*A PDF version of this article can be downloaded here

Introduction

The scenario is the archetypical product of futures studies because it embodies the central principles of the discipline:

• It is vitally important that we think deeply and creatively about the future, or else we run the risk of being surprised and unprepared.
• At the same time, the future is uncertain so we must prepare for multiple plausible futures, not just the one we expect to happen.

Scenarios contain the stories of these multiple futures, from the expected to the wildcard, in forms that are analytically coherent and imaginatively engaging. A good scenario grabs us by the collar and says, “Take a good look at this future. This could be your future. Are you going to be ready?”

As consultants and organizations have come to recognize the value of scenarios, they have also latched onto one scenario technique--a very good one in fact--as the default for all their scenario work. That technique is the Royal Dutch Shell/Global Business Network (GBN) matrix approach, created by Pierre Wack in the 1970s and popularized by Peter Schwartz (1991) in the Art of the Long View and Kees van der Heijden (1996) in Scenarios: The Art of Strategic Conversations. In fact, Millett (2003, p. 18) calls it the “gold standard of corporate scenario generation.”

While the GBN technique is an excellent one, it is regrettable that it has so swept the field that most practitioners do not even know that it is only one of more than two dozen techniques for developing scenarios. There are so many approaches and techniques that go by the term scenario that Millett (2003, p. 16) says that “resolving the confusion over the definitions and methods of scenarios is the first necessary step to bring the value of scenario thinking and development to a wider audience.” A number of overview pieces have been published recently that respond to Millett’s requirements. First, we will address the confusions and definitions, describe our research approach, then review the overview pieces, and finally move into the analysis of the specific scenario techniques.

Confusions

This section addresses three primary confusions in the scenario literature.*

1. Perhaps the most common confusion when discussing scenarios is equating scenario development with scenario planning. We suggest that “scenario planning” has more to do with a complete foresight study, where scenario development is concerned more specifically with creating actual stories about the future. Scenario planning is a far more comprehensive activity, of which scenario development is one aspect.

2. A more subtle confusion is equating the term “scenario” with “alternative future.” In other words, all descriptions of alternative futures are deemed to be scenarios. A more narrow definition of scenario would focus only on stories about alternative futures. With this narrow definition, other forecasting methods might produce alternative futures, but not scenarios. In practice, however, the broader definition of scenario as alternative future, whether they are in story form or not, has prevailed. Thus, the complete collection of methods for scenario development includes almost all forecasting methods since they also produce alternative futures. In fact, very little is said about the actual creation of the stories in most methods. More attention is paid to generating the scenario kernel or logic, which can be done by any number of methods. We decided that it does not make sense to fight the battle for a narrower definition, and thus our list of methods is based on current practice and includes the incorporation of forecasting methods whether or not they produce a story.

3. The third confusion involves equating the terms methods and techniques. These terms are used interchangeably in the literature and in practice. There are subtle differences in the terms, with method being focused more on the steps for carrying out the process and technique focusing more in the particular way in which the steps are carried out. As above, however, we bow to the practicalities that the terms are used interchangeably, and do not see it useful to try and make the distinction at this point.

Definitions

Being a new field, futures studies is blessed with an abundance of creative and entrepreneurial practitioners who develop excellent approaches and methods to suit the needs of their clients. After a while, however, the growth becomes chaotic. One solution, as noted above, is to focus on one technique and stick with that. While that solution does reduce the chaos, it doesn’t make the best use of the techniques that others have created and are using.

But even the most basic vocabulary is used every which way in this field. Therefore, before beginning our review of scenario techniques, we have to decide on what a technique is in the first place, as opposed to an approach, or a method, or a tool. Therefore, we offer the following (small) glossary to distinguish these terms from each other so the reader knows what we are talking about and in hopes that others might use the terms in a similar fashion.

We begin first with a project. The futures project is the largest unit of professional work. It includes the sum total of the objectives, the team, the resources and the methods employed in anticipating and influencing the future. Projects may be simple, involving just one product and technique, or complex, involving many steps each of which produces one or more products and uses one or more techniques.

The process that one employs in conducting a project is the approach. The approach consists of an ordered series of steps to accomplish the objectives of the project. Every project has an approach, whether it is explicitly articulated at the beginning or not. Some approaches are widely practiced, such as the approach to develop a strategic plan.

