BeFi Web Seminar for October 31, 2007
Identifying, Prioritizing, and Using
Objectives for Complex Decisions
© BeFi Forum 2007
by Ralph L. Keeney
Duke University, Research Professor
Consultant U.S. Marketing
and Decisions Group, Inc.
Identifying, Prioritizing, and
Using Objectives for Complex
Decisions
by
Ralph L. Keeney
Fuqua School of Business
Duke University
Durham, North Carolina, USA, 27708-0120
Keeney@duke.edu, KeeneyR@aol.com
1
Objectives of the Presentation
• Describe recent experiments to
generate objectives
• Present some techniques to help
create good sets of objectives
• Demonstrate the relevance of good
objectives to important decisions
• Communicate some useful information!
Have some fun!
2
Objectives Are Critical for Decisions
What Are Objectives? Objectives define
what you hope to achieve
You can state objectives using a verb and an
object
Examples:
• Minimize environmental damage
• Maximize profits
• Save money
• Make my spouse happy
3
Where Should Objectives Come From?
Objectives for a decision should come from
the decision makers
A Key Assumption in Decisions: Individuals
can articulate their objectives.
Can they?
Can they do this well?
4
Experiments to Identify Objectives
Have individuals identify their objectives
for important decisions that they face:
• Selecting an MBA program
• Choosing a dissertation topic
• Selecting an MBA internship
5
Overview of Experiments
Step 1: DMs generate as
many relevant objectives
as they can.
Objectives List
Step 2: DMs see the
master list and check all
relevant objectives for
their decision..
Master List
Objective
DM
DM
DM
DM
….
….
….
Objective A
Objective B
Objective C
Objective D
Objective
Step 3: DMs map
objectives from Step 1 to
the master list. Checked
items that map back are selfgenerated objectives; all
others are recognized.
Step 4: DMs rank or rate the
importance of all checked
objectives.
Master List
Master List
Objective
Objective
Objective
Objective
1
Objective
Objective
Objective B
Objective D
Objective B Objective D
Objective
Objective
Objective
Objective
Objective
5
Objective
2
Objective
Objective
Objective
Objective B
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective
Objective A
DM Objective C
Objective A
DM Objective C
Objective
Objective
_____________
Objective
_____________
Objective
4
3
Objective
Objective
_____________
Objective
_____________
Objective
Objective
_____________
Objective
_____________
6
Selecting an MBA Internship
participants:
33 full-time MBA students
28 completed all parts
conditions:
listed objectives over time with
reminders
consultation with others was suggested
listed objectives mapped to personal list
of recognized objectives
importance
all recognized objectives later rated from
1 (not important) to 9 (extremely important)
7
Master List of Summer Internship Objectives
I would like to choose an internship that...
Improves my attractiveness for full-time job
offers
Helps me make good networking contacts
Gives me pride from landing a prestigious
internship
Is at a company that sponsors work visas for
placement in US offices
Helps me develop my leadership skills
Provides information to help select a job
after graduation
Provides opportunities to interact with senior
managers
Provides a structured program for learning
and training
Uses skills I have learned in my first year of
B-school
Allows me to meet interesting people
Is with a company whose culture I identify
with
Is challenging
Helps me decide what courses and skills I
need to develop next year
Is enjoyable to do
Could lead to a full time offer from that firm
Helps me improve my communication skills
Is a job that I would like to do full-time after
graduation
Compensates me well
Gives me a substantial project of which I can
feel ownership
Allows me to experience a new geographical
area Helps me decide whether the internship
field is good for me long-term
Provides flexibility for personal interests
during the summer
Enhances my resume
Is with an organization that I am passionate
about
Lets me work with a diverse group of people
Enhances my knowledge in a particular
industry
Offers the chance to learn new skills
Is at a well recognized / respected company
Is in a specific location (e.g. near family or
friends)
8
MBA Internship Results
self-generated objectives:
6.8 average
relevant objectives:
14.8 average
recognized objectives:
7.9 average
average importance
rating: from 1 (not
important) to 9
(extremely important)
6.8 for recognized objectives
7.4 for self-generated objectives
2.9 for bogus objective
4.7 for baseline objective
Bogus objective - allows me to mentor other employees
Baseline objective - helps me improve my communication
skills
9
Summary of Experimental Results
For realistic decisions problems:
• Individuals can identify only about half of
their objectives
• They miss objectives that are roughly as
important as those identified
• Techniques can help individuals identify
more of their objectives
These results are consistent with all of my
consulting experience
10
A Procedure to Help Identify
Objectives
• Write down your values (i.e. what you
care about) that may be influenced by
your decision
• Expand list of values with ideas from
others (stakeholders, friends,
knowledgeable individuals)
• State values as objectives
• Organize objectives
11
Techniques to Identify Objectives
•
•
•
•
•
•
•
•
•
Use a wish list
Think about alternatives
Imagine possible consequences
Use different perspectives
Think about strategic objectives
Ask ‘why’ for each objective
List objectives over time (Franklin’s idea)
Do individual thinking first
Ask for ideas of others
12
Organize Objectives
Separate ends from means to establish the
fundamental objectives
Means Objective: an objective whose
importance stems from its contributions to
achieving another objective.
