A user interface for a software package that, like a
dashboard on an automobile, organizes and presents information in a way
that is intended to be easy to read and absorb. Most modern software tools
employ user interfaces that resemble dashboards, but vendors of project portfolio management (PPM) tools
often use the term to ensure that potential customers recognize the
similarity. Typically, and unlike most automobile dashboards, a PPM
dashboard is interactive — if the user clicks on an item, more
detailed information is provided. However, whether or not the information
displayed by a PPM dashboard is "real time," as it is with an automobile
dashboard, depends on the tool and the type of information presented.
Illustrative PPM dashboard
The process of extracting from a (usually large) database
information useful for decision making. Typically, data mining utilizes
software to identify statistical patterns or relationships in
data that may be commercially useful. Information obtained in this way is often used to help
organizations obtain improved understanding of customers and to aid decisions involving product design, marketing, and customer support.
de Bono Six Hats
A thinking tool for group and individual decision making.
The method is designed to help people make better decisions by looking at
the choice from different perspectives, thereby producing a more
comprehensive understanding of considerations and issues. The method is described in the book
Six Thinking Hats by Edward de Bono.
decision analysis (DA)
A body of knowledge and related analytic techniques for
helping decision makers choose among alternatives, taking into account the consequences of choosing each alternative
and the decision maker's preferences. Decision analysis (DA) is concerned with finding effective and practical ways to implement
DA provides methods and aids for addressing
essentially all the steps involved in formal decision making, including
problem definition, information collection, risk assessment, the
identification and screening of alternatives, the evaluation and selection
of alternatives, and the communication and implementation of decisions. I like
to describe DA as a "megatool" for decision
making—unlike most other decision tools, DA is more like a tool box
than a tool. DA is relevant to project
portfolio management because it provides a framework for analyzing
project selection decisions as well as specific methods for quantifying
project value and addressing project risk.
DA was initially developed in the 1960s and 1970s at
Harvard, Stanford, MIT, Michigan, and other major universities. The term
"decision analysis" was coined in 1964 by Ron Howard, a professor at
Stanford University. DA is generally considered a branch of the field of
operations research, but also has links to management science, economics,
systems analysis, psychology and statistical decision theory.
Although DA has continued to advance over the past decades, the basic principles
of the field as first articulated in the 1970's remain relevant. For example, when DA was being practiced mainly at the
Stanford Research Institute (SRI), it was described as being conducted according to a Decision Analysis Cycle, depicted below.
The Decision Analysis Cycle
In the first (deterministic) phase, it was advised that analysis begin by identifying the variables
affecting decision outcomes judged to be important by the decision maker.
Values are assigned, and the sensitivity of decision outcomes to the variables is measured via
sensitivity analysis without
consideration of uncertainty.
The second (probabilistic) phase) was described as starting with the encoding of
probabilities from knowledgeable experts.
This phase also includes the assessment of risk preference, which defines the best course of action in the face of uncertainty.
In the third (informational) phase), the results of the first two phases are synthesized through a calculation of the
value of eliminating uncertainty over each of the important, uncertain variables. The results of the informational phase
show what it is worth in dollar terms to have perfect information. Comparing the value of information with its cost
determines whether additional information should be collected prior to committing to a course of action.
In more recent years the Strategic Decision Group (SDG) and other management consulting firms have recommended
that DA be conducted as a collaborative process called the
Process, illustrated below. DA has continued to be an
area of consulting specialty, and there are journals and professional
societies devoted to the field.
The Dialogue Decision Process
According to DA, a good decision is one that (1)
considers the full range of alternatives that are available to the decision
maker, (2) accounts for what the decision maker believes will be the
potential consequences of choosing each alternative, and (3) is consistent
with the decision-makers preferences for those possible decision
consequences. In other words, making good decisions requires knowing what
you can do, what you believe, and what you want.
DA employs various procedures and tools for understanding
how the actions taken in a decision determine the consequences that may
result, as well as the significance of those consequences relative to the
decision-maker's objectives. To provide this understanding, an analytic
model of the decision is constructed. This decision model is typically composed of
two parts, a consequence model that simulates decision outcomes and a value
model for measuring the decision-maker's preferences for those outcomes.