A generic approach to a comprehensive foresight project is outlined in the following six steps:

Table I. A Generic Approach to a Comprehensive Foresight Project

This approach was used to classify best foresight practices in a forthcoming publication (Hines, 2006).

There are many other examples of comprehensive approaches to foresight. At the Association of Professional Futurists’ 2004 Professional Development Conference, two of these were described:

· The Futures Lab in Austin, TX uses an approach to product and business development that they recently described in Futures Frequencies (Woodgate, 2004).

· The Futures Management Group in Eltville, Germany uses a “lenses” approach to strategy development, as described in Der ZukunftsManager (The Future Manager) (Micic, 2004).

In fact, most professional futurists and consultants use a favorite approach that they have honed over time.

Each approach produces one or more products or deliverables that satisfy the objectives of the project. The product is the final result of the work done in the approach—as a report, a database of trends, scenarios in various forms, a strategic plan and many more. Usually each step in the approach generates a product and together they form the deliverable from the project.

A method or technique is the systematic means that a professional uses to generate a product. We found that method and technique are used rather interchangeably in the literature so it is hard to pick just one. Method carries a solid, organized, even an academic connotation where technique seems to relate more to style than to substance. In a review of terms in articles about scenarios published in Futures over the last few years, authors used both terms although they used technique quite a bit more.[1] So we will go with that for this review.

A tool, another term often confused with method or technique, is more concrete. A tool is a device that provides a mechanical or mental advantage in accomplishing a task. Tools are things like video projectors, questionnaires, worksheets and software programs. By the same token, scenarios and plans are not tools. Some of the best known tools in the field are Michel Godet’s (2003) Toolbox and the Parmenides Foundation’s Eidos tool suite--formerly Think Tools (Lisewski, 2002).

Finally, an exercise or activity is a unit of activity within a lesson performed for the sake of practice and to acquire skill and knowledge. It may be, of course, that the skill or knowledge is applied right away in the same workshop as part of project work.

So much for the general definitions; now we define the topic of this paper—the scenario. Despite its ubiquity, or perhaps because of it, we found more than two dozen separate definitions of scenarios in the literature, and that is probably not all. Suffice to say that a scenario is a product that describes some possible future state and/or that tells the story about how such a state might come about. The former are referred to as end state or even day in the life scenarios; the latter are chain (of events) scenarios or future histories.

Research Approach

The starting point for this research was collecting descriptions of the methods we had amassed over the thirty-year history of teaching scenarios in the Master’s program at the University of Houston. We then supplemented our list with literature and web searches to identify methods that had escaped our attention.

Surveying the scenario development field is no mean feat, but we believe we have captured most of it. The literature contains overview pieces that review the field (e.g., Notten, 2003; Bradfield, 2005; and Borjeson in press) and methodological pieces that describe a specific scenario technique.

We began by scouring the key methodological publications in the field to see what they said about scenarios. Among the sources of this material were:

• Books –Schwartz (1991), van der Heijden (1996), Ringland (1998), Bell (2003), Cornish (2005)
• Collections – Fowles (1978), Fahey (1997), Slaughter (2005), Millennium Project Methodology CD and its Global Scenario collection (2003)
• Journals – Futures, Foresight, Technological Forecasting and Social Change, Futures Research Quarterly, Journal of Futures Studies, The Futurist
• Abstract and Citation Indexes – Future Survey, Business Academic Premier, Social Science Citation Index
• World Wide Web

As one might suspect, this approach generated a number of additional methods, many of which were closely related to methods we had already identified. We revised our initial list and posted queries to several listserves that discuss futures topics, including those of our academic program, the Association of Professional Futurists, and the World Futures Studies Federation. We also asked for general advice about our project, and were very pleased to receive a great deal of helpful feedback and, of course, more methods to consider!

These sources yielded dozens of methodological pieces and cases in which a scenario technique was used and/or in which one or more scenarios were produced.

Overviews

Three articles have appeared recently with a similar purpose—to review the field of scenario development and, if possible, bring some organization and understanding to the field. They do an admirable and useful job of proposing different ways to think about scenarios at a high-level. Our purpose here goes a level deeper to provide further assistance by outlining specific methods/techniques that fit within the high-level categories. We summarize below the excellent contribution that each of these overviews has made to the literature, noting areas we will build on. - Van Notten et al., 2003

Philip van Notten (2003) and his colleagues from the International Centre for Integrative Studies in Maastricht have created a typology of “scenario types.” In the end, they propose three major categories or overarching themes, based on the “why” (project goal), the how (process design) and the what (content). They identify 14 specific characteristics to characterize scenarios (Table II.).

