PhD - Reducing Uncertainty

[this is the first part pf my PhD 'Synthesis' which - after five years of argument - brought together my published work and justified the award of my doctorate]

submitted in partial fulfilment of the requirements of Leeds Metropolitan University for the degree of Doctor of Philosophy

Addressing the lack of standardised approaches to scenario forecasting, and especially of quantified performance reports on the techniques to be recommended, this synthesis describes a body of research and development work over the past decade. The initial aim was to produce a basic, standardised approach to scenario forecasting which might be used by a wider range of organisations. It - qualitatively and quantitatively - explored how such scenario methods are used; and then experimentally developed optimal approaches. The results exposed a gap in strategic planning and new theory for this was, accordingly, developed.

To extend the work further, to global scenarios, a new Hypothesis of Aggregated Expectations was then developed. This new theory underpins scenario forecasting in general, sheds light on why this may be successful in detecting trends in society, and allows more definitive maps of future trends to be produced.

In this submission, therefore, 13 refereed papers - along with extracts from 2 books - describe significant new extensions to the theory and techniques which underpin long-range planning. These are included after the end of the synthesis. Seventeen separate pieces of research were undertaken; eight largely making use of existing methods and nine based on new techniques and, in particular, extending the use of focus groups. Conducted over the ten years up to 1999, using a range of different techniques across different populations, they included direct input from nearly 4,000 respondents. The research methodologies used are explored in detail in the Appendix.

The end result was a standardised and comprehensively tested set of techniques, supported by new theory.

This synthesis, in which the main elements of long-range planning are illustrated by the 'flowcharts' at the beginning of each section, describes and integrates, as well as critically reviewing, these developments.

LIST OF INCLUDED PUBLICATIONS

Paper 1: A Two Decade Test of Product Life Cycle Theory, British Journal of Management, 4, pp 269-274, (1993)

Paper 2: Management's Commitment to Marketing Theory Compared with Actual Practice, Competitive (refereed) Paper to MEG (Marketing Educators Group) Conference, (1996)

Paper 3: Scenarios Made Easy, Long Range Planning, 28 (4), pp81-86, (1995)

Paper 5: (extract from) Marketing Strategy: the Challenge of the External Environment, Sage, Chapter 7, (1998)

Paper 6: Industry Scenarios - Short Termism Revealed, Industrial Management and Data Systems, 96 (8), pp 23-27, (1996)

Paper 7: Long Range Marketing, Journal of Marketing Practice, 4 (6), pp 178-184, (1998)

Paper 8: (extract from) Marketing Strategy: the Challenge of the External Environment, Sage, Chapter 11, (1998)

Paper 9: Robust Strategies in a Day, Management Decision, 35 (3), pp 219-223, (1997)

Paper 10: From Scenarios to Robust Strategies: The Links Managers Make, Futures Research Quarterly, 17 (2), (2001)

Paper 13: A General Hypothesis of Aggregated Expectations, Technological Forecasting and Social Change, 55, pp 145-154, (1997)

Paper 15: Organisational Futures: Unprepared for the Surprises to Come, Management Decision, 37 (5), pp 411-416, (1999)

Paper 16: Life-Long-Learning is the Future, Foresight, 1 (5), pp 125-130, (1999)

INTRODUCTION

This synthesis, therefore, describes a body of research and development work which - over more than a decade - addressed these shortcomings. Its initial aim was to produce a basic, standardised approach to scenario forecasting which might be used by a wider range of organisations. This work qualitatively and quantitatively explored how such scenario methods are used; and then experimentally developed optimal approaches. The results, quantified by further research, exposed a gap in strategic planning, where corporate strategy did not easily allow for the incorporation of long-range forecasts; thus demanding the development of new theory.

In order to extend the work to global scenarios, the new Hypothesis of Aggregated Expectations - described in detail in the later section - was developed. This theory also underpins scenario forecasting in general; and sheds significant light on why this may be successful in detecting trends in society. The new approaches then developed allow more definitive maps of future trends to be produced.

