Lee Merkhofer Consulting Priority Systems

Glossary of Technical Terms Used in Project Portfolio Management

Mastering a new field of study requires learning the language, especially learning the meanings of the technical terms commonly used by practitioners in the field. According to one estimate, a technical professional must learn at least 3,000 terms specific to his or her field. Technical jargon is often confusing to newcomers, but technical terminology allows those with expertise to communicate precisely and efficiently. If you misunderstand or misuse technical terms, you will be seen by others as lacking credibility and experience. You won't be recognized as knowledgeable within a technical area until you learn its technical jargon.

In the case of project portfolio management (PPM), learning the language can be particularly difficult. PPM uses terms from diverse fields. Some of the terms involve tricky concepts, and some denote complex mathematical calculations. PPM tool promoters have in some cases invented new terms to describe variations on traditional techniques. If you take the time to understand the relevant technical concepts, you'll be able to cut through the marketing hype and benefit from honest discussions about how you and your organization can benefit from PPM.

This glossary provides explanations for terms used here and elsewhere to describe PPM methods and tools. Be aware that some of these terms may have different meanings in other contexts.

Please contact me if you would like to suggest additional terms or if you wish to offer or request clarifications.





























access permissions

Specifications that control who can read and/or alter computer files or directories. Project portfolio management (PPM) tools, especially web-based tools, like other sensitive software intended for multiple users, often utilize access permissions to ensure that users are only able to view and alter information contained in the software in ways that are appropriate given their specified roles.

additive equation

Also called an additive function, a function f(x1, x2,...,xN) with some number of arguments N, such that for any set of argument values x1, x2,... ,xN for which the function is defined,

f(x1, x2,...,xN) = f1(x1) + f2(x2) + ... + fN(xN)

In other words, an additive equation is an equation that can be expressed as a summation of functions each of which has just a single argument.

additive independence

Also called utility additive independence, the condition that ensures a utility function has the form of an additive utility function. If additive independence applies, the utility function may be expressed as:

U(x1,x2...xN) = w1U1(x1) + w2U2(x2) ... + wNUN(xN)

Two attributes, 1 and 2, are called additive independent, if the preference between two lotteries (defined by a joint probability distribution on the two attributes) depends only on their marginal probability distributions (the marginal probability distribution on attribute 1 and the marginal probability distribution on attribute 2).

Ugly bumper

Ugly bumper

An example, though contrived, may help clarify the test for utility additive independence. Suppose you are interested in purchasing a high-performance European sports car. For simplicity, assume you care only about 3 attributes: cost, horsepower, and looks. You've identified several cars that interest you. Since utility functions are only appropriate if there is uncertainty over outcomes, assume that the government is considering two new regulations. One is an environmental law that would require you to install on your selected car an emissions-reducing device that would cut the car's horsepower by 10%. The other is a safety regulation that would require you to attach a plastic bumper at the rear end of your car that, in your opinion, would make your selected car less attractive.

The lotteries below could be used by you to determine whether the horsepower and looks attributes are additive independent (the cost of the auto is assumed to be same for each lottery).

Example utility independence test

Example lotteries for testing for utility additive independence.

If you are indifferent between these two lotteries, your utility function will be additive in horsepower and looks, otherwise not.

If there are more than two attributes in your utility function, as in the above example, then you must repeat the test for all pairs of measures. If any test indicates that the additive form cannot be assumed, then you'll either need to use a more complex form for the function, or you'll need to change the definitions of objectives and their performance measures until independence does apply.

Although a highly technical concept, additive independence is important for constructing value models for projects whose outcomes are uncertain. Unfortunately, it is common for organizations to use weight-and-add aggregation equations for prioritizing projects in situations where additive independence does not apply. If this error is made, the prioritization model will not correctly rank projects.

additive utility function

Analogous to an additive value function, but applies to a utility function rather than a value function. In other words, an additive utility function is a utility function that is expressed as an additive function of single attribute utility functions. If a utility function is additive, its assessment is much easier. In this situation, the assessment of the multi-attribute utility function does not require considering preference tradeoffs among more than two attributes simultaneously, nor is it necessary to consider lotteries with more than one level of an attribute being varied. In fact, the assessment can be accomplished by assessing the single attribute utility functions and the weights.