Probabilistic reasoning is used to
quantify risk and determine whether additional information should be
collected before committing to a course of action. The models allow
sensitivity analysis, a
process that identifies the issues that make the most difference and helps
decision makers avoid "paralysis by analysis."
DA is generally focused on three different (but
overlapping) types of decisions: (1) one-time decisions where alternatives
must achieve multiple, competing objectives, (2) sequential decisions where
uncertainties and learning play an important role, and (3) portfolio
decisions where the goal is to provide a model and process for selecting a
portfolio of related alternatives.
The latter application is sometimes termed portfolio decision analysis (PDA), and is the
approach to project portfolio management that is described and recommended
in this website.
Regardless of the type of decision being analyzed, a
major task for the decision analyst is constructing a value model that
allows the overall desirability of alternatives to be computed based on how
they perform relative to a set of evaluation measures, or attributes. Multi-attribute utility analysis (MUA) is often
used to construct the value model. To represent decision timing and
uncertainty, sequential decisions are analyzed using decision trees and influence diagrams. Decision analyses
of project decisions often include
calculations of the expected net present
value (ENPV) of candidate projects and may include valuations based on real options analysis.
decision expert (DEX)
A multi-criteria analysis
(MCA) decision-making method implemented as an interactive computer
program that runs under the MS-DOS operating system on IBM-compatible
computers. The underlying methods were developed in the early 1980's
by a research team led by M. Bohanec, I. Bratko, and V. Rajkovic.
DEX works with
utility functions defined by users of the software.
The approach involves decomposing a decision problem into a
hierarchy of smaller sub-problems represented in a tree-like
Illustrative attribute structure defined in DEX
DEX differs from other MCA decision support systems in that
it uses qualitative, symbolic attributes instead of quantitative ones.
Aggregation equations in
DEX (e.g., value functions) are
defined by "if-then" decision rules rather than numerically (e.g.,as a
weighted scoring model).
Fuzzy logic and probability distributions can
be defined to accommodate uncertainty and incomplete information.
A person who decides things, especially at a high level in an organization.
In the context of
multi-criteria analysis, the decision maker is one of the two
somewhat idealized actors, the other being the analyst, who are assigned specific responsibilities for the application of a
formal approach for solving decision problems. The decision maker is the individual or group of individuals
assumed to have responsibility for identifying the need to make a decision and to have authority for
making choices. The decision maker
is also assumed to have responsibility for identifying decision objectives and
for directly or indirectly providing the preference structure needed by the analyst to evaluate and rank
decision alternatives. The analyst, on the other hand, is responsible for conducting a wide range of activities directed
toward problem formulation and the quantitative and qualitative analysis of the problem The analyst
constructs a mathematical model of the decision problem, and, depending on the solution approach used by the analyst,
may encode the
decision maker's preference structure as a multi-attribute utility function.
is responsible for communicating the results of the analysis to the decision maker and the decision maker must then
decide the acceptability of that solution.
A prioritization matrix that is
used for the purpose of aiding the selection of an alternative rather than
for the purpose of prioritizing alternatives (i.e., the alternative in the
matrix with the highest weighted-total score would be recommended for
A sample decision matrix
An analytic representation (model) of a choice among
possible actions or alternatives. A decision model represents the factors
relevant to decision making and their relationships.
Decision models fall into two categories:
descriptive and prescriptive (a prescriptive decision model
is also called a "normative" model). A descriptive decision model
describes how people typically make decisions and seeks to explain how
various factors influence decision-making behavior, including why people
often make sub-optimal decisions.
In the context of project
portfolio management, the term decision model usually refers to a
prescriptive decision model designed to indicate how people or
organizations should choose based on principles of logic and rationality.