Table II. Van Notten Scenario Typology (2003, p. 426)

Their contribution is notable, and it could well be used to study the field of scenario development further. Their attributes, however, relate more to the overall scenario project than to the specific scenario technique(s) used. Process Design contains four attributes that are closer to the techniques employed, but they are general and do not call out the specific techniques. Characteristic VI Data, for instance, classifies scenario designs as either qualitative or quantitative; but that is still very general since there are many ways to conduct qualitative and quantitative scenarios. They have created a comprehensive and useful mechanism for analyzing and comparing scenarios. As valuable as this contribution is, it does not review the actual techniques that futurists use to generate scenarios. - Bradfield et al., 2005

Bradfield (2005) and his colleagues propose “to resolve the confusion over ‘the definitions and methods of scenarios,’” or at least begin to do so. Their approach is historical, tracing the evolution of three schools of scenario development from their origins to the present day. Two of these schools originate in Anglophone countries (US and UK) and one in France.

After describing how Herman Kahn originally introduced the concept of scenario development during his time at RAND, they describe two Anglo schools of scenario development with radically different approaches. The first is the “intuitive logics” school described above as the Shell/GBN method that now dominates scenario development in the US and many other countries. The second is the “probabilistic modified trends” school, originated by Olaf Helmer and Ted Gordon. That “school” is actually an amalgam of two quite different techniques: Trend Impact Analysis that Ted Gordon used at The Futures Group and Cross-Impact Analysis that has been used in many different contexts. Both of these techniques are quantitative, as opposed to the Shell/GBN technique, and they were developed by the same people, but that is pretty much where their similarity ends.

Continental Europe uses a different approach originally developed by Gaston Berger and Bertrand de Jouvenel known as “La Prospective” and now carried on by Michel Godet among others. Godet (2003) has developed a number of useful computer-based tools to analyze structural conditions and stakeholder positions. He also has two tools that generate scenarios—MORPHOL and SMIC PROB-EXPERT. MORPHOL is a computer version of morphological analysis (as described below), and SMIC PROB-EXPERT is a form of cross-impact with some variation.

So Bradfield’s analysis proposes a useful framework for thinking about scenarios at a high level. Notten’s taxonomy proposes attributes of scenarios where Bradfield propose actual high level categories. Their three macro-categories are conceptually useful, but do not do justice to the range of techniques available for scenario development. - Börjeson et al., in press

The final review will be coming out in 2006. Lena Börjeson (Börjeson et. al., in press) and her colleagues from Sweden create a typology of scenario techniques based on Amara’s classification of different types of futures—the probable, possible and preferable futures. Predictive scenarios answer the question: “What will happen?” Exploratory scenarios answer: “What can happen?” Normative scenarios answer: “How can a specific target be reached?” They divide each of these into two sub-categories to make six types of scenarios, as depicted in Figure 1.

Figure 1. Borjeson Scenario Typology (2006, p. 3)

Within their categories, they classify scenario techniques--the focus of our analysis--according to their purpose:

• Generating techniques are techniques for generating and collecting ideas, knowledge and views regarding some part of the future, consisting of common data gathering techniques such as workshops and surveys.
• Integrating techniques integrate parts into wholes using models based on quantitative assessments of probability or relationship, such as time series analysis and systems models.
• Consistency techniques ensure consistency among different forecasts such as morphological analysis and cross-impact analysis.

The latter classification comes closest to serving our purpose here since it identifies some specific scenario techniques although it treats them at a general level that does not allow an analysis of the advantages and disadvantages of each.

In the end, therefore, we still have more work to do, to identify the specific techniques that futurists use to generate scenarios and give some sense of their advantage and their use.

Scenario techniques

Now onto the key purpose of this article—the categorization and discussion of scenario techniques. While authors, such as the ones above, have characterized techniques according to some high-level attributes, none has actually classified the actual techniques in use. That is the purpose of this section. Based on our review of the literature, we have discovered eight general categories (types) of scenario techniques with two to three variations for each type, resulting in more than two dozen techniques overall. There are, of course, variations of the variations. Some techniques are also hard to classify because they contain processes from different categories. Despite these difficulties, we believe that having such a list is a good step toward alleviating the confusion over scenario techniques.