The end result was a standardised and comprehensively tested set of techniques, supported by new theory, which covers the long-range planning needs of almost all medium to large organisations.

This synthesis describes and integrates, as well as critically reviewing, these developments.

The post-1945 boom in business theory, up to the start of the 1970s, led to a focus on quantitative forecasting as a major component of planning The main principle then adopted was that existing trends could be extrapolated[3].

In the relatively stable business climate of that boom period, the techniques for investigating historical trends were mainly numerical, and largely based on time-series analysis[4]. The seemingly mathematical accuracy of the techniques was seductive and, in the relatively stable decade of the 1960s, it often appeared that they worked well. With the advent of greater computing power, this approach was eventually extended to the production of econometric models (Levenbach & Schultz, 1987) - based on the solution of large numbers of simultaneous equations[5].

Qualitative Forecasting

The 1970s brought an end to this stability[6]. Where unpredictable shifts in the environment were then seen to be possible, the basis for extrapolation of existing trends over the longer term was undermined.

The resulting need to understand long-range developments in the external environment initially inspired moves to develop new approaches to environmental analysis. The conventional steps in this process of addressing the wider environment were[7] a steady progression (and concentration) from information obtained by a general audit of the wider environment to identification of the key strategy needed to deal with the opportunities and threats emerging from analysis of this information[8]. The heart of this process, starting with the audit, has been called scanning. This is a wide-ranging activity which is an essential precursor for all long-range planning.

Qualitative Techniques

Futurology… following their initial funding by the military after the Second World War, the 1960s and 1970s were the decades which proved to be the high point of the work of the individual futurologist, including those at Rand and the Stanford Research Institute (SRI) as these expanded into commercial work. In particular this approach was publicized by Herman Kahn (1967), of Rand Corporation and then the Hudson Institute, and Naisbitt (1982). By the use of their imagination, such futurologists attempted - often almost as a form of science fiction - to see what the possible range of alternative futures might be; so that suitable contingency plans could be developed. It made a virtue of the expertise of the individual gurus[9].

Judgmental Forecasting

In practice, even now most long-range forecasts still emerge from the personal judgement of an individual, although this typically happens informally - and unrecognized - as part of the decision-making process undertaken by individual managers.

Apart from demographic projections (Wallace, 1999) which can be at least in part predictable, most long-range forecasts - influenced by unpredictable factors which mathematical techniques cannot handle - now are essentially based on the personal judgement of individuals (Diffenbach, 1983). The various techniques available, therefore, focus at one extreme providing the best possible historical data and in helping the individual use these to visualize the future and, at the other, in averaging out individual misjudgements (Makridakis & Wheelright, 1989).

Jury Method…thus the simplest method of 'averaging' such judgements, to avoid individual bias, is to use a jury; a number of individuals - typically experts - meeting together. Such juries can, however, be swayed by strong-minded individuals; especially if they make use of their status within an organisation (Makridakis & Wheelright, 1989).

Delphi Method…assessing the importance of identified events which may occur in the future, this was developed in the late 1960s by Olaf Helmer and Ted Gordon[10]. Its purpose was to overcome the distortions that can emerge when simply acquiring the views of experts[11].

In our work with our US collaborators (Glenn & Gordon 1998), we have found that this - well researched - process can usefully complement the widening perspective provided by scenario forecasts; and we have successfully used it as one input to our own global scenarios.

On the other hand, the Delphi method has been criticized as time consuming and difficult to organize. The delay in receiving replies from the various rounds can be considerable, participants drop out and the way the questions were put can introduce bias into the responses. In particular, there is pressure for a consensus even though this may not produce convergence on an agreed forecast (Fildes, 1987)[12]^.

