environment characterized by uncertain demand.

Chapter 3: Deployment of Manufacturing Flexibility: an Empirical Analysis of the Automotive Industry.

Empirically tests the findings of the second chapter. Evidence suggests that product and volume flexibility may be linked.

Chapter 4: Capacity Investment and the Interplay between Volume Flexibility and Product Flexibility.

Analytically explores the intertwined nature of volume flexibility and product flexibility.

Chapter 2

Strategic Technology Choice and Capacity Investment under Demand Uncertainty.

PT Cruiser

Chrysler

Town & Country

Honda

Odyssey CR-V

Ford

Ford Freestar Ford Escape

Research Questions …• Does the technology investment decision (flexible vs

dedicated) depend on what competition is doing?• Is the impact of problem parameters different with and

without competition?• Will every firm adopt flexible technology in the

equilibrium?

..and answers• It does• It is• No

The Model

One flexible

or

two dedicatedplants

One flexible

or

two dedicated plants

i j

2y

1y

Two markets

Uncertain Demand Curve

Uncertain Demand Curve

2 2 2

2 2 2 1

,

.

i jQ q q

P A Q Q

1 1 1

1 1 1 2

,

.

i jQ q q

P A Q Q

Decision timeline for each of two firmscompeting in two markets

time

Decide choice of technology, Dedicated (D) or Flexible (F)

fic

ic

Choose capacity

: Cost of flexible capacity per unit

: Cost of dedicated capacity per unit

Decide production qty

for both markets

q1i, q2i

Technology Game

Production Game

Capacity Game

Prices determined as per Cournot competition. Profits gleaned

Demand Curve realized

Flexible Firm: one decision

Ded. Firm: two decisions

The Technology Game

D

F

F

D

Firm i

Firm j

dj

di ,

fj

fi ,

mjf

mid ,

mjd

mif ,

The Technology Game Profits

2

1

2

)1(3

2

y

jiyd cc

)1(16

)(

)1(18

)42( 221

221

ifjm

d

cc

)1(18

2

)1(9

)2(12

22

21

2

1

2

y

fifjyf cc

)1(8

2

)1(32

)(

)1(18

)42(12

22

21

221

221

fijmf

cc

Firm profit in (D,D) market

Firm profit in (F,F) market

Dedicated Firm profit in (D,F) market

Flexible Firm profit in (F,D) market

The Stochastic Effect Profits (symmetric costs and distribution)

D F

D

F

9

2 ,

9

2 22 cc

9

2

9 ,

9

2

9

2222ff cc

9

22 ,

9

22

4

222 cccc ff

2 222 2 2 2

, 9 4 9f fc c c c

9

2

9

22fc

Stochastic Deterministic

Infeasible Region

2

fc

c

critfc

The Best Response Functions - When

Competitor invests in dedicated technology

,,cc f

Dedicated

Flexible

,,cc fM

Monopolist

Flexible

Dedicated

Infeasible Region

2

fc

c

critfc

Dedicated

Flexible

,,cc f ,,cc fM

Monopolist

The Best Response Functions - When

Competitor invests in flexible technology

Flexible

Dedicated

The Nash Equilibrium

(D,D)

2

fc

c

Infeasible Region

(F,F)

critfc

,,cc fM

,,cc f

,,cc f

(D,F) and (F,D)

Pure Flexible Market

Mixed Market

The Nash Equilibrium

Other Effects

• Market size effect– Pulls threshold curves down.– Additional (F,F) and (D,D) equilibrium is

simultaneously possible.

• Product Substitutability Effect.– Amplifies both the stochastic and market size

effects

• The Cost Effect.– Induced by asymmetries in the costs of firms.

(D,D)

(F,F)

(D,F) and (F,D)

(D,D) and (F,F)21

fic ficfMc

2T

c

fc

Equilibrium with market size effect

Other Effects

• Market size effect– Pulls threshold curves down.– Additional (F,F) and (D,D) equilibrium is

simultaneously possible.