The rest of this section describes each of these categories and the specific techniques in it, noting how each one varies from the pure type.

1. Judgment (genius forecasting, visualization, role playing, Coates & Jarratt)

Judgmental techniques are the easiest to describe and probably the most common since what most people, even professional futurists, generally assert what they believe the future will or could be without much if any methodological support. As the name implies, judgmental techniques rely primarily on the judgment of the individual or group describing the future. While they may use information, analogy and reasoning in supporting their claim, pure judgmental techniques have none of the methodological scaffolding that appears in the other categories. Unaided judgment is probably used most often, but judgment aided with some technique also appears.

a. Genius forecasting come from Herman Kahn, the original scenarist, is also the archetypical genius forecaster. Blessed with high intelligence, an assertive personality and the research capabilities of the RAND Corporation, Kahn (1962) was the first person to encourage people to “think the unthinkable,” first about the consequences of nuclear war and then about every manner of future condition.

b. Visualization is the use of relaxation and meditative techniques to quiet the analytical mind and allow more intuitive images of the future to surface. Individuals typically use a calming narrative, called an induction, to promote relaxation and gently direct the mind to different aspects of the future. Oliver Markley promoted such techniques, first with Willis Harman at SRI in the 1970s and then by teaching and practicing the technique for 20 years at the University of Houston-Clear Lake. (See Markley 1988).

c. Role playing is a form of group judgment. It puts a group of people into a future situation and asks them to act the same as those in that situation would. The original role-playing scenarios were the war games conducted by the U.S. and (probably) the Soviet militaries in the 1950s, simulating the tensions and negotiations leading to a nuclear attack. Today role playing is common in emergency preparedness and for those preparing for dangerous technical missions, such as pilots, astronauts or nuclear operators. (See Jarva 2000)

d. Coates & Jarratt shared the scenario technique that they used in their highly successful consulting practice. It contains elements of more formal techniques described below, but it is basically a more complex, but straightforward form of judgmental forecast. Briefly, the steps involve 1) identifying the domain and the time frame, 2) identifying conditions or variables of concern in that domain, 3) generating four to six scenario themes “that illustrate the most significant kinds of potential future developments,” 4) estimate the value of the condition or variable under each theme, and finally 5) write the scenario. (See Coates 2000).

2. Baseline/expected (Trend extrapolation, Manoa, Systems Scenarios, Trend Impact Analysis)

The second category produces one and only one scenario, the expected or baseline future. We call this scenario the baseline because is the foundation of all the alternative scenarios. Futurists often discount the expected future because it rarely occurs in its full form. In fact, they make their living pointing out that surprising developments are common and are, in fact, more likely than the expected. Herman Kahn reportedly captured this principle in his often-quote phrase, “The most likely future isn’t.”

Nevertheless, the expected future is a plausible future state, and so the description of this state qualifies as a scenario. In fact, it is the most plausible scenario of all because, even though surprises will surely change the future in some ways, it will not change it in all ways. In fact, one of the most surprising developments to futurists, steeped in change and uncertainty, is that things do not often change as fast or as surprisingly as they anticipate. One who takes stock of the world today must admit that it is more like the world of the 1950s than futurists expected, despite the appearance of nuclear power, spaceflight, cell phones and the Internet.

The modal technique in this category is simply to measure existing trends and extrapolate their effects into the future. One can do this by judgment or, if empirical data is available, by mathematical techniques. Next to pure judgment, trend extrapolation is the most common scenario technique--more people, more cars, more computers, more wealth, more liberties, etc. In fact, Herman Kahn (1979) made the rather outlandish claim that he had identified the 15 trends that he believed drove most of human history. His multifold trends included such undeniable trends as the accumulation of scientific and technical knowledge, the greater military capability of developed nations and the growing dominance of Western culture throughout the world. Though surprises are perhaps inevitable, most trends will describe most of the future into the medium- or even the long-term.