Scenarios in Theory

Michele Godet (1982) criticized the bias which tends to lurk behind normal forecasts His solution was to use scenarios (‘La Prospective’) to handle the uncertainty of the future. The technique focuses on the uncertainty itself and thereby avoids most of the problems by not predicting exactly what will happen. It recognizes that there may be many different possible outcomes, and describes them. The ‘scenario[13]’ is, in this context, merely a reporting framework within which all of these outcomes may be described

Possibly as a result of the very sophisticated approaches[14], and of the difficult techniques they employed, scenarios initially earned a reputation for being difficult and costly to use. Even so, the theoretical importance of the use of alternative scenarios, to help address the uncertainty implicit in long-range forecasts, was dramatically underlined by the widespread confusion which followed the Oil Shock of 1973. As a result, many of the larger organisations started to use the technique in one form or another[15].

Practical Development of Scenarios

The practical development of scenario forecasting to guide strategy, rather than for more limited academic uses, was started in 1971, at the Royal Dutch Shell group of companies (hereafter, for simplicity, referred to as ‘Shell’), by Pierre Wack (1985). It too, was given impetus by the Oil Shock two years later. Shell has, since that time, led the commercial world in the use of scenarios and in the development of more practical techniques to support scenarios[16].

In one way or another, Wack's work has formed the basis for much of the scenario work undertaken in practice across recent decades. However, the generalized nature of these scenarios (such that they can incorporate almost any other form of forecast input) means that there are a wide range of different implementations[17].

Where most of the previous forecasting methods were to a degree standardized, and the work of different groups could be compared, scenario forecasting is characterized by the variety of different processes which may be used. Almost every group of practitioners has its own version. Even so, the approach seems to work across a wide variety of organizational situations, and a great strength of scenarios is that they can accept such a range of processes - even where the forecasting processes are uniquely tailored to the needs of the individual client.

The disadvantage is that such work cannot be easily compared between groups and in particular, little or no theory has been developed to underpin the practice of this work. The work reported in this synthesis attempts to provide some standardization and especially to suggest some supporting theory explaining how the processes determine what emerges from the 'black box' of scenario forecasting.

The initial split was between normative scenarios, used by the earlier practitioners and which relate to what are the desired outcomes, and exploratory scenarios, preferred more recently and which only report on likely developments.

They can also be broadly grouped into two main categories of usage: those that 'force' the resultant scenarios and the 'freeform' ones favoured by Pierre Wack.

‘Forced’ Scenarios

Thus, stimulated by Wack's work, but not following the details of this, other approaches - typically those recommended by consultancies - can be based upon creating four-box matrices, derived from the two (or possibly more) dimensions which have emerged during the process of examining the uncertainties; identification of which is at the heart of almost all scenario approaches[18]. The outcome (with one dimension stretching from A to B and the other from C to D, as used by British Airways - Moyer, 1997) will be four scenarios. A+C, C+B, D+B, A+D[19]. [20]

'Freeform Scenarios'

The other - more widely reported developments have been direct continuations of Shell's (Wack's) work with 'brainstorming' (freeform) approaches; and, especially, the simpler form used in its work with its operating companies across the world. Much of this has been undertaken by 'alumni' from the Shell planning group[21]^,[22].

Discussion

My own initial approach was based directly upon the 'Shell' free-form methodology; or at least the simplest version of this which its planning group used with their participating managers coming from outside of the planning department itself. Even so, my work originally had the prime objective of simplifying this further, so that it might be used more widely; by organisations which did not have specialised expertise available.

At the same time I recognised that not merely was their no standardisation, or theoretical underpinning, but that no significant research into the actual use of scenario forecasting had been reported or, to the best of my knowledge, carried out.

Accordingly, another key objective - which emerged as the work progressed - was to undertake the widest possible range of research into the performance of the various techniques involved. The outcome of this research was, in particular, used to incrementally improve the effectiveness of the processes - and to start to develop relevant theory.

[1] Or at least of those of the overall 'categories[1]:

inevitable outcomes

inevitable changes with unknown outcomes

unknown changes

‘wild’ cards

which are susceptible to the techniques involved. The first of those listed is already the province of corporate strategy. The last is, apart from 'creeping catastrophes', largely unpredictable by any means.

