|
Term
|
Explanation
|
|
multi-criteria analysis (MCA)
|
An umbrella term used to describe techniques that use more than one criterion to evaluate and judge performance. Multi-criteria
decision aiding techniques typically involve weighting the criteria to reflect the relative importance attributed to each criterion. Examples range from simplistic "rate and weight"
techniques to rigorous multi-attribute utility analysis. Other examples of MCA techniques include AHP, the balanced scorecard, ELECTRE, PROMETHEE, and SMART. Many project portfolio management tools use some form of MCA to rank or evaluate proposed projects.
|
|
multiobjective linear programming
|
A variation of the linear programming problem formulation wherein more than one linear objective function is specified. For example, choose values for the 3 variables x1, x2, and
x3 that maximize
y = ax1 + bx2 + cx3
and
z = cx1 + dx2 + ex3
subject to one or more linear constraints. As illustration, x1, x2, and x3 might be the
investment levels for 3 proposed new product projects, y might be total profit, and z might be total sales. If N is the available budget, the
constraint would be expressed by an equation of the form:
x1 + x2 + x3 ≤ N .
With a multiobjective linear program, no prioritization of the multiple objectives
is expressed or implied. In the example, it is merely assumed that more profit is preferred over less profit and more sales is preferred over fewer sales. Multiobjective linear programs
may be solved using a multiobjective analog of the simplex method for solving ordinary linear programs.
A key issue for multiobjective linear programs is that there is typically no solution that simultaneously optimizes all objectives. In the case of the example, there
may be no combination of funding levels for the 3 projects that produces the greatest possible profit and sales (for example, more funding for a new, low-cost product might increase
sales but reduce profits). Accordingly, multiobjective linear programming seeks a list of "non-dominated solutions." A non-dominated solution is one from which it is impossible to
improve performance on one objective without some sacrifice in at least one other objective. Thus, in the example, there might be several combinations of project funding choices which
would produce different levels of profit and sales. The approach is to list each funding combination wherein it is impossible to increase one (e.g., profit) without producing a decrease
in the other (sales).
In principal, multiobjective linear programming appears to have a major advantage relative to other multiobjective methods, such as multi-attribute utility analysis, that require as input difficult judgments regarding decision-maker willingness to make tradeoffs. The concept with multiobjective linear programming is to present to decision makers a variety of potential solutions, with no judgment about
which is preferred. In practice, however, there is a problem. Oftentimes, there will be a number of fixed non-dominated solutions, plus an infinite number of additional solutions that
represent linear combinations of the fixed solutions. If there are more than two objectives, it may not be possible to even plot the non-dominated solutions, making it very difficult
for decision makers to comprehend the options let alone choose from among those options. (See goal programming for another multiobjective approach
related to linear programming that avoids this problem.)
|
|
multi-project management
|
The simultaneous planning, monitoring, and management of multiple projects interconnected by their demand for common resources.
Multi-project management is more difficult than the management of a single project due to the need for integration and coordination. Tools for multi-project management generally provide
a system for standardized project workflows and reporting. However, in contrast with tools for project portfolio management, tools for multi-project
management do not typically provide capabilities for project prioritization or portfolio optimization.
|
|
N
|
|
|
|
natural scale
|
See scale.
|
|
net present value (NPV)
|
The traditional method for quantifying the financial attractiveness of a project based on equating project value to the present value (PV) of the project's present and future cash flows. NPV, also called discounted cash flow (DCF), represents the amount in today's dollars
by which all income projected from the project exceeds all costs.
Basically, NPV attempts to answer the question, "What is the equivalent, lump-sum worth of this project?" According to NPV logic, given two projects, the one with the
larger NPV should be preferred. Also, any project with a positive NPV should be viewed as a good investment.
NPV computes the present value for a project by discounting estimated future incremental cash inflows and outflows. The discount rate is chosen to represent a required rate of return or target yield for the capital invested, which is often chosen to the company's
weighted average cost of capital (WACC).
To accurately calculate a project's NPV, it is necessary to estimate the life-cycle cash flows that would result from doing the project— not just the project
costs, but also all of the financial benefits that would result from the project. For example, if a project improves productivity, the future cost savings that would result should be
included in the estimated cash stream. Cost estimates must reflect the total cost of ownership (TCO) perspective. Thus, costs include project
investment costs, future operating costs, and any "exit" costs associated with ultimately phasing out whatever it is that the project produces. Since taxes and the capital structure of
the firm can have a significant impact, cash flows should ideally be calculated "after-tax," accounting for depreciation, working capital,
and other considerations. More precisely, what should be estimated is the impact of the project on the company's so-called free cash flow, the
value actually available to shareholders and debt holders.
NPV cannot be used directly as a reasonable metric for ranking projects because it ignores the size of the projects being
compared. Larger projects tend to have larger NPV's. Thus, projects with large NPV's tend to consume greater portions of the available budget. However, NPV can be used to translate the
financial benefits expected from a project into an equivalent dollar value which, if divided by the project cost, can be a useful metric for ranking projects (assuming the projects are
independent).