We have identified two variations on trend extrapolation, one that elaborates the baseline scenario using futures techniques and one that adjusts it given the occurrence of potential future events:

a. The Manoa technique was invented by Wendy Schultz and other students at the University of Hawaii at Manoa studying with Jim Dator. It is a concatenation of futures techniques to explore the implications and interconnections among trends. The technique requires an individual or group to work with three strong, nearly indisputable trends. Those trends are elaborated in two ways. The first way is to discover the implications of each of the trends separately using a futures wheel. (A futures wheel is essentially a mind-map where each trend forms the center and successive levels of implications are brainstormed from that.) The second way is to discover the interactions among the three trends using a qualitative cross-impact matrix. (A cross-matrix is a square matrix, in this case with one row and column for each trend. The cells are filled with the impacts or effects of one trend (the row) on another (the column).) After these exercises, individuals are left with a rich store of material from which they can answer specific questions about this future or even write a complete scenario. Schultz used this technique with the Hawaii Services Council in 1993 (See Schultz 1993).

Two of Dr. Schultz’s students Sandra Burchsted and Christian Crews (2003) also developed a variation of the Manoa technique that they call Systemic Scenarios. Rather than use the cross-impact matrix as a way to identify the interactions among the trends, they show the relationships among the implications from different trends using a causal model which shows the dynamic interactions among the implications and hence the trends (See Burchsted 2003).

3. Elaboration of fixed scenarios (Incasting, SRI)

The third category begins the explicit consideration of multiple scenarios. Most scenario techniques develop the scenarios from scratch, but these begin with scenarios that are decided ahead of time. The intention then is to elaborate the scenario logic or kernel, the simplest statement of what the scenario is about. The advantage is that participants do not have to struggle with the uncertainties of the future. All they have to do is articulate the implications of given alternative futures.

a. Incasting is a simple matter of having participants divide into small groups and read a paragraph that describes a rather extreme version of an alternative future. Examples would be a green future, a high-tech one, or one dominated by multi-national corporations. They are then asked to describe the impacts on a series of domains, such as law, politics, family life, entertainment, education, work, etc. One interesting variation during the debrief is not to tell the other participants the nature of the underlying scenario, but rather have them guess what is from its effects. Incasting is a good technique to illustrate how the world could be different given paths that the world could take. (See Schultz n.d.)

b. The SRI Matrix was one of the first explicit scenario techniques following Kahn’s introduction of genius forecasting and trend extrapolation. It was developed at the Stanford Research Institute (now SRI) and used by Peter Schwartz, Jay Ogilvy, and Paul Hawken in their late 70s book Seven Tomorrows. The SRI technique also begins with a fixed number of scenarios, usually four, but they are not expressed as paragraphs. The scenarios are identified as titles to columns in a matrix, such as the expected future, the worst case, the best case, and a highly different alternative. The titles vary by practitioner and by engagement. The dimensions of the world are then listed in the rows, such as population, environment, technology, etc., or other domains that are more specific to the engagement. Participants then fill in the cells with the state of that domain in that scenario. The whole scenario is elaborated in each column, and the differences for a specific domain across the scenarios are elaborated in each row. (See Hawken 1982)

4. Event sequences (Probability Trees, Sociovision, Divergence Mapping)

Most people think of the past as a series of events, in one’s life or in history. So we can think of the future that way too, except that we do not know which events will occur and which ones will not. Each event then has a probability of occurrence. If a potential event happens, the future goes one way; if not, then another. The future branches at each of those points depending on whether the event occurs or does not. In fact, more than one thing can happen in which case the future has three paths from that point. String a number of those branches together, and one has a probability tree. Two variations of probability trees were discovered: one uses the branches to create scenario themes and the other builds the sequences after developing the events (Lisewski, 2002; Buckley, 1999; Covaliu, 1995).

a. Probability Tree has the same form as a decision tree, except the branches in a decision tree are not what could happen, but what decisions we will make at each branch. The tree ends at different future conditions depending on the path. And if one knows the probability of each branch, one can calculate the probability of arriving at that final state as the product of the probabilities of the branches that occurred along the way. Those probabilities sum to 100% since one of them is bound to occur. Probability trees are used in risk management, particularly when risk managers and planners have to assess the probability of multiple risks happening in the same time frame. The Eidos tool suite from the Parmenides Foundation (formerly ThinkTools) contains a tool for building and evaluating probability trees.

b. Sociovision begins with a standard probability tree. Examining the tree, however, may reveal certain branches that have a common character. Perhaps many of them are less likely or more preferred, or they may be driven by one particular stakeholder or condition. Gathering those branches together creates a coherent scenario of how the future might develop, complete with the events that make up the story. The probability tree then acts as an input that reveals some overall macro themes that might not be apparent to the participants at the beginning. (See de Vries 2001)

c. Divergence mapping was described by Willis Harman in his book An Incomplete Guide to the Future. It consists of brainstorming a set of events that could change the future. His “map” allows for up to 22 of those events, but more are clearly possible. These events are arrayed in a fan-life structure with four arcs, each of which represents a longer time horizon. Events from earlier time horizons are then linked with later ones in a plausible sequence that forms the storyline of a scenario. (See Harman 1976)