additive value function

Also called additive value model, a value function expressed as a weighted sum of single-attribute value functions:

V(x1,x2...xN) = w1V1(x1) + w2V2(x2) ... + wNVN(xN)

The additive value function, like other value functions, assigns a number indicating a decision maker's preference for an alternative or outcome characterized by attributes, denoted in the equation as x1,x2...xN. With the additive function, the value so assigned is a weighted sum of the decision maker's preferences over the individual attributes, as specified by the single attribute value functions, Vi(xi).

The additive form of a value function is often used in models for prioritizing or selecting projects because it is relatively easy to create and it produces an intuitive model form wherein different components of value are weighted and added. As an example, a government transportation agency might be considering projects addressing a segment of a highway. Suppose there are only two objectives: to reduce the commute time for travelers and to reduce the frequency of accidents. With an additive value model, single attribute value functions would be constructed to indicate the value of reducing commute time and the value of reducing the frequency of accidents. A project's total value would then be computed by weighting and adding its commute time value and its accident reduction value.

The single-attribute value functions, Vi(xi), are, in effect, scaling functions that translate the individual attribute measures, expressed in their natural units (e.g., annual numbers of accidents), into units of value. As in the above example, attributes typically indicate levels of achievement relative to objectives. By convention, value is typically expressed on a scale of 0-to-1 (or 0-to-100), where a value of 0 is assigned to the least desirable level that the attribute could have (e.g., zero advancement of the objective). Alternatively, the value zero is sometimes assigned to the least desirable attribute level achieved by any of the alternatives under consideration. Likewise, the value of 1 (or 100) is typically assigned to the best-possible attribute performance (or the best performance achieved by any of the alternatives). Typically, weights are scaled to sum to 100 (or 1), so that value, as expressed by the additive model, ranges from 0 to 100 (or zero to 1). Regardless, the resulting value number is directly proportional to preference; for example, an alternative with a computed value of 100 is twice as desirable as one with a computed value of 50. In technical terms, it is a cardinal utility.

Value functions, like utility functions, may be encoded from decision makers by seeking answers to questions regarding preferences for outcomes. The number of required questions and answers increases greatly with the number of attributes used. However, if the value function is additive, it may be constructed in two steps: (1) assessing the single attribute value functions and (2) assessing swing weights. The weights may be assessed using the swing weight method.

For the additive value function to apply, independence assumptions must hold. The most common approach to proving additivity for a value function is to demonstrate mutual preferential independence and difference independence. Preferential independence basically requires that the decision maker's preferences for achieving any level of performance against any objective does not depend on the performance achieved against any other objective.

Value functions specified over single attributes are sometimes referred to as conditional value functions. Conditional value functions express preferences over one attribute assuming other attributes are set to specific levels. However, with the additive value function, the particular levels of other attribute are not important to the assessment of single-attribute value functions. With additivity, these value functions will be the same for any levels of the other attributes. Mutual preferential independence thus makes it possible to discuss single value functions over each of the attributes.


As used in agile software development, refers to software development practices and methodologies wherein requirements and solutions evolve through collaboration among cross-functional teams, typically involving customers, designers, and developers. The goal is to promote timely, responsive solutions based on frequent reassessment of requirements and interim deliverables. The most common agile method, known as Scrum, utilizes multiple small teams working in an intensive and independent manner, with daily meetings, execution of tasks in brief, high-intensity work sessions, and frequent reviews aimed at identifying ideas for improvement. Similarly, agile, as used in agile project management, refers to project management methodologies that emphasize collaboration, quick delivery time, and ability to respond to changing requirements. Six sigma and PRINCE2 are examples of a methodologies that incorporate principles of agile project management.

aggregation equation

As used on this website, an equation used to compute a metric (number) for prioritizing (ranking) projects. Most methods for prioritizing projects are multi-criteria methods, since it is typically assumed that multiple criteria are relevant to determining project priorities. Thus, an aggregation equation typically includes multiple variables, each of which represents a different consideration for choosing projects. An aggregation equation also may include weights meant to represent some concept of the relative importance of the various criteria.