Fundamental Objective: objective that defines
a basic reason for caring about a decision.
Example:
• Means Objective – maximize sales
• Fundamental Objective – maximize profits
13
Means-Ends Objectives Network
CO Air Quality Standards
CO
concentrations
health
impacts
CO
doses
breathing
rate
CO
emissions
CO
dispersion
body activity
construction
schedule
costs
maintenance
requirements
fines for
violators
Adapted from Keeney, 1992
14
Fundamental Objectives Hierarchy
CO Air Quality Standards
fatal
heart
attacks
nonfatal
health impacts
angina
attacks
peripheral
vascular
attacks
regulation
cost
capital
equipment
operations
costs
enforcement
cost
health
cost
Adapted from Keeney, 1992
direct
(e.g., treatment)
indirect
(e.g., lost
opportunity)
15
Summary
• Identifying objectives for a decision is
the foundation for thinking about or
analyzing that decision
• Identifying objectives for decisions is
not easy
• Therefore, the activity of identifying
objectives for important decisions is
worthy of thought, time, and effort
• Techniques to facilitate this process
are useful
16
Evaluating Customer
Acquisitions at American
Express Using Multiple
Objectives
by
Ralph L. Keeney
Duke University
and
Qing Lin
American Express
17
Outline of Presentation
• Overview of Cardmember Acquisition
Decisions
• Our Project: Add Market Share to Profit in
Customer Acquisition Models
• Our Approach: Assess an Objective
Function to Quantify American Express
Values
• Implementation of the New Model
• Business Results
18
Cardmember Acquisition Decisions at
American Express
Front
Front End
End Decision
Decision
for
Active
Channels
for Active Channels
Consumer
Response
For Each Responder / Applicant
Applicant
Information
Approve
and assign
credit limits
For Each Prospect
Response
Prospect
Information
Cardmember
Behaviors
•
•
•
•
•
•
Credit Risk
Attrition
Revol. Bal.
Spending
AR days
# of Sub
Decline
Offer Product 1
Product 2
Do Not Solicit
No Response
19
Decision Infrastructure for Acquisition
Since late 1980s, American Express has been leveraging decision
science in its cardmember acquisition process to maximize long-term
profitability
Max NPV
Optimal
Decision
Rules
Business
Requirement
(i.e. budget
constraint)
Implemented in
Direct
Marketing and
Application
Processes
Optimization
Customer
Information
Evaluation Models
• Predicting
Customer Behavior
• Assessing Longterm Profitability
s.t.
20
Our Project: Including Contribution to
Market Share in Decision Models
In 1995, the company started to emphasize on strategic
significance of market share growth. The cardmember acquisition
business needed to expand its decision infrastructure to include
market share growth.
Expand Acquisition
Models to Include
Market Share Growth
Evaluation Models
•
Customer
Information
•
Predicting Customer
Behav ior
Assessing Long-term
Profitability
• Contributions to
Market Share Growth
Optimization
Max Multiple
Objective
Utility Function
s.t. Business Requirement
(i.e. budget constraint)
Optimal Decision
Rules
Implemented in Direct
Marketing and
Application Processes
21
Create and Assess a Multiple Objective Utility
Function for American Express’ Card
Business
• Identify and select objectives for assessment
• Select functional form for the utility function
• Assess the required information
– Much interaction with senior managers throughout
the process
– Cycled through three times
22
Relationships Among Key Objectives
Relevant to Acquisitions
Short Term
Cash Flow
Acquisition
Expenses
Shareholder
Reaction
Investment
Opportunities
New M embers
for
Charge Cards
Total Spending
Using Charge
Cards
M arket
Share of
Spending
New M embers
for
Credit Cards
Total Spending
Using Credit
Cards
M arket Share
of Total
Outstanding
Directly Influenced
by Acquisition
Decisions
Network
Effects
Profitability
Over Time for
Current Business
M arket
Effects
Corporate
Image
(“
” means “influences”)
23
Additive Multiattribute Utility Function
4
U(X1, X2, X3, X4) = Ki Ui(Xi)
i=1
X1 = short term cash flow
X2 = profitability over time
X3 = market share of charge volume
X4 = market share of lending
24
The Measures for the Utility Function
Measure
X1 = Cash flow NPV for next 3 years
($ 1997 billions)
X2 = Long-term profit NPV
($ 1997 billions)
X3 = Charge volume market share
(percent in 1999)
X4 = Lending outstanding market
share (percent in 1/1/2000)
Range
Bad
Good
$1.2
$3
$6
$12
12%
20%
3%
6%
25
Prioritizing Objectives – Ranking Ranges of
Consequences
Measure
X1 = Cash flow NPV for next 3 years
($ 1997 billions)
X2 = Long-term profit NPV
($ 1997 billions)
X3 = Charge volume market share
(percent in 1999)
X4 = Lending outstanding market
share (percent in 1/1/2000)
Range
Bad
Good
$1.2
$3
Rank
4
$6
$12
1
12%
20%
2
3%
6%
3
26
Assessing Value Tradeoffs
20
Charge
volume
market
share
(percent)
A
Assessed as
Indifferent
18
16
14
12
6
D
F
H G
E C
7
8
9
11
10
12
Long-term Profit NVP (billions $)
27
Summary of Assessed Value
Tradeoffs
(X1 = $3 billion, X2 = $6 billion) ~
(X1 = $1.2 billion, X2 = $7 billion)
(X2 = $6 billion, X3 = 20%) ~
(X2 = $9 billion, X3 = 12%)
(X2 = $6 billion, X4 = 6%) ~
(X2 = $8 billion, X4 = 3%)
( ~ means is indifferent to)
28
The Utility Function
U(X1,X2,X3,X4) = K1U1 (X1) + K2U2 (X2) + K3U3 (X3) + K4U4 (X4),
where
K1 = 0.052, K2 = 0.625, K3 = 0.204, K4 = 0.119
and the Ui are assessed next.