• Product Substitutability Effect.– Amplifies both the stochastic and market size

effects

• The Cost Effect.– Induced by asymmetries in the costs of firms.

ic

jc

(D,D)

(F,F)

(D,F) and (F,D)(F,D

)

(F,D)

(F,D)

fic

fic

fjc

fjc

cost

2T

I

V

IV

III

VIVII

The cost effect

fjfi cc

fic

fjc

2T

c

cost

fic

fjc

Summary and Conclusions• The paper covers three levels of firm decisions: strategic

(technology investment), tactical (capacity investment) and operational (production decisions).

• Distilled the impact of competition on the technology choice of firms– Flexibility is more valuable if competitor uses dedicated

technology, less valuable if competitor uses flexible technology– Technology choice decision cannot be made in isolation.– Flexible and dedicated technologies can co-exist in equilibrium.

• The differential Impact (under competition) of:– Product substitution– Market Size– Costs

Chapter 3

Deployment of Manufacturing Flexibility: an Empirical Analysis of the Automotive Industry.

The Hypotheses

• H1: The use of flexibility is associated with higher uncertainty in demand for individual products.

• H2: The use of flexibility is associated with lower demand correlation for individual products.

• H3a: The use of flexibility is associated with a larger number of flexible competitors.

• H3b: Under moderate demand uncertainty, the use of flexibility is associated with fewer flexible competitors.

Hypotheses (cont)..

• H4: The use of flexibility is associated with lower mean demand for products.

• H5: Flexibility is associated with lower difference in mean demand (demand differential) for products.

• H6a: Under high demand uncertainty, the use of flexibility is associated with higher product substitutability in the marketplace

• H6b: Under a low demand differential, the use of flexibility is associated with lower product substitutability in the market place.

The Data

• Primary Sources– Harbour Reports– Ward’s Automotive.

• The “Big Three” US Manufacturers.• Years 1996-2003.• Over 70 manufacturing facilities in North

America.• Unit of analysis is a given plant in a given year

(plant-year combinations, 483 in numbers).

Measures

• Flexibility: “Demonstrated” vs. “Potential”

Assembly Line Flexibility (ALF): 1 if the number of platforms manufactured in a plant is greater than the number of assembly lines, and 0 otherwise.

• Other Ways?

0%

5%

10%

15%

20%

25%

30%

35%

40%

45%

1995 1996 1997 1998 1999 2000 2001 2002 2003 2004

Years

% F

lexi

ble

Cap

acit

y GM

FORD

DCX

Observed Flexibility Over Time

Measures (cont)..

• Demand Uncertainty: Coefficient of Variation of de-seasoned monthly sales.

• Correlation.• Mean demand.• Demand Differential.

• Competition: number of flexible competitors.

• Substitutability. Price difference.

Control Variables

• Plant Capacity

• Plant Utilization.

• Manufacturer dummies.

The Analysis:Descriptive Statistics

Minimum Maximum Mean Std. Dev.

Mean Demand 471 79836 17045 16133

Demand Uncertainty

0.03 1.56 0.45 0.28

Correlation -1.00 1.00 0.3632 .4630

Competition 0.00 7.00 1.35 1.27

Demand Differential

46 511365 31018 62764

Substitution 8 26352 2677 3619

Capacity 33088 327120 208602 61527

Utilization 0.15 1.67 0.91 0.28

Competition Utilization Correlation SubstitutionDemand

Uncertainty

Demand Differ.