For example, a decision model based on decision analysis seeks to identify
choices that are logically consistent with the available alternatives,
preferences, and beliefs of the decision maker. Such a model would include
performance measures that
indicate the degree to which the possible outcomes of choosing each
alternative would achieve each decision objective, and a value model indicating the relative
preference the decision maker assigns to such performance. In addition to
helping decision makers make good choices, decision models can often be
"mined" to provide additional information and insights, including
sensitivities and value of information
A broad spectrum of related concepts and techniques used
for the systematic study of the process of decision making, including the
study of how decision should rationally be made. Decision science draws
from numerous fields of study, including economics, decision theory, decision analysis, negotiation
analysis, management science, and operations research.
decision support system
A computer software program and associated database
intended to aid managers in making decisions. A DSS may include a decision model, simulation programs, and algorithms. Alternatively, a DSS may merely
provide information useful for supporting decisions. The term is an old one
that is not used much now, as its definition is so broad as to not be
particularly useful. A project portfolio
management tool is an example of a DSS.
Typically means utility theory. Sometimes, however, the
term refers more broadly to theory relevant both to how decision ought to
be made (normative decision theory )
and how they actually are made (descriptive decision theory.
A graphic representation of a decision problem wherein
the alternatives to decisions and possible outcomes to uncertainties are
represented sequentially by nodes and branches in a tree-like diagram. Many
software tools are available for constructing and analyzing decision
The sample decision tree below is a popular text book
example for a problem that might be faced by an oil wildcatter (a person
who drills exploration oil wells in areas not known to be oil fields).
An oil wildcatter's decision tree
According to the software tool used to construct this
tree decision nodes are shown as yellow squares, uncertainty or "chance"
nodes are shown as green circles, and terminal nodes are represented by
blue triangles. The initial decision, shown at the left of the diagram, is
whether or not to conduct a seismic test on the land prior to deciding
whether or not to drill. If the seismic test is conducted it will indicate
whether the site has (a) no structure-bad, (b) open structure-so-so, or (c)
closed structure-best. Regardless of whether a seismic test is conducted,
the wildcatter must decide whether or not to drill. If a well is drilled,
it will come up dry, wet, or soaking in oil. Each path through the tree
corresponds to a different sequence of choices and uncertainties and leads
to different profit or loss and, perhaps, other outcomes important to the
Because decision trees can be very large, for display
purposes it is common to use shorthand ways to avoid having to draw
duplicate parts of the tree. In the case of the above tree the node "a" in
the lower part of the tree is used to indicate that the nodes and branches
beyond this point are exactly the same as those beyond the node in the
upper portion of the tree labeled "a".
A decision tree can be quantified by associating
probabilities to the uncertainties and computing the value of the outcomes
associated with each path through the tree. To illustrate, the tree below
is an expanded version of the wildcatter's decision tree that explicitly
shows all of the possible paths through the tree. By convention,
probabilities are placed under the branches of the chance nodes and the
equivalent monetary values corresponding to the various paths through the
tree are placed under the terminal nodes. The monetary values may be
determined by discounting the cashflows or, if non-financial impacts are to
be considered, a method such as multi-attribute
utility analysis may be used to compute an equivalent monetary worth
that combines both financial and non-financial project outcomes.
Expanded oil wildcatter's decision tree
When a decision tree is quantified by
associating probabilities with chance nodes and values to terminal nodes,
the tree becomes a stochastic
decision model that can be
solved using a computational approach known as tree rollback. Tree
rollback is based on dynamic
programming and indicates the optimal choices contingent on the outcome
of each uncertainty. The rollback process begins with the values associated
with the terminal nodes of the tree and works backward to the initial
decision node. Rollback values are calculated and placed in brackets
besides the corresponding nodes. Depending on the node type, rollback
values are determined as follows:
- At a terminal node, the rollback value is simply the value associated
with the terminal node.
- At a chance node, the rollback value is the expected value (the probability-weighted average)
of the values on the branches emanating from the node.
- At a decision node, the rollback value is set equal to the highest
rollback value (or minimum cost, if negative) on the branches emanating
from the node.
To illustrate, the monetary values on the 3 top terminal
nodes are as follows: -$70,000 (if the well is dry), $50,123.4 (if the well
is wet), and $200,037 (if the well is soaking). Using the probabilities
shown under the branches, the expected value obtained from rollback is
$20,044.5 = 0.5 * -$70,000 + 0.3 * $50,123.4 + 0.2 * $200,037). This is the
value shown in the brackets to the left of the top-most green node.