The exact formula used for the aggregation equation differs greatly from method to method, since different prioritization methods use different logics for ranking projects. Thus, determining what metric a prioritization method uses and how that metric is calculated is of central importance to understanding and judging the quality of a prioritization method.

According to decision theory, independent projects should be prioritized based on the ratio of project value to project cost. According to multi objective decision analysis (MODA), the discipline concerned with implementing decision theory for situations where the decision maker has multiple objectives, project value may be computed using a function called a utility function (if the prioritization method includes consideration of uncertainty and risk) or a value function (if the prioritization method ignores risk and uncertainty). MODA provides a step-by-step process for deriving utility and value functions and, thereby, provides a way to determine an aggregation equation for prioritizing projects based on decision theory. Because the methods for valuing and prioritizing projects advocated on this website are often based on MODA, by aggregation equation I typically mean a utility function or value function derived using the methods of MODA.

aggregated indices randomization method (AIRM)

A multi-criteria analysis (MCA) method applicable to decisions with uncertainty. AIRM evaluates "objects" using multiple, single-index criteria interpreted as being characteristics of the object associated with a single or specific point of view. The single points of view are synthesized into one, general point of view via an aggregation equation. The aggregated index value is determined by the single indices values and varies depending on weights, which are interpreted as measures of the relative significance of the corresponding single criteria indices. Only ordinal information about weights is assumed, so that weights can be considered random and restricted to intervals based on a system of equations defined by equalities and inequalities.

The distinguishing feature of AIRM is its ability to compute the single indices, weights, and aggregation equation using non-numeric (ordinal), non-exact (interval) and incomplete information. AIRM was developed in the early 20th century by a Russian naval applied mathematician.


A procedure composed of a sequence of instructions or steps for solving a problem (typically a problem expressed mathematically).


A course of action actively being considered in the context of a decision making process. Projects are the alternatives for the organization's project portfolio. Alternatives are actions available to a decision maker to achieve objectives.


A systematic approach to building understanding based on identifying and investigating the component parts of a whole and their relationships. The word can be traced to ancient Greek terms meaning "breaking up" and "a loosening". The technique has been applied in the study of mathematics and logic for thousands of years, though analysis as a formal concept is a more recent development.

Today, the most powerful methods of analysis involve applying theories and techniques developed within specific fields of expertise, such as those from the natural science, social science, and decision science. Project prioritization and portfolio optimization, for instance, rely heavily on analytic methods from the field of decision analysis.


A person who conducts analysis, short for systems analyst or decision analyst. See decision maker for more information,

analytic hierarchy process (AHP)

A popular, theory-based, decision-aiding approach, developed by Thomas Saaty in the early 1970s. Like multi-attribute utility analysis (MUA) and outranking methods, AHP is a form of multi-criteria analysis. The approach involves decomposing a decision problem into a hierarchy of sub-problems. As illustrated below, the hierarchy contains the decision goal, the objectives or criteria to be achieved, and the alternatives to be evaluated using the criteria. Once the hierarchy is built, the various elements of the hierarchy are evaluated, typically using paired comparison. As described below, AHP provides methods for quantifying the elements of the hierarchy and for evaluating the alternatives. Some project portfolio management (PPM) tools use AHP for project prioritization. There are many variations of AHP and how it is applied, making it difficult to provide general statements about AHP's effectiveness for prioritizing projects.

MACBETH decision structure

AHP decomposes a decision problem into a hierarchy of subproblems.