Mean Demand

Utilization 0.02

Correlation -0.02 0.08

Substitution -0.17** 0.02 -0.11*

Demand Uncertai

nty0.006 0.33** 0.05 -0.06

Demand Different

ial0.22** 0.13** -0.08 -0.12** 0.29**

Mean Demand -0.07 0.24** 0.10* 0.03 0.56** 0.16**

Capacity 0.24** 0.21** 0.13** -0.19** 0.10* -0.01 -0.14**

Correlations

Univariate Test

Univariate Test of Differences in Means (dependent variable: ALF)

Mean: ALF = 0Mean: ALF

= 1P-value for test of mean difference

Competition 1.25 1.90 0.00

Correlation 0.38 0.24 0.01

Demand Uncertainty

0.44 0.48 0.37

Demand Differential

31150.90 30236.34 0.91

Mean Demand 17538.14 14138.23 0.10

Logit Regression (N=483)Column 1 Column 2 Column 3

Intercept -1.899*** (0.645) -2.00*** (0.704) -2.078*** (0.766)

Competition 0.3243*** (0.098) 0.276** (0.123)

Demand Correlation -0.600** (0.275) -0.598** (0.284)

Demand Uncertainty 1.367** (0.596) 1.612** (0.694)

Demand Differential -0.00 (0.00) -0.00 (0.00)

Mean Demand -0.00 (0.00) -0.00 (0.00)

Substitution -0.00 (0.00)

Substitution × Coefficient of Variation.

0.00 (0.00)

Substitution × Difference of Means.

-0.00 (0.00)

Utilization -0.900** (0.481) -0.907* (0.543) -0.962* (0.550)

Capacity 0.00 (0.00) 0.00 (0.00) 0.00 (0.00)

GM 0.189 (0.356) 0.137 (0.389) 0.162 (0.405)

Ford 0.152 (0.283) 0.331 (0.40) 0.399 (0.416)

Significance 0.22 0.007 0.0047

Logit Regression

• Evidence suggests that plants that are observed to be flexible have:– Higher demand uncertainty (H1).– Lower correlation (H2).– Higher (flexible) competition (H3a).

• Control Variables:– Flexible plants have lower utilization.– No significant differences between the “big three”.

Productivity Analysis

• Study the implications of deploying flexibility, measured against extant theories.

• Hours per Vehicle (HPV) as a measure of productivity.

YEAR 2003 2002 2001 2000 1999 1998 1997

50

40

30

20

DCX

GM

Ford

Mean HPV

HPV over the years

Mismatches..

• Measure mismatch benchmarked against six environmental variables: – Demand Uncertainty (H1)– Demand Correlation (H2)– Flexible Competition (H3a)– Competition with moderate uncertainty (H3b).– Mean Demand (H4).– Demand differential (H5).

Regression

• Regress HPV against these six mismatches (OLS).

• Control Variables:– Flexibility– Utilization– Plant Capacity– Companies– Years– Number of Chassis Configurations.

N=375 Controls only All Variables

Intercept 7.995*** (0.359) 7.899*** (0.349)

Mismatch: Competition -0.115*** (0.031)

Mismatch: Competition and Moderate Uncertainty

0.057 (0.098)

Mismatch: Uncertainty 0.043** (0.026)

Mismatch: Correlation 0.015 (0.025)

Mismatch: Mean Demand 0.006 (0.026)

Mismatch: Demand Differential -0.062***(0.023)

GM -0.122*** (0.026) -0.154*** (0.027)

FORD -0.183*** (0.027) -0.210*** (0.028)

Assembly Line Flexibility 0.068** (0.036) 0.046(0.037)

Body Line Flexibility 0.126*** (0.042) 0.154*** (0.042)

Chassis 0.022*** (0.005) 0.014*** (0.005)

LOG (Capacity) -0.366*** (0.028) -0.371*** (0.029)

LOG (Utilization) -0.223*** (0.028) -0.234*** (0.029)

Year 1998 0.011 (0.061) 0.050 (0.060)

Year 1999 -0.055 (0.061) -0.025 (0.060)

Year 2000 -0.066 (0.061) -0.027 (0.060)

Year 2001 -0.119** (0.061) -0.089 (0.060)

Year 2002 -0.178*** (0.061) -0.152** (0.060)

Year 2003 -0.223*** (0.061) -0.186*** (0.060)

Adjusted R2 0.592 0.614

F-Statistic 42.829*** 32.316***

Results

• In the absence of the environmental variables, flexible plants have significantly higher HPV than inflexible plants.