Rollback values are repeatedly computed in this way until you reach the
first node in the tree. As shown, the rollback value for the "no seismic
test" option ($20,044.5) is less than the value of doing the test
($22,586.9), so the value-maximizing choice is to begin by conducting the
test. The tree also shows that if the test indicates no structure, the
optimal decision is not to drill. Conversely, if the test shows open or
closed structure, the optimal decision is to drill.
Decision trees provide an additional advantage that they
can be analyzed to provide value of
information calculations that indicate the worth of obtaining
additional information that would help to resolve uncertainties before
committing to a decision.
Decision trees are used in some project portfolio management tools, especially
tools aimed at pharmaceutical product development and other sorts of
R&D, as a means for addressing risk.
Also called a decision variable, a variable in a
decision model representing the
choice that is to be made in the context of a decision problem. As with
other decision models, the decision units for a project priority system
should be defined as part of the framing process. Among other things, framing requires
requires determining which quantities are to be treated as
decision variables and which ones are to be taken as fixed. The
quantities whose values are fixed (or relatively fixed) are typically called parameters.
Decision units are important because they determine the
granularity of the analysis, including spatial, temporal, and intensity
assumptions. When shopping for project
portfolio management tools, it is very important to understand the
restrictions that are placed on the definition of decision units. For
example, if you need a tool to help you prioritize capital projects, it
might be reasonable to consider one that evaluates "fund" versus "don't
fund" options for each project. However, such a tool might be useless for
evaluating maintenance projects if the appropriate decision unit is the
choice among alternative 5-year spending plans for programs consisting of
groups of similar assets.
A method used by a group of people to reduce biases in
estimates made by the group or other group judgments. The Delphi method is
typically conducted as an iterative process wherein a facilitator
repeatedly obtains judgments from each member of the group interspersed
with feedback of group responses and opinions. Group members typically
remain anonymous with regard to the opinions expressed and interact through
the facilitator in order to reduce biases in the estimates produced.
A condition that, if satisfied, ensures that the utility function describing a person's
preferences among alternatives with uncertain outcomes will have either a
linear or exponential form. The required condition is that if any arbitrary
amount D is added to the value of each possible outcomes to an alternative,
then the certain equivalent (what the
uncertain gamble is worth to the individual) will increase by exactly that
Dempster-Shafer (D-S) theory
A theory for reasoning about uncertainty similar to, but
different than probability theory. D-S theory shows how evidence from
different sources can be combined to arrive at a degree of belief
represented by a mathematical function called a belief function.
A belief function, like a probability function, assigns
a number between zero and one to propositions. However, while a probability
indicates the likelihood of the proposition, a degree of belief assigned by
a belief function indicates a degree of support provided by the evidence
for the proposition. A belief of 0 means no support for the proposition, a
belief of 1 means full support. The assigned numbers differ from
probability in that belief in a proposition (the proposition is true) and
its negation (the proposition is false) need not sum to 1, and both values
can be 0 (meaning no evidence for or against the proposition). The
mathematics provided by the theory for combining evidence and aggregating
beliefs can be regarded as a generalization Bayes theorem for updating subjective
probability distributions based on new information. The major difference
between the two theories is that the D-S theory is capable of combining
evidence and dealing with ignorance in the evidence combination process
The theory owes its name to work by Arthur Dempster and
his student, Glenn Shafer. Dempster developed an initial version of the
theory in 1967 in the context of statistical inference as a means for
combining evidence from different sources to arrive at a degree of belief.
Shafer then expanded the theory into a general framework for representing
epistemic uncertainty (uncertainty due to lack of knowledge). Since then,
numerous researchers, especially in the field of artificial intelligence,
have continued to expand and refine the theory.
Belief functions are most often used for combining expert
opinions, since experts often differ in their opinions with differing
degrees of credibility. Applications include auditing, medical diagnoses,
and other applications where information is gathered from semi-reliable
Also called a dependency structure matrix (DSM), a
method for documenting and analyzing the interdependencies between programs, projects, or project tasks. The capability for
constructing a dependency matrix is provided by some project management tools and some project portfolio management (PPM) tools . A
few PPM tools take into account the interdependencies expressed in a
dependency matrix when computing optimal project portfolios.