As originally defined by Saaty, AHP involves asking decision makers to express their preferences over pairs of elements in the hierarchy with respect to the hierarchy element above them. For example, starting at the level of alternatives in the hierarchy, the subject might be asked, "With regard to improving financial performance, do you prefer alternative A or alternative B, and by how much?" A nine-point ratio scale is recommended for specifying strength of preference (each score on the scale is defined qualitatively, for instance, 1 = "equal preference" 3 = "slight preference," 5 = "strong preference," etc., see this example). Reportably, Saaty established 9 levels for his scale because his experiments showed that individuals cannot compare more than seven, plus or minus two, objects at a time,

Understanding the mathematical procedure by which a ranking of alternatives is obtained requires an understanding of matrix algebra. The key things to know are that a square, (n × n) matrix can be multiplied by another matrix of the same size to produce a product (n × n) matrix. Likewise, a (1 × n) vector can be multiplied by an (n × n) matrix to produce another (1 × n) vector. An eigenvector of a matrix is a vector that when multiplied by that matrix simply scales the entries in the vector; in other words, the vector doesn't change direction, it merely changes size. The eigenvalue associated with the eigenvector is the amount by which the vector entries are scaled.

Starting with the level in the hierarchy represented by alternatives, the results of the pairwise comparisons are entered into matrices, one matrix for each criterion. For example, if there are 6 alternatives, the matrix for comparing alternatives with respect to a criterion will have 36 cells within which the results of comparisons are placed. The comparison s are expressed using a nine-level ratio scale as described above. If, for instance, alternative A is judged to be "extremely more desirable" than alternative B, then, according to the scale, a "9" would be placed in the cell (above and to the right of the diagonal) corresponding to A's row and B's column. The scores summarizing the results of the other alternative comparisons are likewise placed in their appropriate cells (above and to the right of the diagonal). Down the diagonal of the matrix, "1's" are placed, indicating that every alternative is "equally preferred" to itself. Below the diagonal, the reciprocal results are placed; if A was judged "extremely more desirable than B," then B must be "extremely less desirable than A," in which case "1/9" would be placed in the corresponding cell. The matrix of alternative comparisons (called a comparison matrix) for each criterion represents a linear system of equations (a matrix equation) describing the relative desirability relationships among the alternatives.

Pairwise comparisons are likewise made at the next higher level of the hierarchy to obtain a system of equations for the importance of the criteria. For example, criterion 1 may be judged two times as important as criterion 2, criterion 2 judged three times as important as criterion 3, and criterion 1 judged four times as important as criterion 3 (Note the inconsistency, since the first two judgments suggest that criterion 1 ought to be six times as important as criterion 3). An important characteristic of AHP is that it allows for inconsistencies in input judgments.

To determine the relative weights of the criteria, an eigenvector for the comparison matrix is found. One easy way to obtain an eigenvector of a matrix using a computer algorithm involves repeatedly raising the matrix to successive powers using matrix multiplication and then computing the eigenvector of the matrix. In other words, the pairwise matrix obtained by comparing the criteria is multiplied by itself to obtain a product matrix. An eigenvector is found for the product matrix by summing the entries in the rows of the product matrix, totaling the row sums, and then normalizing the row totals by dividing the row sums by the row totals. This process is repeated by squaring the product matrix and then normalizing the row totals from the new product matrix in the same way to obtain the eigenvector. The process is continued until the resulting eigenvector solution does not change much from that obtained in the previous iteration. At this point, the computed eigenvector provides the relative weights of the criteria.

In the same way as described above, an eigenvector is found for of the alternative pairwise comparison matrices. The eigenvector in each case provides the relative ranking of alternatives under each criterion. Finally, multiplying the matrix showing the ranking of each alternative under each criterion by the vector of criteria weights gives the overall alternative desirabilities and rankings.

AHP is similar in many ways to MUA. Both are widely used and both reflect a philosophy that alternatives should be selected based on how much they are preferred by decision makers. The approaches, however, reflect somewhat different philosophies. MUA is regarded as a purely "prescriptive" approach in that it is aimed at identifying the "correct" decision that a "rational" decision maker would take. AHP, in contrast, is at least somewhat "descriptive" in that it expects errors and inconsistencies in the inputs that people provide and seeks a solution that best accommodates those errors. It can be shown mathematically that the eigenvector approach described above provides this solution. The differences in philosophy and methods for eliciting preferences have spurned a lively debate among academics and practitioners over the pros and cons of the two approaches.