Accordingly, the majority of the work described in this synthesis has focused on the development of practical new techniques for long-range planning in the second and third categories (Mercer 2001a-e),.

[2] It is not clear why there has been so little research in this field. However, perhaps due the judgmental emphasis involved, qualitative research in general has been the subject more of anecdotal reports than specific research and scenario work has, in particular, been mainly the province of consultancies who for reasons of commercial confidentiality have traditionally not published their results.

[3] This form of forecasting is still the staple diet of most corporate strategy, which focuses on the relatively short term future - two to three years ahead. As such, the mathematical techniques involved are rigorous and well understood, though Makridakis (1988) disputes their effectiveness, especially for handling uncertainty.

[4] For example exponential smoothing (Gardner, 1987), regression analysis techniques (such as ARMA - Auto Regressive Moving Average) and Box-Jenkins methodology (Makridakis & Wheelright, 1989)

[5] Such models still represent the main tool used for national and international macro-economic forecasts - especially those related to interest rate changes. Reports in the media suggest, however, that they are more often wrong than right - especially in terms of predicting the onset of recessions (Woodall, 2002) - as might be expected of such trend-based models in an uncertain environment. On the other hand, the greatest advantage of such models in this context may be the 'what if' ability, where it is possible to model the changes which might result from specific financial actions, but they also can be used to communicate 'government expectations' - to the wider audience - in the hope that they become self-fulfilling!

[6] Most notably with the ending of the Bretton Woods agreement on currency stability, which led to the emergence of the ultimately unmanageable global financial markets, closely followed by the oil shock of 1973.

[7] As illustrated by Johnson and Scholes (1988), who described them in the context of corporate strategy.

[8] In the context of the typology described by Aguilar (1967), which also sees research about the external environment progressing through ever narrowing perspectives from 'Undirected Viewing' (scanning) down to 'Formal Search' (market research), the former is the widest-ranging activity. Naisbitt (1982) describes the one scientific approach - 'content analysis' - of tracking newspaper coverage; but the resources needed for this are usually only available to national intelligence agencies. For most organisations scanning just encompasses the activities which the organisation uses, formally and informally, to keep abreast of changes in the wider environment which will affect its future; such as simply following the daily news media.

[9] In fact, key elements from the forecasts of the leading gurus , especially those of Herman Kahn (1967) himself and later of the Tofflers (1970), did stand the test of time; so, assuming a suitably informed guru can be found (a significant shortcoming in practice), this approach may still work.

In the wrong hands, though, it was much less successful. As Ringland (1998) reports (based on Schnaars, 1989), of the 355 predictions made by 27 top TRW scientists in 1966, "…nearly every prediction was wrong".

[10] Then of the American RAND Corporation but now with our collaborators - the American Committee of the United Nations University (ACUNU)

[11] In the original Delphi concept, the participants - the experts - remain anonymous, their identity known only to the Delphi study organizer. During the exercise, the participants respond to a questionnaire in which specific events are identified and the expert is asked to make a judgement of the probability of the event occurring before certain dates

The results of the first round of responses are statistically analyzed and fed back to all participants. If they desire, they can then decide to modify their first response (Holroyd, 1998). In this way, a wide range of opinion is captured in the first set of responses and amended in a second, or even a third, round of questions. It is found that such an approach often leads to a narrowing of the range of responses - which can usefully complement the widening which results from the use of scenarios - and that the final statistical analysis of the results can be a good guide to the eventual future outcome (Twiss, 1980), especially where the possible alternatives are clearly specified.

[12] In order to avoid this problem, Gordon recommended a follow-up face-to-face session by the organizer(s) with the participants - a practice he still follows - and delays can now be reduced by the use of web-based communications.

[13] In the later context of its commercial use, Ringland (1998) nicely defines scenarios by "The emphasis in creating a set of stories of the future…for testing business plans…", though she also stresses - as we have done with our own students - "The essence of this is to find ways of changing mindsets so that managers can anticipate futures and prepare for them." Thus, although there are a wide range of different techniques which may be used, the ultimate outcome is usually the production of such stories.