• Adjusting for deviations from the benchmarks determined by the six environmental variables, flexibility is no longer significant.

• Flexibility by itself does not cause lower productivity

The six benchmarks

• Uncertainty: not matching flexibility deployment to environmental uncertainty decreases productivity.

• Competition: Responding to flexible competition with flexibility decreases productivity.

• Demand Differential: Contrary to theory.

The Control Variables

• Plants with higher capacity and utilization have higher productivity.

• Productivity has been increasing over the past years.

• GM and Ford have higher productivity than DCX.

Summary

• One of the first studies to formalize the deployment of manufacturing flexibility.– Demand uncertainty– Correlation.

• Though flexibility is used as a competitive weapon (flexible plants have higher flexible competition), evidence also suggests that this could be a cause of lower productivity.

• Flexible plants have lower utilization, a possible reason is the presence of volume flexibility in conjunction with product flexibility.

Chapter 4

Capacity Investment and the Interplay between Volume Flexibility and Product Flexibility.

Product Flexible Technologywith volume flexibility (VP)

K

K+εK-ε

Demand for product 1

Demand for product 2

product 1

product 2

Product Flexible (P) Technology

Demand for product 1

Demand for product 2

product 1

product 2

Total Capacity fixed

Capacity Allocated to Product 1

Capacity Allocated to Product 2

Product Flexibility

-ε

-ε

+ε

+ε

Demand for product 1

Demand for product 2

product 1

product 2

Volume Flexible Technology (V)

The Dedicated (D) Technology

Demand for product 1

Demand for product 2

product 1

product 2

Volume and Product Flexibility• Both types of flexibility help cope with demand

uncertainty.– Ample literature on capacity investment into product

flexibility.– Virtually non-existent literature on volume flexibility.– When would a firm prefer one flexibility to another?

• A plant may possess (to some extent) both flexibility types.– No analytical models combining two flexibility types.– When would a firm add one flexibility over another?

The Model

Choice of Technology

D,V,P or VP

i

2y

1y

Two markets

Uncertain Demand Curve

Uncertain Demand Curve

2111 qqAP

1222 qqAP

,

Decision timeline

time

Decide choice of technology, Dedicated (D), Product-Flexible (P), Volume-Flexible (V), Vol & Prod-Flexible (VP),

xc

Choose capacity

: Cost of capacity per unit

x={D,V,P,VP}

Adjust and/or allocate capacity

Choice of Technology

ProductionCapacity Investment

Demand Curves realized

The Problem Formulation

Firm i invests in V technology Firm i invests in VP technology

2

1

2~

2

1

~

)3(

~~

21,

)(

)(

max21

yyvyv

y

yvvyyvyv

vvvvAKK

v

KKc

KKKA

KKcEvv

vpvpvp

vpvp

yyvpvpyyvpyvp

vpvpvpAK

vp

Kqqts

KKc

qqqA

KcEvp

~

21

2~

2

1)3(

..

)(

)(

}{max

As c→,

Volume Flexible →Dedicated

As c→,

Vol-Product Flexible → Product Flexible

Frictional cost of capacity adjustment

Expected Profits

)1(4

)(

)1(4

)()(2

221

22

21

dd

d

cc

)1(8

2

)1(4

)(

)1(4

)()(12

22

21

2

221

22

21

pp

p

cc

))1((4

2)1)((

2)1(4

)(

)1(4

)()(22

1222

21

2

2

221

22

21

c

c

c

ccc vvvv

)21)(1(4

)(2)1)((

4)1(4

)(

)1(4

)()(12

22

21

2

2

221

22

21

c

cc

c

ccc vpvpvpvp

D

P

V

VP

DeterministicStochastic

Leverage

The cost thresholds

Variance

Cost

DedicatedVolume Flexible

Dedicated Vs Volume Flexible

Dedicated Vs Product Flexible

Dedicated

Product Flexible

How do these thresholds behave?