AHP's chief advantage is ease of application. The paired comparison questioning process can be presented pictorially, and decision makers typically find it easy to make the necessary judgments. AHP produces a decision model that is simpler than that typically produced with MUA in that AHP does not require the intermediate step of estimating the consequences of projects. AHP is particularly well suited to group decision making since the group members do not need to be in agreement. Among AHP's outputs are measures of the consistency in the input judgments, which can help participants reduce the inconsistencies expressed in their comparisons. Also, it is relatively easy for a group to structure a complex decision into the hierarchical structure required by AHP, and the process helps participants to understand the problem as well as each others' thoughts and opinions.

Saaty has provided an axiomatic theory to support the basic approach to applying AHP. However, the theory has been criticized based on the observation that AHP can produce "rank reversals," a result wherein the addition of a new alternative can change the ranking of existing alternatives, even though the new alternative does not influence the costs or benefits of the existing alternatives. For example, it is possible with a project prioritization system based on AHP to have the proposal of a new project that might be ranked 8th cause a project ranked 4th to move up to 3rd. While such behavior is undesirable and raises questions regarding the logical defensibility of AHP, supporters have pointed out that rank reversals rarely occur in practice.

AHP's cons include the very large number of comparisons that must be made if there are a large number of alternatives and/or objectives to be considered (that number can grow exponentially with the size of the decision problem). For this reason, applications of AHP to project prioritization often use a variation of AHP in which the preference comparisons are expressed for objectives, not for the projects. For example, "Compare the relative importance of the objectives 'improve time to market' and 'improve financial performance.'" The answers are used to derive a set of weights interpreted to represent the relative importance of the objectives. Projects are then scored to indicate their contributions to each objective (e.g., no contribution = 0, slight contribution = 0.1,..., excellent contribution = 1). The scores are weighted and added to obtain an overall measure for ranking projects.

Unfortunately, this common approach to applying AHP may not accurately prioritize projects (it also violates the axioms of Saaty's theory). It is not possible to obtain meaningful preference comparisons between objectives unless the amounts of improvement are specified (e.g., How can I say how much I prefer the objective "improve financial performance" without knowing by how much financial performance would be improved?). Also, it is not correct to weight and add performance scores unless the value of achieving a given level of performance on one measure does not depend on the level of performance achieved on any other measure, a condition known as preferential independence.

An alternative approach to applying AHP overcomes the above problems by using the so-called swing weight method to define specific amounts of improvements for expressing preferences and by applying tests to ensure that preferential independence holds. In effect, the resulting approach to developing the model for valuing projects is then consistent with MUA, however, Saaty's AHP's pairwise comparison technique is used to determine the weights.

A final limitation important for applications to project prioritization is AHP's inability to directly address risk. AHP does not allow probabilities to be assigned to reflect uncertainty over project performance. Decision makers can implicitly factor uncertainty into their pair-wise comparisons, however, this is difficult to do. Although fuzzy logic has been proposed as a means for addressing ambiguity within AHP, nearly all applications of AHP to project evaluation are conducted assuming most likely scenarios for project performance.

analytic network process

A generalized form of the analytic hierarchy process (AHP) used in multi-criteria analysis. Whereas AHP structures a decision problem into a hierarchy of components consisting of a goal, decision criteria, and alternatives, ANP structures the problem as a more general network, a form that allows for feedbacks and other forms of interdependencies among the components.


Term used to describe the methods of analysis used to apply data and mathematical logic to help make decisions. The term usually refers to the application of more sophisticated forms of analysis. Project portfolio optimization is an application of analytics.

application lifecycle management (ALM)

The process of managing the phases of the life of a software application, including planning, coding, development, testing, release, and support.

Application lifecycle management

Various tools are available to support ALM, addressing such issues as requirements management, architecture, coding, testing, tracking and release management. Project portfolio management tools aimed at IT project portfolios typically include or may be supplied with modules for supporting elements of ALM.

application portfolio management (APM)

The management of an organization's various computer software applications as a portfolio. APM may be viewed as a special case of project portfolio management (PPM) wherein the "projects" are investments intended to obtain optimal performance from the firm's portfolio of computer software applications. PPM tools aimed at IT project portfolios may include specialized techniques for managing software applications.

application program

Any self-contained software program that performs a specific function directly for the user. This is in contrast to system software such as computer operating systems that provide services to support application programs.