The rows and columns in a dependency matrix are labeled
with projects and/or programs and the cells represent dependencies. A
number placed in a cell indicates an estimated degree of dependency between
the project represented by the column and the project represented by the
The precise meaning of the numbers entered into the cells
is not standardized and differs depending on the application and the nature
of the dependencies. For example, in some versions, a zero or one is
written into each cell—zero indicates there is no connection between
the projects, one indicates that the project corresponding to the column
cannot be conducted unless the project corresponding to the row is
completed. In a version of the tool for projects to create new products,
users indicate in the diagonal cells the amount of revenue produced if the
corresponding projects are conducted alone; the numbers in the other cells
in the column indicate the amount of revenue produced if the column project
and the corresponding row projects are jointly conducted. In another
version, users indicate in the cell corresponding to a column and row the
"impact," expressed as a percent, that the row project has on the column
project, where impact could relate to impact on benefit, cost, schedule, or
Some PPM tools use a dependency matrix for portfolio
monitoring. For example, "traffic light" icons may be associated with
projects. A red light may indicate that a problem is forecast for a project
because a problem has occurred for one of the projects on which it is
dependent. A green light may indicate that a project is dependent on other
projects, but that so far no problems have been encountered by those
A project with
objectives or deliverables that are only slightly or incrementally
different from those provided by the organization's other projects. The
term is typically used for projects intended to modify existing products or
services, or that create new products that derive naturally from the firms
other product offerings. Projects that provide an add-ons, new packaging,
or manufacturing efficiencies are typically viewed as derivative
A software application that runs stand alone on a desktop
or laptop computer..
Non random. Typically applied to describe a model or
method of analysis whose outputs are fully determined by its inputs, with
no uncertainty or possibility of an alternative outcome. Compare with
The independence condition that, together with
preferential independence ensures
that a cardinal value function can be constructed with
the additive form.
Attribute X is said to be difference independent of other attributes if the preference difference
between any two attribute bundles that differ only in the level of X does not depend on the common levels of the remaining attributes.
In the context of choosing among job offers, for example, suppose location and salary are relevant attributes. If you prefer
working in New York city twice as much
as working in San Antonio at a salary of $50K, then you must continue to maintain that to same preference ratio (prefer New York
twice as much as San Antonio) if the salary offer is $80k.
discounted cash flow (DCF)
See net present value
A rate, expressed as a percentage, used to compare the
value of obtaining future versus present outcomes. The discount rate
appears in the formula for net present
value, where it is denoted r. The discount rate indicates
time preference in that it
specifies the return that would be required to make a decision maker
indifferent to delaying an outcome (e.g., If you could earn a 10% return,
would you be willing to postpone receiving your paycheck for a year?).
In business settings where a business has the opportunity to borrow funds, the The discount rate
is typically chosen to be the the average amount that the organization must pay to obtain funds (i.e., opportunity cost associated with making the investment)—If I can earn a return r per year from investing, then I
won't be willing to accept less than r if my current investment
delays my cash flow by a year. Discount rates are important to project portfolio management because they provide
a means for comparing projects that produce delayed costs and benefits.
discrete risk, discrete
An event, circumstance or condition that may or may not
occur, which would a project or its outcomes (e.g., technology failure,
strike, discovery of unexpected hazardous conditions, labor strike). A
discrete uncertainty may be characterized by the probability of occurrence
and the estimated consequences should the event occur. For comparison, see
A measurement scale
that allows for the assignment of only a finite number of possible values,
for example, a one-to-ten scale that does not allow assigning non-integer
values. For comparison, see continuous scale.
To approximate a continuous variable (one that could take
on an infinite number of possible values defined over some range) by a
finite number of possibilities. A common application is discretizing a
distribution so that it can be represented by a small number of
branches in a decision tree. If
the variable is uncertain and characterized by a continuous cumulative probability
distribution, denoted F(x), the continuous distribution is approximated
by a discrete distribution P(x) (a probability mass function) that assigns
a probability to each of the finite number of possible outcomes. Typically,
the discrete approximation is chosen to have the same mean, standard
deviation, and skew as the continuous distribution.
Discretizing a continuous probability distribution
Continuous variables are typically discretized for the
purpose of simplifying analyses, and many project portfolio management tools use the
technique, for example, when discrete scales are used to define the
outcomes of variables that are in reality continuous.
