 A Decision Analysis is an
analysis that uses expertise or knowledge about actions and outcomes in the form of a
decision tree to determine the "best" strategy for managing risks in some
category.1
Examples of expertise or knowledge include:
- What are the sequences of possible decisions
and outcomes in a promotional campaign for a new product.2
- What are the options and the possible damage
awards, in pre-trial negotiations -- from the point of view of the plaintiff, and from the
point of view of the defendant.3
- What design choices and tests are needed in
the building of a waste treatment system.4
- What locations and risks should be
considered in a feasibility study for locating a manufacturing facility.
- What are the sequences of possible
decisions, including the decision to perform coronary artery bypass graft surgery, and
possible outcomes for a patient with documented coronary artery disease?5
The source of the expertise or knowledge is typically one
or more persons and the field's literature. The knowledge is organized in the form of a decision
tree with nodes connecting decisions and outcomes, or actions with results.
Sophisticated Decision Analyses can incorporate uncertainty, allowing for subjective
or intuitive decision making.
Here's an illustrative
example for a Promotional Campaign (from S.M. Lee et al., 1985, with
modifications).1
Nodes N1, N4, N5 are decision nodes.
Nodes N2, N3, and N6 through N11 are outcome nodes.
Nodes N12 through N24 are final outcome nodes. Only node N2
involves a cost. Briefly, a node is a decision node if its branches are decisions; and it
is an outcome node if its branches are outcomes or chance events. All final nodes must be
outcome nodes.
Here are some sample assumptions for the Promotional
Campaign.
|
NODE |
DESCRIPTION |
PROB. |
COST, BENEFIT, OR PAYOFF |
|
From |
To |
|
N1 |
Start
|
-- |
-- |
-- |
|
N2 |
We
Introduce the Product |
|
R&D Cost of $80K |
|
|
N3 |
We
Do Not Introduce the Product |
|
|
|
|
N4 |
Our
Competitor Introduces a Competing Product |
70% |
|
|
|
N5 |
Our Competitor
Does Not Introduce a Competing Product |
30% |
|
|
|
N6 |
We
Launch a Major Promotional Campaign, etc. |
|
|
|
|
N12 |
We
Launch a Big Promotional Campaign |
60% |
$50K |
$80K |
|
N13 |
We
Launch a Medium Promotional Campaign |
30% |
$80K |
$100 |
|
N14 |
We
Launch a Small Promotional Campaign |
10% |
$140 |
$180 |
Note that the probabilities of N4 and N5 add
to 100%, and those of N12, N13, and N14 also add to 100%. The product R&D cost
associated with N2 is assumed to be $80K. N12, N13, and N14 are three of 13 possible final
outcomes. The payoffs (profits) of the promotional campaign are assumed to range between
$50K and $80K for N12, between $80K and $100K for N13, and between $140K and $180K for
N14. The probabilities and payoffs for nodes N15 through N24 are not shown.
Briefly, the logic for the Promotional Campaign is
as follows:
- We have two decisions: Either we introduce
our product (N2), or we don't (N3). If we introduce our product, we incur $80K in R&D
or consulting costs. N2 and N3 are decision nodes.
- Our action to introduce the product can
induce our competitor to introduce a competing product. Thus N2 can have two outcomes: Our
competitor introduces a competing product (N4), or does not (N5). Based on our knowledge
of the marketplace, our competitor, and some marketing intelligence, we assess the
probability of N4 to be 70%, and that of N5 to be 30%. N4 and N5 are outcome nodes
or chance nodes.
- The final outcomes of the Promotional
Campaign can depend, among other things, on our actions, our competitor's actions, the
size of our campaign, and the size of our competitor's campaign. Thus, we can analyze the
final outcomes in terms of three possible promotional campaigns: We launch a big campaign
(N12), we launch a medium campaign (N13), or we launch a small campaign (N14). If we
launch a big campaign, we estimate that our profit will be in the range from $50K to $80K,
with a probability of 60%, etc.
The "best" strategy depends on the criterion
used. In a marketing analysis, the criterion is typically the Expected Monetary Value
(EMV). The "best" strategy is the one for which the EMV is maximum.