Comparing D,V and P - Large Correlation

xpv ccc

,,,,2* ccv ,,,2*pc

Dedicated

Volume

Flexibility

Volume

Flexibility

Variance

D>(V,P) V>D>P

V>P>D

Cost of flexibility

With large aggregate uncertainty in demand Volume Flexibility is more useful

,,,2*pc

DedicatedVolume Flexibility

Volume

Flexibility

Product Flexibility

Variance

xpv ccc

D>(V,P)

V>D>P

V>P>D

P>V>D

,,,,2* ccv

Comparing D,V and P - Medium Correlation

Cost of flexibility

,,,2*pc

Volume Flexibility

Dedicated Product

Flexibility Product

Flexibility

Variance

xpv ccc

D>(V,P)

P>D>V

V>P>D

P>V>D

,,,,2* ccv

Comparing D,V and P - Low Correlation

Cost of flexibility

With small aggregate uncertainty in demand Product Flexibility is more useful

Technology upgrade (addition)

Aggregate uncertainty is fixed

Profit VP - Profit P0,

Profit VP - Profit P0.

Incremental value of Volume Flexibility:

Aggregate uncertainty is fixed

Profit VP - Profit V0,

Profit VP - Profit V0.

Incremental value of Product Flexibility:

Additional volume flexibility helps when aggregate demand uncertainty is large but individual demand uncertainty does not matter.

Additional product flexibility helps when aggregate demand uncertainty is small and individual demand uncertainty is large.

Key Findings• Match flexibility to the environment (preference):

– Product flexibility - individual demand uncertainty.– Volume flexibility - aggregate demand uncertainty.

– Product flexibility - substitutable products (VCR and DVD Player)– Volume flexibility - complementary products (VCR and TV)

• Incremental Product flexibility may be harmful even if it is costless (V>VP).

• Linking: Quick Response (volume flexibility) and Variety Postponement (product flexibility).

• Empirical study on the adoption of flexibility in the automotive industry is in progress.

Appendix

Total Capacity not fixed

Capacity Allocated to Product 1

Capacity Allocated to Product 2

Vol-Product Flexibility

Flexibilities as “Building Blocks” and Technologies

Volume Flexible

Product Flexible

D

V

P

VP

X X

X

X

Dedicated

Volume Flexible

Product Flexible

Volume and Product Flexible

The Problem Formulation

2

1)3(

21,

)(

max21

yyddyydyd

ddddAKK

d

KKKA

KKcEdd

..

)(max

}{max

21

2

1)3(

, 21

ppp

yyppyypy

qqp

pppAK

p

Kqqts

qqqA

KcE

pp

p

Firm i invests in D technology Firm i invests in P technology

Model of Volume Flexibility

Adjusted Capacity,

Cost of flexibility

Kyv

2~

yvyv KKc

K+ε=K~

K

yvK~

K-ε=K~

Frictional Cost of Volume Flexibility

(Source: The Second Century, Holweg and Pil)Capacity

LevelChange over year

Change over week

50% 75.5 % 78.7%

80% 90.0% 92.2%

100% 100% 100%

110% 105.1% 106.1%

Financials

• The typical scale of operation is about 200-250,000 vehicles per year.

• In year 2002, the average incentive for the US automotive industry was $1873 per vehicle (The Second Century, Holweg and Pil).– The average incentives for the Big Three was $2300

per vehicle.

• The pretax profit per vehicle ranged from $226 (DCX) to $2069 (Nissan) per vehicle in 2002 (The Harbour Report, 2004).

Product Flexible Technologywith volume flexibility (VP)

K

K+εK-ε

Demand for product 1

Demand for product 2

product 1

product 2

Product Flexible (P) Technology

Demand for product 1

Demand for product 2

product 1

product 2

Total Capacity fixed

Capacity Allocated to Product 1

Capacity Allocated to Product 2

Product Flexibility

-ε

-ε

+ε

+ε

Demand for product 1

Demand for product 2

product 1

product 2

Volume Flexible Technology (V)

The Dedicated (D) Technology

Demand for product 1

Demand for product 2

product 1

product 2