29
Summary of Assessed Value
Judgments for Single Utility Functions
1. (X1 = $2.4B) ~ (X1 = $3B, 0.5; X1 = $1.2B, 0.5)
2. (X2 = $10B) ~ (X2 = $12B, 0.5; X2 = $6B, 0.5)
3. (X3 = 17%) ~ (X3 = 20%, 0.5; X3 = 12%, 0.5)
3. (X3 = 16%) ~ (X3 = 17%, 0.5; X3 = 12%, 0.5)
4. (X4 = 4.75%) ~ (X4 = 6%, 0.5; X4 = 3%, 0.5)
( ~ means is indifferent to)
30
Individual Utility Functions
100
100
utility
utility
0
$1.2
0
$2.1
$3.0
$6
$12
Profitability over time (NPV)
Short term cash flow (NPV)
utility
$9
100
100
50
utility 50
0
0
12
16
20
Market share-charge volume (%)
3
4.5
6
Lending outstanding market share (%)
31
The Single Utility Function
U(X1,X2,X3,X4) = K1U1 (X1) + K2U2 (X2) + K3U3 (X3) + K4U4 (X4 )
U1(X1) = -0.3051(1-e0.8074(X1-1.2)), 1.2 X1 3.0,
U2(X2) = -0.3051(1-e0.2422(X2-6)), 6 X2 12,
-0.01955(1-e0.656(X3-12)), 12 X3 17,
U3(X3) = +0.5 + 0.6526(1--0.4844(X3-17)), 17 X 20,
3
U4(X4) = -1.0268(1-e0.2267(X4-3)), 3 X4 6
32
The Revised New Business Acquisition Model
Customer Behavior Models
Economic Models
Response
Acquisition
Information
• Credit Data
• Demographic Data
• …...
Probability of Default
Probability of Voluntary Attrition
Default Balance
Cardmember
Profitability
Expected
Profit
Spending/Revolving Balance
Decision Variables
Product & Channel
•Charge Card vs. Optima
•Direct M ail vs Telemarketing
Market Share
33
Implementation
• Implemented in 1995
• Used for over 50 million prospective
customers
• Facilitated group decisions among
Marketing, Risk Management and Finance
• Significantly altered prospect targeting
strategies, focusing more on higher usage
and higher risk group
• Significantly increased customer acquisition
volume, exceeding business goals of
acquisition for 1996 by a large number
34
What Occurred
• In first six-months of 1997, AX market share
of domestic credit card market increased
from 18.31% to 18.90%, whereas it had lost
market share over the previous ten-year
period.
• Translating the value of the increase of
18.31% to 18.90% into NPV of the current
business using the business utility function
yields $447 million.
35
Impact of This Work
• Cause and effect is hard to discern.
Other factors such as the economy and
actions of competitors affect market
share.
• An explicit focus on market share may
have significantly helped.
36
Two Quotes
“You’ve got to give credit to American Express
for overcoming huge obstacles to growth
through creative marketing…”
H. Spencer Nilson, publisher of Nilson Report
in WSJ (Sept. 22, 1997)
“Market share has been a very key focus for
our card business over the last three years
and I think what you’re seeing is that we
have had a very strong effect in increasing
spending on our card products and our cards
in force.”
Kenneth I. Chenault, AX President
in WSJ (Sept. 22, 1997)
37
Some References
Bond, S D., Carlson, K.A., and R.L. Keeney, Generating Objectives: Can
Decision Makers Articulate What They Want?, Management Science,
to appear.
Keeney, R.L. Value-Focused Thinking: A Path to Creative
Decisionmaking, Harvard University Press, Cambridge, MA 1992.
Keeney, R.L. and Q. Lin, Evaluating Customer Acquisition at American
Express Using Multiple Objectives, Interfaces, 30, Vol. 5, 31-33, 2001.
Keeney, R.L. and R. M. Oliver, Designing Win-Win Financial Loan
Products for Consumers and Businesses, Operational Research, 56,
1030-1040, 2005.
38
PRESENTED BY
Shlomo Benartzi
Co-Founder, BeFi
Associate Professor Co-chair of the
Behavioral Decision Making Group
The Anderson School at UCLA
Warren Cormier
© BeFi Forum 2007
Co-Founder, BeFi
President, Boston Research Group