Clearly, not all decisions are based on money. Actually,
the most important decisions in life often have nothing to do with money. Consider, for
example, the decision of whether to perform Coronary Artery Bypass Graft Surgery
or to manage the disease medically (substantially simplified, for illustrative purposes,
from M.C. Weinstein et al., 1980).7 In this case, the decision tree can include
the following decisions and outcomes (only a few nodes are listed)
|
NODE |
DESCRIPTION |
PROB. |
UTILITY
(QALY) |
|
From |
To |
|
N1 |
Start |
-- |
-- |
-- |
|
N2 |
Perform Coronary Artery Bypass Graft Surgery |
|
|
|
|
N3 |
Manage the Disease Medically |
|
|
|
|
N4 |
Patient Dies |
10% |
0 |
0 |
|
N5 |
Patient Survives Eight To Ten Years, Four of Which Involve Severe
Angina, etc. |
90% |
7.2 |
10.2 |
The outcomes can involve complications
(e.g., stroke, infection, etc.) and chronic limitations. Some complications can
be fatal, others non-fatal. The utility of an outcome depends on the condition and
preferences of the patient (e.g., years of survival, and degree of pain relief). The
probabilities of the outcomes can depend on the patient's characteristics and those of the
treating institution. The final outcomes can be assessed in terms:
Death
Years Of Life With Pain
Years Of Life Without Pain, etc.
The outcomes can change with time. The utility of the
outcomes can be expressed in terms of quality-adjusted life years (QALY). |
The knowledge for the decision problem can
be generally available, privileged, or a proprietary trade secret. The knowledge can be
objective or subjective, certain or uncertain.
The output of the Decision Analysis is the EMV of each
node. The path with the highest EMVs determines the "best" strategy.
| Macroknow Decision Analysis
Services translate your objective or subjective knowledge of the decision problem at hand
into Expected Monetary or Utility Values.
What do you
get from Macroknow? You get access to a private website
featuring:
- The Decision Tree and the EMVs (or Expected Utilities) of
the nodes, in both graphical and tabular forms. The "best" strategy is
identified.
- If requested, colorful sensitivity
graphs depicting how the EMVs (or Expected Utilities)
vary with changes in costs, benefits, and payoffs (or utilities).
Only you or those you authorize can access the private
website on the Internet.
You can use the website to communicate the
results of the calculations to your board of directors, your team, your customers, your
public relations department, your legal department, etc.
The power of technology must be
tempered with prudence. Prudence requires that you note the
following:
- You must always exercise care in the
choice of assumptions and in the interpretation and use of calculated results.
- Calculations can only be one
component, out of many, in the decision making process.
- All calculations and methods have
limited domains of adequacy.
- There is no known method to certify
the absolute correctness of any complex computer program.
- Most important, there is no
substitute for a circumspect judgement.8
|
Sources:
1 For a thorough treatment, see Howard Raiffa, Decision Analysis. New
York, NY: Random House, 1968.
2 See Sang M. Lee, Laurence J. Moore, and Bernard W. Taylor III, Management
Science, 2nd ed. Dubuque, IA: Wm. C. Brown Publishers, 1981, 1985, at 431-434
[Decision Tree for Promotion of New Product].
3 See Howard Raiffa, The Art and Science of Negotiation. The President
and Fellows of Harvard College, 1982. Cambridge, MA: Harvard University Press, at 66-77
{Settling Out of Court].
4 See Alfredo H.S. Ang and Wilson H. Tang, Probability Concepts in
Engineering Planning and Design. Vol. II-Decision, Risk, and Reliability. New York,
NY: John Wiley & Sons, Inc., 1984, at 17-32 [Waste Treatment System].
5 Milton C. Weinstein, Harvey V. Fineberg, et al. Clinical Decision
Analysis. Philadelphia, PA: W.B. Saunders Company, 1980.
6 See 2 above.
7 See 5 above.
8 See Howard Raiffa, Decision Analysis. New York, NY: Random House,
1968, at 268-272 [Pros and Cons of Decision Analysis].
|