By requiring that a single, nominal cash flow be identified for the project, NPV ignores uncertainty. This limitation may be addressed by using NPV in conjunction with
simulation. If alternative project cash flows are simulated, they may be used to generate a range or probability distribution of possible project NPV's. This distribution may be interpreted as describing the uncertainty over the actual
value that the project will generate. See expected net present value (ENPV) for more discussion.
The major limitation of NPV for project prioritization is that it underestimates the true value of projects (e.g., the impact of a project on shareholder value) because it leaves out "intangible" project benefits that are difficult to express as
incremental cash flows. One such omitted component of value is "option value," the value associated with options embedded in or created by the project which may allow management to
better respond to future risks and take advantage of future opportunities. See real options analysis for an explanation of option value. One
method for addressing this bias is to add to the estimated project cash flows additional dollar amounts that represent the equivalent dollar value of the non-financial project benefits.
See multi-attribute utility analysis for a method for doing this.
Another problem with NPV is that it is not always clear what discount rate should be chosen. According to finance theory, the correct rate is a risk adjusted discount rate equal to the return available from investing in securities equivalent to the risk of the project being
evaluated. Research on real options shows that the discount rate ought to be adjusted over time depending on how uncertainties are resolved and on the project-management strategy. Using
a constant discount rate for a project implicitly assumes that uncertainty increases over time in a specific way (geometrically). If the discount rate is adjusted upwards to account for
the risk of the project, there will be a bias toward short-term, quick payoff projects because project benefits that occur in the more distant future will be severely discounted.
|
|
network model
|
A model composed of multiple, relatively independent sub-models linked or interact in specified ways. Thus, network models are often used to represent systems composed
of multiple, interacting components or subsystems.
A network model is distinct from a hierarchical model. The latter has a top-down, tree structure such that each subsystem is linked to at most one "parent"
subsystem. With a network, each sub-model may have links to multiple parents.
The term network model is also used to describe a database structured as a collection of records with relationships among the data represented by links.
|
|
nominal group technique
|
A group decision-making method, developed originally by Andre Delbecq and Andrew Vandeven, that helps prevent the domination of discussion by a single person,
encourages more passive group members to participate, and results in a set of prioritized solutions or recommendations.
The basic process begins with each participant suggesting possible solutions or choices, typically recorded on a flip chart. After duplicate ideas are eliminated, each
person ranks the solutions and then anonymously votes points according to his or her ranking (e.g., 5 points to the first choice, 4 points to the second choice, etc.). The total points
each solution receives determines the group ranking, and the top ranked alternative is assumed to be the group's choice.
|
|
O
|
|
|
|
objective
|
Also referred to as decision objective, an explicit statement of a desired goal of implementing a decision, such as a decision whether to conduct a project. Decision analysts recommend that objectives be SMART—Specific, Measurable, Attainable, Relevant, and Time Based. Such objectives provide a
basis for defining attributes and performance measures as required by some multi-criteria analysis methods, including multi-attribute utility analysis. To be technically complete, the specification of
a decision objective requires specifying the object of value, its context, and the direction of preference. For example, "Increase (direction of preference) customer satisfaction
(object of value) regarding our company's products (context)."
|
|
objective function
|
A mathematical statement in equation form that defines a dependent variable (output) which is to be maximized or minimized through the assignment of optimal values to
independent variables (equation inputs).
|
|
objectives hierarchy
|
A graphic display of the objectives for a decision structured as a hierarchy. Objectives in the uppermost levels of the diagram
reflect broad or overarching values. Progress towards those objectives is achieved by meeting lower-level, subordinate objectives. The process of identifying and structuring objectives
into an objectives hierarchy typically leads to an improved understanding of what one should care about in a decision context. Creating an objectives hierarchy is also a useful initial
step for creating a decision model as recommended by formal decision framing processes.
Objectives hierarchies are constructed "from the top down," beginning with the specification of the overall fundamental objective. Typically, the overall fundamental
objective is relatively easy to identify—it defines the reason for interest in the decision situation. For example, my tutorial on prioritization
presents an objectives hierarchy for a decision of what to pack for a vacation. The overall fundamental objective is specified to be "maximize enjoyment."
Once the overall fundamental objective has been identified, it can be related to more specific objectives that explain what is required to achieve the overall
objective. The more specific objectives can often be obtained by asking "What do you mean by the higher-level objective? Or, "What, more precisely, needs to be achieved to achieve the
higher-level objective?" The resulting lower-level objectives can be further related to more detailed objectives, thereby building the hierarchical structure. Oftentimes, as in the
example below, the objectives in the hierarchy can be structured into different types or categories of objectives.
