Mark Cronshaw PhD MBA
Quantitative methods for decision making have great potential for improving
business success. They are not well known today. They may soon become so
commonplace that people will wonder why business decisions were ever made
in any other way.
Quantitative and analytical tools are available to improve business
These tools increased the value of Smith Kline Beecham’s drug
portfolio by $2.6 billion1,2
They deal with uncertainty and limited information
They can incorporate different opinions
Use of these quantitative tools
Increases the chance of a good outcome
Often unearths valuable new alternatives
1. “Smart Organizations Perform Better” D. & J. Matheson, Research Technology Management, Jul-Aug 2001 p. 53
2. “How Smith Kline Beecham Makes Better Resource-Allocation Decisions”, P. Sharpe & T. Keelin, Harvard Business
Review, Mar-Apr 1998, v76, n2, pp. 45-53
Jan. 27, 2003 303 618 9279 2
Should we introduce a new product?
Should we deploy the new technology recommended by a vendor?
How much should we invest …
To reduce our costs?
To enhance our call centers?
How should we raise new capital?
Should we outsource?
Should we add new marketing channels, like e-commerce?
Should we acquire, merge or divest?
Jan. 27, 2003 303 618 9279 3
What Makes Decisions Hard?
How do you make decisions?
Jan. 27, 2003 303 618 9279 4
Typical method Analytical method
Identify alternatives Inspire creativity
May only consider one alternative
Consider many alternatives
Verbal pros & cons Sensitivity analysis
Deterministic business case Explicit consideration of risk
Value of information
May be conflictual
Choose alternative Clear, easy, fast
Uncertainty and doubt persist
Analytical method spends more time identifying and evaluating alternatives
Often unearths attractive, previously over-looked choices
The final choice is easy after careful evaluation of alternatives
Jan. 27, 2003 303 618 9279 5
No consensus Charismatic leader
Opposing ideas cancel out One strong opinion Better business
Stressful Creates movement decisions
Go nowhere May not be best direction
Partial agreement Alignment
General alignment Coherent action
Slow progress Rapid progress
Some opposing ideas Focused effort
Jan. 27, 2003 303 618 9279 6
How to Create Alignment: a Mini-Case
XYZ Corp. is deciding whether to introduce a “basic” or “advanced”
type of new product
Some believe that “advanced” is better because
It demonstrates XYZ’s technological leadership
It is more robust to competitive challenges
It can do more for purchasers
Others believe that “basic” is better because
It is not clear how much buyers will pay for the advanced features
The advanced product will cost more to produce
Technical difficulties could delay introduction of the advanced product
Debates on the pros and cons of each are time consuming
Jan. 27, 2003 303 618 9279 7
for basic Power
type Profit Total
Diagram clearly illustrates factors relevant to the decision. Builds consensus.
All parties can see what factors have been included, and can add others if desired
Rectangle is a decision node, where XYZ chooses
Ellipse is a chance node, representing uncertainty beyond XYZ’s control
Different data can be provided by different subject matter experts within XYZ
Jan. 27, 2003 303 618 9279 8
Based on assumptions and models described in the appendix
Best apparent decision: Introduce the basic product. The outcome is
uncertain, but has an expected value of $76M.
Risk: If the market is sophisticated, then the advanced product would yield
a much better payoff of $170M. But … the advanced product only yields
$10M if the market is naïve.
Value of perfect information: Could do market research before introducing
the product. If the market research is perfectly accurate, then the expected
value rises by $20M.
Value of imperfect information: No research is perfect! If the research is
accurate 75% of the time, then it is worth only $1.4M. If it is accurate with a
probability of 73% or less, then it has no value.