5. Backcasting (Horizon Mission Methodology, Impact of Future Technologies, Future Mapping)

Most people think of the future as extending from the present, a natural extension of the timeline running from the past and through the present. But that perspective has its disadvantages, chief among which is the future then carries all the “baggage” of the past and the present with it into the future. The baggage limits creativity and might create futures that are too safe, not as bold as the actual future turns out to be.

An antidote to carrying too much baggage is to leap out into the future, jab a stake in the ground, and then work backward on how we might get there. The first step then is to envision a future state at the time horizon. It can be plausible or fantastical, preferred or catastrophic; but having established that state as a beachhead, it is easier to “connect the dots” from the present to the future (or back again) than it is to imagine the events leading to an unknown future. The technique is “backcasting,” (Robinson, 1990) as opposed to “forecasting,” for obvious reasons.

a. Horizon Mission Methodology -- One of the most well-known and purest forms of backcasting was developed by the late John Anderson at the National Aeronautics and Space Administration (NASA). Anderson’s technique was designed to help NASA engineers decide on R&D pathways that might yield some return. Forecasting from the present, engineers were often bound by their disciplinary backgrounds to recommend incremental rather than breakthrough research. To counteract that tendency, Anderson first had engineers envision a fantastical mission (a horizon mission), one that was completely infeasible given today’s technology. A favorite of his was a one-day mission to Jupiter. That trip today would take several months by the fastest route using the most powerful rockets. So a one day trip was fantastical indeed. Having overcome the “giggle factor,” Anderson then asked the engineers to “decompose” that mission into its component parts. In other words, “Supposing that such a mission had actually taken place, what technologies would be required?” Given the components of the mission, he then asked them to decompose each of those components using the same question, “What technologies would it require?” Arriving at the present, engineers found that they had some near-term R&D opportunities that might not get them to Jupiter in a day, but they might create other breakthroughs in space exploration. Working backward got them out of the present and into the future in a big way! (See Hojer 1999)

b. Impact of Future Technologies – The IBM Corporation has developed and is now marketing a backcasting technique for the same purpose—making investment decisions in future R&D technology. The technique, called the Impact of Future Technologies (IoFT), begins at the same place that Anderson’s does with a highly capable vision of the future. IoFT differs from HMM, however, in starting from multiple elaborated scenarios of the future rather than just one simple mission. Working backward from those scenarios, a team of knowledgeable scientists identifies signposts that are defined as scientific or technological breakthroughs that would be required for one or more of the scenarios to come true. IBM does not recommend that the client work to create the breakthroughs because they are so massive that even the most capable client would contribute little to their occurrence. What is more, breakthroughs are by definition unpredictable, particularly when they will occur, so that they recommend rather that the client monitor for the occurrence of the breakthrough and then deploy a contingent strategy during a subsequent window of opportunity for exploiting the capabilities of the breakthrough. (See Strong 2006)

c. Future Mapping was developed by David Mason of Northeast Consulting. It is a variant of the pre-defined scenario technique in which he pre-defined, not only the end-states, but also the events leading up to them. Participant teams then select and arrange the events that lead to each end-state. The technique offers participants a deeper understanding of how events can interact to create different futures and how different end-states can occur from the same set of events. (See Mason 2003)

6. Dimensions of uncertainty (Morphological analysis, Field Anomaly Relaxation, GBN, MORPHOL, OS/SE)

The reason for using scenarios in the first place is the uncertainty inherent in predictive forecasting. We never have all the information; theories of human behavior are never as good as theories of physical phenomena, and finally we have to deal with systems in chaos and/or emergent states that are inherently unpredictable. Scenarios in this section, then, are constructed by first identifying specific sources of uncertainty and using those as the basis for alternative futures, depending on how the uncertainties play out.

a. GBN Matrix has become the default scenario technique since Peter Schwartz (1991) published his best-seller, The Art of the Long View. The matrix is based on two dimensions of uncertainty or polarities. The four cells represent alternatively the four combinations of the poles of the two uncertainties, each of which contains a kernel or logic of a plausible future. Each kernel is then elaborated into a complete story or other presentation, and the implications for the focal issue or decision are discussed.