application programming interface (API)

An interface provided by a project portfolio management tool or other application program that defines how that program can access the computer's operating system or other programs to request data or services.

application server

Also called appserver, a type of server designed to install, operate and host applications and associated services for end users. Many project portfolio management tools rely on application servers. The user interface is often web based and accessed through a browser, but it may be through other means as well.

artificial intelligence

A branch of computer science concerned with making computers behave like humans. Artificial intelligence is concerned with the theory and development of computer systems able to perform tasks that normally require human intelligence, such as visual perception, speech recognition, and translation between languages. Prioritizing projects, for example, could be considered an application involving elements of artificial intelligence.

asset management

The deliberate, long-term management of an organization's assets, usually physical assets, with the goal of maximizing their contribution to the achievement of the organization's objectives. Asset management is of critical importance to organizationsthat that acquire and utilize expensive assets. Examples of organizations that view asset management as an important function include companies in the energy, mining, and oil and gas industries. Asset management typically involves making decisions about when to create and acquire assets, how to use them, their repair or replacement, and their ultimate disposal. Project portfolio management applied to asset intensive organizations often focuses on asset management and may be referred to as such.


A characteristic of an alternative or the outcome to a choice of an alternative viewed as important to decision makers. An attribute is a potentially useful means for quantifying and evaluating achievable levels of an objective. In the context of project prioritization, attributes are measurable characteristics relevant to the evaluation of candidate projects; they are metrics, which may be specified objectively (i.e., observed and measured) or subjectively (i.e., estimated based on judgment). An attribute, when assigned a number expressed in some appropriate unit of measurement or in accordance with some scale, quantifies some consideration relevant to prioritizing or otherwise evaluating that project, for example, its cost, how long it would take to implement, or what outcomes it might produce. The number assigned to an attribute may be a point estimate; that is, a single number intended to represent a best estimate. Alternatively, if there is uncertainty, a range or probability distribution might be assigned to the attribute. In addition to being measurable, attributes should be unambiguous, meaning that there is no ambiguity in interpreting the meaning of the attribute or the numbers that are assigned to it. In the context of multi-attribute utility analysis (MUA) attributes are typically defined to quantify the degree to which an alternative achieves various decision-maker objectives.

Attributes may involve natural scales, constructed scales, or proxy measures. An attribute with a natural scale quantifies the attribute based on commonly used, widely accepted units. For example, cost, expressed in dollars (or euros, rubles, yen, etc.) is a natural measure for project cost. Number of fatalities is a natural measure if loss of life is an attribute of concern. Most natural measures are expressed in some unit of measurement (e.g., dollars, fatalities) and can be counted or physically measured.

A constructed scale is a scale that defines different levels for the attribute in terms of descriptions or definitions for each level of the scale. For example, while it might be difficult to come up with a natural measure for "corporate brand image," it might be possible to construct a 1-to-5 scale consisting of 5 different verbal statements describing various negative-to-positive customer perceptions of the corporation. A defined impact scale is a constructed scale wherein the scale numbers are associated with narrative descriptions of discrete levels of impact. In other words, the scale levels describe the "delta" or amount of change that might result from the selection of an alternative.

A proxy measure is an indirect measure selected because there is a presumed relationship that exists between it and the relevant project characteristic. For example, if the company concerned about its image participates in a market analysis survey comparing customer perceptions, its ranking relative to its competitors might serve as a proxy for its brand image. Like natural measures, a proxy measure usually involves a scale that is in general use and that can be counted or physically measured.

The specification of attributes and their associated measures is a critical step in the construction of a decision model, as the choices made strongly affect the accuracy, defensibility, practicality, and usefulness of the model. A variety of techniques are available for identifying an appropriate set of measurable attributes, including influence diagrams (graphical representations of the different factors and relationships important to understanding the concern) or simple brainstorming (e.g., asking, "What information would you like to have to evaluate this concern?").

audit trail

In the context of project portfolio management (non-accounting sense), evidence in the form of records, references, data or documents that enable a user to trace the path of assumptions made, data changes, decisions, or other past actions that are critical to understanding the results obtained.