Use judgment and past experience to assess the accuracy of market research
Do research if value exceeds cost, otherwise introduce basic product without
Jan. 27, 2003 303 618 9279 9
Benefits of Quantitative Methods
Create consensus about the right action in a short amount of time
(sometimes just a few days)
Incorporate many diverging opinions in the decision process
Use the best resources in your company for gathering information
Bayesian methods make the most of limited information
Avoid costs and delays from pursuit of information with limited value
Develop a clear vision
Use this clarity of vision to
Increase the chance of a good outcome
Reduce the chance of costly mistakes
Identify previously unconsidered alternatives
Even a small improvement in a high value project is worth $$
Jan. 27, 2003 303 618 9279 10
How to Make Better Decisions
When the stakes are high:
Use analytical and quantitative methods
Clearly identify current and future decisions
Define your objective(s)
Include uncertainty due to external factors
Cronshaw Consulting can
Partner with you for major decisions
Jan. 27, 2003 303 618 9279 11
Typical schedule for a major decision
Risk and risk attitudes
Alternative influence diagram
Jan. 27, 2003 303 618 9279 12
Typical Schedule for a Major Decision
1 Establish decision context
1 Identify core decision team and objectives
1 Team facilitation
2-6 1 - on -1 follow up interviews
5-15 Create preliminary model and presentation
1 Present preliminary results to team
5-10 Modify model and presentation
2 Presentations to team and leadership
? Socialize results and begin implementation
Small decisions can be completed in less time
Jan. 27, 2003 303 618 9279 13
Suppose that decision makers agree on the following. (This is a
simplified version of the variety of opinions that might be considered.)
There are two types of customer – power users and average users.
The exact mix is unknown.
XYZ thinks that there are two possible market scenarios
Sophisticated (20% probability): 10 M power users, 4 M average users
Naïve (80% probability): 2 M power users, 8 M average users
40% of power users will buy basic product
No average users will buy advanced product
Profit margin per unit: basic $10, advanced $20
Fixed cost of product introduction: basic $10 M, advanced $30 M
Jan. 27, 2003 303 618 9279 14
Basic [76.4] 20%
Product_type Naive 78
Advanced  20%
It is best to introduce the basic product.
There will be only slight uncertainty about the outcome: $70 or $78 M.
The statistical expected value is $76.4 M
Note the risk associated with the advanced product: upside and downside.
Is there some way to capture the upside?
Jan. 27, 2003 303 618 9279 15
Product_type Profit $M
Market_report 20% Advanced
Do market research to identify market type before making product decision
Market research raises the expected value from $76.4 to $96.4 M
Avoids introducing the basic product if the market is sophisticated
Perfect market research is worth $20 M
But no market research is perfect!
Jan. 27, 2003 303 618 9279 16
Imperfect Market Research Profit $M
Basic [74.5714] .429
Sophisticated [78.5714] .571
Advanced [78.5714] .429
Basic [77.3846] .077 70
Naive [77.3846] .923
Advanced [22.3077] .077
Suppose that the market research is correct with probability 75%
Do research and then decide on which product
The expected value drops to $77.8 M, so this imperfect information is only
There is a 57% chance of having chosen the wrong product when the
research reported a sophisticated market!
Jan. 27, 2003 303 618 9279 17
Imperfect Market Research – Tree Flipping
Market_type .200 Naive
Research_report .350 Naive .20
Jan. 27, 2003 303 618 9279 18
Even minor inaccuracy reduces the value of information
Expected value of information
60% 65% 70% 75% 80% 85% 90% 95% 100%
Probability of accurate report
Jan. 27, 2003 303 618 9279 19
Every Action is Intrinsically Risky
Perfect info Imperfect info
Quantitative methods are ideal for making decisions in the face of uncertainty
Can easily incorporate uncertainty about many factors
Can consider other risk attitudes instead of the risk neutrality implied by expected value
Jan. 27, 2003 303 618 9279 20
Choice A Choice B
Suppose that you were picking between choices A & B above. Which would you pick?
Most people would choose B.
The expected value of choice A is $12 M, which is more than the expected value of choice B.
Using expected value to make selections is not consistent with most people’s choices!
Different people have different risk attitudes; often risk aversion, e.g., choose B over A
There is no absolute right choice, cf., vanilla or chocolate.
There are models of risk attitudes that represent a decision-maker’s true preferences.
Such models further improve decision making.
It is not possible to make good business decisions without a consideration of risk attitudes.
Jan. 27, 2003 303 618 9279 21
Decision with more uncertainty
P er buyer
users P rofit
type Profit Total
Jan. 27, 2003 303 618 9279 22