b. Morphological Analysis (MA), Field Anomaly Relaxation (FAR) are more traditional versions of the same technique. The difference is that they contain any number of uncertainties and any number of alternative states for each uncertainty so that GBN is a actually subset of MA/FAR. The uncertainties are portrayed as a set of columns in which each column represents a dimension of uncertainty and contains any number of alternatives. One creates a scenario kernel/logic from the MA/FAR layout by picking one alternative from each column. Of course, that is easier said than done because a standard layout with five dimensions of uncertainty, each with three alternatives, generates 3⁵ or almost 250 different scenario kernels. While MA and FAR are more complicated and hence less common, they do overcome the difficulty that it is devilishly hard to capture the uncertainties of the future in just two dimensions. (See Coyle, 2003, 1994; Duczynski, 2000; Eriksson 2002; Rhyne 1974,1981, & 1995)

c. Option Development and Option Evaluation is part of the Eidos tool set distributed by the Parmenides Foundation that manages the complexity of morphological analysis. Option Development is the program that lays out the dimensions of uncertainty and the alternatives associated with each one. Open Evaluation uses a compatibility matrix of all the alternatives against all the other alternatives to calculate the consistency of each combination of alternatives. The program then ranks them according to their consistency.

d. MORPHOL is a computer program that also manages the complexity of morphological analysis. Developed by Michel Godet, a prominent futurist in Europe, MORPHOL performs the standard morphological analysis, but it then reduces the total number of combinations based on user-defined exclusions (impossible combinations) and preferences (more likely combinations). It also provides an indicator of the probability of each scenario compared to the mean probability of all scenario sets based on the user-defined joint probability of each of the alternatives in the set. (See Godet 1996)

7. Cross-impact analysis (SMIC PROF-EXPERT, IFS)

One objective of identifying various future conditions, events and even whole scenarios is not just to identify their characteristics and implications, but actually to calculate their relative probabilities of occurrence. One can judge the single probability of a condition or an event using judgmental means. But the more people making the judgment and the more expert they are, presumably the better their collective judgment.

Most analysts, however, are keenly aware that the probability of any one event is, to some extent, contingent on the occurrence of other events. Placing these events in a square matrix with each condition or event occupying one row and one column, one can display, not only the initial probability assigned to a condition or event, but also the conditional probabilities of the condition or event given the occurrence of any other condition or event. Using these estimates, a random number between 0 and 1 is chosen. Events with a probability above that number are said to occur; those below are not. The probabilities of all events then adjusted (up or down) based on the contingent probabilities in the matrix. Running the matrix many times produces a distribution of probabilities for each that can be used to estimate the probability of that event given the possible occurrence of the other events.

The most well-known use of cross-impact analysis was a program conducted called INTERAX, conducted by Selwyn Enzer (1981) at the University of Southern California. Enzer constructed a cross-impact matrix of many global trends and potential events that participants would discuss at an annual workshop.

a. SMIC-PROB-EXPERT is a cross-impact analysis developed by Michel Godet with an important variation. The cross-impact matrix of conditional probabilities is constructed by experts, but their estimates often do not conform to the laws of probability, such as P(x) must equal P(x|y) · P(y) + P(x|~y) · P(~y). SMIC adjusts the probabilities suggested by the experts so they conform to such laws. The PROB-EXPERT portion of this technique creates a hierarchical rank of scenarios based on their probability. Finally, it allows one to draw diagrams of clusters of scenarios and experts, showing which scenarios are most alike, which experts judged the probabilities most alike and even which scenarios are most favored by which experts. (See Godet 2003)

b. Interactive Future Simulation was developed at the Battelle Memorial Institute to calculate the quantitative conditions associated with different scenarios. IFS begins with a set of variables, called Descriptors, that are important for understanding the future rather than with events or binary conditions as the other techniques do. It divides the range of each variable into three alternatives—high, medium and low—and assigns an initial probability to each of those alternatives. It then constructs a cross-impact matrix in which the cells are the influence of each alternative on each other alternative on a scale from -2 to +2. A Monte Carlo simulation runs the impacts many times over generating different combinations of scenarios with different frequencies of occurrence. The final probability of each of the alternatives (the ranges of the target variables) is then calculated based on the number of times that that alternative appears in the scenario combinations generated. (See www.battelle.org)