The inclusion of different types of investments within a
portfolio. Diversification is commonly used in financial investing to
reduce risk. Similarly, diversification tends to reduce uncertainty over
the total return generated by a portfolio of risky projects. Thus, project
diversification is often good for project portfolios. However,
diversification is not as effective when uncertain project returns are
highly correlated. For example, if
many projects would be similarly and significantly affected by the same
uncertainty (e.g., an economic recession or a change in currency exchange
rates), diversification will not be as effective at reducing portfolio
Also called dominance rule, a multi criteria decision making method wherein
are identified and eliminated from further consideration. The method proceeds by comparing the first two alternatives. If one
of the alternatives performs as well or better than the other on every criterion,
it is said to dominate the other alternative,
and the dominated alternative is discarded. Next, the undiscarded alternatives are compared against the third alternative.
If any of the alternatives are found to be dominated, they are discarded. The method proceeds until all dominated alternatives
have been eliminated.
The dominance method typically results in multiple
non-dominated alternatives. Therefore, so, if a ranking of the undiscarded alternatives is required, a
compensatory method is typically
applied for this purpose.
A alternative to a decision that is inferior to at least
one other alternative with respect to every relevant decision criterion. A dominated alternative has
disadvantages without any advantages. A decision matrix is useful for
identifying dominated alternatives. Eliminating dominated alternatives is a
typical screening step prior to more involved, multi-criteria analysis
do minimum scenario
The baseline scenario against which the additional benefits and costs of the with project scenario are measured
(often a synonym for the ‘without project’ scenario). The do minimum scenario is project option that includes all the necessary
realistic level of maintenance costs and a minimum amount of investment costs or necessary improvements needed in order to avoid
or delay serious deterioration of the organization's assets or to comply with safety standards. Do
The maximum amount of possible loss in a given decision
A term used to describe the action of moving from summary
level information to the more detailed information on which the summary is
based. Tools for project portfolio management (PPM) are often
advertised as providing dashboards
with drill down capability—clicking on summary information on the
dashboard causes the user to navigate to a more detailed level or record.
Drill down capability is made possible by arranging data in hierarchies
that start with general information and encompass increasing levels of
Refers to the level of care and analysis that should be
reasonably conducted prior to making business investment decisions. The
term is most often applied in the context of venture capital investment and
business mergers and acquisitions, where due diligence is regarded as the
essential means for preventing avoidable harm to the investing parties.
However, the concept applies to any important decision. Before investing
scarce resources to conduct a major project, a deliberate, documented
process should be undertaken to uncover and understand all of the
information relevant to the choice.
Due diligence is not just important for making good
decisions; it is also important for the defense of those decisions. In
legal disputes involving situations where projects have gone badly or
created significant health, environmental, or economic losses, due
diligence has been established as a legal obligation, and demonstrating due
diligence represents an important legal defense. Courts, however, have held
that due diligence requires not merely showing that the standard of care
for the decision was normal for the industry, but proving that what was
done is what a "reasonable and prudent" professional within the area would
do. Failure to meet this standard can give rise to civil and criminal
In the context of project
portfolio management, due diligence means that organizations should
institute a deliberate, documented, quality process for making major
project investment decisions. The
process should include and ultimately be based on an evaluation of the
consequences of doing versus not doing the project and the risks involved.
Due diligence for project selection doesn't just improve decision quality,
it ensures that tough choices can be defended in hindsight. Even if a
simplistic project evaluation technique, such as strategic alignment, could be shown
to be the "normal approach," organizations (especially those that depend on
projects for success or that operate assets or sell products that produce
public risks) would be well advised to utilize better and more defensible
project evaluation methods.
A type of mathematical solution technique wherein a
complex problem is broken into a series of interconnected and
similarly-structured sub-problems in such a way that the sequential
solution of the sub-problems results in the solution to the complex
problem. Dynamic programming is used as a solution technique in some
project portfolio management tools,
especially tools designed for projects requiring a series of decisions,
such as new product development projects. When applicable, dynamic
programming is a divide-and-conquer approach that can efficiently generate
solutions to complex problems.