[14] The earliest use of scenarios tended to be the province of academics, before it spread to the leading think-tanks. Thus, though the concept was first introduced as ‘La Prospective’ by Berger (1964), and the word ‘scenario’ itself was reportedly first used by Herman Kahn(1967), the theoretical foundations of scenario forecasting were mainly developed in the 1970s, especially by Godet (1987) himself. By the early 1980s these approaches had developed into a sophisticated forecasting technique which was primarily recommended for the analysis of complex systems by integrating the output from other qualitative approaches using techniques such as cross-impact analysis (Godet, 1979).

[15] Indeed, just ten years later, Diffenhach (1983) reported that ‘alternate scenarios’ were the third most popular technique for long-range forecasting - used by 68% of the large organisations in his 1977 survey of 66 firms from the Fortune 500.

Percentage of companies reporting

use of techniques

Expert opinion 86

Trend extrapolation 83

Alternate scenarios 68

Single scenarios 55

Simulation models 55

Brainstorming 45

Causal models 32

Delphi projections 29

Cross-impact analysis 27

Input-output analysis 26

Exponential forecasting 21

Signal monitoring 12

Relevance trees 6

Morphological analysis 5

Since that time, however, the numbers of commercial organisations employing scenarios has fallen dramatically, though some governments still do make extensive use of them.

[16] SRI (the Stanford Research Institute think-tank) has also followed very much the same path (Kleiner, 1966), though its work has been less influential.

[17] Such as those - based on the use of two dimensional (4-way) boxes - used by British Airways (Moyer, 1997) or - based on interviews with an organisation's senior executives - preferred by North East Consulting (Herman, 1997) or that - based on use of its own expert consultants - implemented (Schwarz, 1991) by the Global Business Network (GBN).

[18] These can also be derived from Repertory Grid Research (Sampson, 1986).

[19] This type of approach can become very sophisticated. It is reported (Millett, 1992)that Battelle Institute assign three levels to each factor, derived from a number of Nominal Group sessions (Langford, 1994), before using a computerized cross-impact matrix program (BASICS) to resolve these. Michel Godet (1982) adopts a similar approach with his own MICMAC cross-impact analysis program.

At a simpler level, in the mid 1990s British Airways (Moyer, 1997)used the two dimensions of ‘growth’ (a crucial feature of their market) and ‘governance’ (the regulation/ deregulation which is just as important to their activities around the world) as the basis for their scenarios, though they then focused on just two outcomes - 'Wild Gardens’ (with growth as the main focus) and ‘New Structures’ (where governance/regulation predominated). A further variation is to look at the outcomes in terms of 'risk’ and ‘reward’ to the organisation. In the event, this looks very much like the well-known Ansoff (1957) Matrix, which was typical of the techniques which had emerged in the 1960s.

[20] Although these approaches follow many of the principles of scenario forecasting in general, they are very dependent on the level of expertise needed to select the original dimensions. As such, the consultancies who typically employ them usually market themselves on the basis on their specialized (guru) expertise in this field. We chose not to follow such 'forcing' approaches since the necessary expertise is not generally available in most organisations and, in view of the wide variety of starting points employed, they cannot easily be standardized or supported by theory.

[21] Thus, the leading consultancy, GBN (Global Business Network), was started by Peter Schwarz (1991), the leading author on the subject but also a Shell alumnus as were a number of its consultants. This also true of some now working in academic institutions. Thus, Kees van der Heijden (1997), isone of a number of managers from Shell who have subsequently entered the academic world to offer their experience to a wider audience. He describes the benefits to be obtained from using scenarios in terms of their value for the institutional learning process.

[22] Although all of these 'freeform' approaches, based loosely on different approaches to brainstorming, are now used in practice and have indeed proved to be practical for use by the larger multinationals, they still involve the use of sophisticated techniques. Accordingly, all of them demand teams of highly trained experts; either in house, as at Shell and ICL (Ringland, 1998), or in consultancies, as at SRI, GBN and Battelle.