8. Modeling (Trend Impact Analysis, Sensitivity Analysis, Dynamic Scenarios)

Systems models are used primarily for baseline forecasting—i.e., predicting the expected future. Based on equations that relate the effects of some variables on others, the output is usually the expected value of target variables at the time horizon or graphs that show the change of those variables between the present and the time horizon. But any technique that can generate a single-valued prediction of the future can also produce scenarios by varying the inputs and/or the structure of the models that generate the prediction.

a. Trend Impact Analysis (TIA) is a method for adjusting the baseline trend given the occurrence of a potential future event. The TIA was invented by Ted Gordon at The Futures Group. It involves a trend and a potential event that acts to perturb the original trend trajectory. Three different points of impact are identified and estimated—time to first noticeable impact (when the trend first departs from its original trajectory), time to maximum impact (when the trend is farthest from its original values), and time to steady-state or constant impact (when the effect of the event is fully integrated into the trajectory of the trend). The size of the maximum and steady-state impacts are also estimated. A new trend line is then calculated (an alternative scenario) and compared to the original baseline trajectory. TIA has been used by the Federal Aviation Administration, Federal Bureau of Investigation, Joint Chiefs of Staff, National Science Foundation, Department of Energy, Department of Transportation, the State of California, and other U.S. agencies. (See Gordon 2003)

b. Sensitivity analysis varies one of the three parts of a systems model that can be varied:

i. The value of exogenous variables that drive model. Exogenous variables, also called boundary conditions, influence other variables in the model, but they are not themselves influenced by those variables. In other words, they are set outside the model, in the model’s environment. The interest rate set by the Federal Reserve and the tax rate set by the Congress are typical exogenous variables to the models of the U.S. economy. One can vary each or both to see how they affect output variables like GDP or employment. The analysis then measures how “sensitive” the model is to changes in the boundary conditions. Each of those variations is a scenario.

ii. The parameters that define the effect of variables on each other. The equations in the models that define future values of dependent variables (Y) are constructed from independent variables (X) adjusted by a coefficients (b) in the form, Y = a + bX. The value of the coefficient is based on the historical relationship of X and Y. But there is considerable uncertainty, even for the most well-supported coefficients. What is more, the value of the coefficient can change completely if the historical relation between X and Y changes. So one can vary parameters in the model to define different scenarios.

iii. The variables in the model itself. Models consist of variables that represent the real world, but the choice of variables to include in the model is also a matter of some dispute. One can vary the actual structure of the model and its equations by adding or removing variables to see the effects on the output variables. So each of these changes can produce alternative descriptions of the future—i.e., scenarios. (See Saltelli 2004)

c. Dynamic scenarios are a combination of scenario development and systems analysis, in that order. The first step is the ordinary process of generating scenario themes or kernels by clustering events of a similar type from a brainstormed universe of all plausible future events. Each of those themes then defined a system which is mapped using causal models. The variables that appeared in many different models were brought together in a meta-model that purported to map the whole domain. The individual themes were then elaborated using different values for the uncertainties in those models.(See Ward 2003; Schriefer & Sales 2006)

Observations and evaluation

Having described the techniques individually, this section compares the scenario techniques with each other. The first table compares the starting point, process, and products of the different scenario techniques.

• The starting points range from completely open to beginning with draft scenario logics. The open approaches begin with an environmental scanning process to produce the raw material that will be crafted into scenario logics. The other extreme is to begin with scenario logics and either elaborate or customize them in order to explore their implications. In between are techniques that begin either with the dominant driving trends or with key uncertainties.

• The processes summarize how the methods are actually carried out. As expected, here is where we see the greatest distinctions among techniques. It is what separates one from another.

• Products also vary by technique. Most techniques produce different numbers of scenarios. Most produce kernels or logics; others produce probabilities of different alternative conditions, and still others produce elaborated stories or end-state descriptions. The common approach of about twenty years ago of producing best case, worst case, and a middle version is no longer used today. The problem with this approach is that clients almost always selected the middle ground, and were thus losing out on the value of expanding their thinking to consider a broader range of futures possibilities.

As described in the confusions above, however, some of the products are scenarios, but not stories about the future in the narrow sense. These range from probabilities of end states to adjusted trend values to ideas for investment strategies.

Table III. Comparing Starting Points, Process and Products of the Scenario Techniques