An example objectives hierarchy (for a public utility)
Sometimes, the process selected for structuring objectives does not impose any formal requirements for how objectives are defined. When this is the case, the resulting
structure is likely to contain a mix of objectives, goals, constraints, and metrics. Also, both fundamental or ends objectives and means objectives
(objectives that specify a particular means by which the desired end might be achieved) may appear in the same structure. Objectives hierarchies with such characteristics are often
referred to as value trees or value maps. Value maps are not very useful for creating decision models because they can't easily be converted into quantitative equations
for measuring project value.
However, if objectives are defined and structured so as to conform to certain principles, including being fundamental, non redundant and non-overlapping, well-defined,
measurable, and preferentially independent, the resulting objectives hierarchy provides a basis for constructing a decision model
in accordance with the requirements of multi-attribute utility analysis. In other words, objectives hierarchies (especially when used with influence diagrams) provide a method for defining performance measures and other metrics for
assessing alternatives and for deriving appropriate equations for combining the metrics into a measure of the value to be derived from each available option (i.e., obtaining appropriate
aggregation equations).
Some project management software tools and a few project portfolio management tools include aids for
creating objectives hierarchies.
|
|
open source
|
Relates to computer software for which the source code is freely available. Project portfolio
management tools that are open source are generally licensed and made available so as to allow modifications and redistribution of its source code.
|
|
opportunity cost
|
A cost attributed to the use of a scarce resource within a project. The opportunity cost is the value that could be obtained if the resource were employed in the best
alternative usage. Thus, it is the value of the opportunity foregone if the resource is used in a proposed application.
|
|
optimal
|
As used here in the context of project portfolio management (PPM), optimal means the project portfolio providing the greatest possible value to the organization subject to applicable constraints (e.g., limits
on available capital and other resources). Obviously, identifying optimal portfolios requires formal analysis, and this is (or ought to be) one of the primary goals of tools intended to support PPM.
|
|
optimization
|
A mathematical process of identifying from an allowed set of possibilities the choice or choices that are in some defined sense best. In the case of project portfolio management, the goal is to select from the set of feasible projects the best subset without exceeding the
constraints that limit the projects that can be conducted. For example, portfolio optimization might seek the set of projects that collectively provides the greatest total value to the organization and that can be conducted given the constraint on available budget. As described in the paper on mathematical methods, various mathematical techniques may be used to solve portfolio optimization problems.
Compared to prioritization, optimization is a more general and powerful tool for selecting projects. Prioritization seeks
to rank or list projects based on some measure of attractiveness assigned or calculated for each project (e.g., the project's benefit-to-cost ratio). However, if there are multiple
constraints that affect project selection or dependencies among projects, priorities are not unique (e.g., the incremental value gained from adding a project to the portfolio compared
to the incremental cost depends on the other projects that are included within the portfolio and also on the impact on ability to achieve the various constraints that may be binding).
Optimization takes a more holistic approach, it merely tells you, based on the constraints you have set, whether or not individual projects are in or out of the optimal project set.
There is no individual measure of project attractiveness; the contribution of individual investments is measured within the bundle. While prioritization can, in some cases, approximate
an optimal solution, a truly optimized result will always identify the most valuable portfolio. See the comparison of efficient frontiers versus ranking
curves for more discussion.
|
|
optimization engine
|
Also called optimizer or solver, an algorithm for performing optimization. The
term most frequently refers to software or the component of a software program that automates the solution to an optimization problem. Users "run" the optimization engine to find
optimal solutions, and they may re-run the engine multiple times to see how the solution changes based on specifying different objective
functions or constraints.
Optimization engines are distinguished by the nature of the mathematical problems that they are designed to solve, for example, linear
programming, quadratic programming, etc. Some project portfolio management (ppm) tools are advertised as
containing an optimization engine. The component so referred to might be code written specifically for the ppm tool, or an optimizer "plug in" provided by a third-party suppler such as
FICO's "Xpress" or OptQuest's "OptFolio".
|
|
organ pipe chart
|
A chart, provided by some project portfolio management (PPM) tools, showing the priority order for adding projects to a portfolio. The columns ("pipes") in the chart show the funding levels (the y-axis) at which the various projects (listed across the top) are
recommended for funding. The organ chart is most useful in situations where the are interdependencies among projects, because it shows increments in funding at which groups of
interdependent projects should be added simultaneously. Also, if the tool solves for the efficient frontier, the chart will show funding levels at which particular projects funded at
higher portfolio funding levels are omitted from the portfolio to make room for other projects that utilize more of the budget so as to create the greatest possible value.
An example organ pipe chart
|
|
outranking (OR) method
|
A category of multi-criteria analysis decision-aiding methods that seek alternatives that, based on paired comparisons, dominate ("outrank") other alternatives. In contrast to the analytic hierarchy process (AHP)
and multi-attribute utility analysis (MUA), which focus on building utility functions for evaluating
alternatives, outranking methods base results on data obtained from criterion by criterion comparisons conducted using ordinal scales. The two most well-known outranking methods are ELECTRE and PROMETHEE.
|
|
|
Previous