Share with your friends


Analytics Magazine

Operations Research: Taming Hard Problems With Soft O.R.

Spring 2009
Soft” methodologies tackle messy problems that traditional operations research/management science can’t touch, so why isn’t it promoted in the U.S.?

“Soft” methodologies tackle messy problems that traditional operations research/management science can’t touch, so why isn’t it promoted in the U.S.?

By John Mingers

Over the last 40 years, methodologies have been developed to deal with “wicked problems” or “messes” that are beyond the reach of the traditional, mathematical modeling methods of operations research (O.R.). These methodologies are structured and rigorous but non-mathematical. Examples include Soft Systems Methodology (SSM), cognitive mapping/SODA and the Strategic Choice Approach (SCA). Collectively they are known as Soft O.R., Soft Systems or Problem Structuring Methods (PSMs), and they have been accepted as an important part of O.R. almost everywhere with one notable exception the United States.

The following three problems illustrate the kind of difficult situations that Soft O.R. was developed to deal with and make the point that Soft O.R. is more effective in certain situations than traditional O.R. One involves children’s healthcare in the United Kingdom, the second a polluted river system in India and the third risk management for Europe’s largest carnival. They have been chosen from many because they are typical of messy situations that have been tackled using Soft O.R. yet are very different from each other, and because they have been written up in reputable journals or books.

In 1997 the Salford and Trafford Health Authority in Manchester (U.K.) wanted to develop a more integrated approach to its children’s services. Consultants produced several documents and one of the organizations involved, Salford Community Trust, became the coordinator. A subsequent project considered the further shape of the Salford service, building on the work of the existing working group. The goal was to produce a service specification that “operationalizes” the proposals made by the Health Authority [1].

While this may sound straightforward, on investigation it turned out to be complex and messy [2]: there was no agreed definition of what a “service specification” was; there was no agreement about what services were to be included within the scope of the project; requirements on a whole range of issues were ambiguous; several different agencies were involved with children’s welfare, and it was known that they disagreed about the future direction of children’s services as well as how the project should be tackled; and there were significant political aspects of the situation at both local and national levels.

Moving to India, the Cooum River in Chennai is slow-moving and polluted with debris, organic sludge and raw sewage [3]. This is a long-standing problem that involves several government agencies including: the Public Works Department, Chennai Metropolitan Development Authority, the Slum Clearance Board and Metrowater. Various attempts have been made to improve the situation, but these have generally been piecemeal engineering projects that have worked locally and in the short-term but failed in the long-term. The situation is complex both in terms of the physical environment (drought/monsoon, flat topology, sand bars, tidal action) and the social environment (population growth, poverty, institutional culture, jurisdictional conflicts, peoples’ behavior). There is also considerable uncertainty both about the main processes and relationships within the system, and about the availability and reliability of data. Attempts to improve this situation must go beyond the physical ecosystem to include the social and political interactions.

The third example concerns risk management at the Notting Hill Carnival [4], a huge street party combining music, dancing, a procession and street trading organized by the West Indian community. It lasts three days and attracts over a million participants, making it the largest street festival in Europe and perhaps the second largest (to Rio) in the world. Clearly there are many interest groups associated with such an event, at least some of which may have historically antagonistic relationships: the Metropolitan Police, the West Indian community (itself split in several ways), the local residents, the Local Government Authority, shop owners and the participants themselves.

Many risks are associated with this situation including threats to public order (several carnivals during the 1970/80s resulted in outbreaks of violence), public safety (e.g., through crushing), environmental health (toilet arrangements, food safety) and crime, especially theft. Project organizers and main stakeholders met to consider ways of redesigning the carnival to take into account its changing nature and changes in expectations and legal requirements.

Although these examples are very different, they all exemplify particular characteristics of “problem situations” or wicked problems that make the traditional mathematical modeling tools of O.R. ineffective. In particular:

  • The “problem” itself is not well-defined with agreed objectives such that efficient means to achieve the objectives can be constructed. In the above examples, even non-optimizing methods such as critical path analysis or simulation could not be used.
  • The situations all involve several interested parties whether they are departments within the organization or cooperating (or conflicting) external bodies. These generally hold different perspectives about the problem situation.
  • There are many uncertainties and often a lack of reliable (or indeed any) data.
  • “Success” can only be seen in terms of agreement among the parties involved to undertaking particular courses of action. The process is primarily one of learning and negotiation rather than the technical solution of a problem. It often involves facilitated workshops of concerned stakeholders.

Swamp Vs. High Ground

These kind of complex and messy problem situations have long been recognized by the likes of Ackoff [5] who termed them “messes” as opposed to “problems,” Rittel [6] (“wicked” vs. “tame” problems), Schon [7] (“swamp” vs. “high ground”), Ravetz [8] (“practical” vs. “technical” problems) and Checkland [9] (“soft” vs. “hard” approaches). I would argue that they are in fact very common. The reader need only reflect on his or her own personal experience at work or at home. I would also argue that these problems are usually significant and their resolution, or sometimes dissolution, has wide ranging effects. Or, to put it another way, strategic problems — i.e., those that are not short-term and narrowly focused — are usually complex and messy.

Thus, the argument I put forward in this article is as follows:

  • A variety of methods, collectively called Soft O.R., have been developed to help tackle these messy strategic problems. They are not mathematical, but they are nevertheless rigorous and they have been very successful in practice.
  • Their general characteristics are [10]: they allow a range of distinctive views and objectives without collapsing them into a single measure; they encourage the active participation of stakeholders in the modeling process; models are generally non-quantitative and transparent to the participants; significant uncertainty is expected and tolerated; they aim for exploration, learning and commitment rather than optimization.
  • After some initial skepticism from the O.R. discipline they have now been fully accepted by both practitioners and academics throughout the world with the exception of the United States.
  • This is an undesirable situation because it splits the discipline of O.R. in two, and because it denies U.S. practitioners knowledge of these powerful methods that would help O.R. to “venture outside the O.R. comfort zone” as former INFORMS President Brenda Dietrich recently urged [11].

soft Operations Research methods

The Development of Soft O.R. Methods

The limitations of traditional mathematical O.R. methods became apparent during the 1960s and 1970s. C. West Churchman, in an editorial in Management Science in 1967 [12], brought Rittel’s concept of wicked problems to attention: “social problems which are ill-formulated, where the information is confusing, where there are many clients and decision-makers with conflicting values, and where the ramifications in the whole system are thoroughly confusing.” Ackoff’s [13, 14, 15] searing critiques of the development of O.R. up to 1979 are well known. And it was during this period that the main Soft O.R. methods were developed by academic/practitioners in response to practical engagements with real problems.

Peter Checkland, who was appointed Professor of Systems at Lancaster University in 1969, developed the foundations of Soft Systems Methodology (SSM) during the ensuing decade through a long series of industrial projects [16]. He saw his task as taking traditional, hard-systems engineering methodologies (e.g., Hall [17]) and transforming them to be able to deal with the humanness of human beings, highlighting the importance of irrationality, creativity and values [18]. The development of SSM has been well-documented in three books [19, 20, 21], the second of which (“SSM in Action”) is wholly concerned with applications of SSM.

In brief overview, the developed form of SSM involves the following stages:

  • Discover as much as possible about the problem situation, especially its history, the nature of the consultancy engagement and possible issues, the prevailing culture and the power and politics. Express this in the form of rich pictures.
  • Develop systemic models of purposeful human activity that explicitly embody particular viewpoints or perspectives relevant to the situation.
  • Express these logically in terms of root definitions and conceptual activity models.
  • Use the models as a way of questioning and exploring the situation to structure a debate between involved parties about desirable and feasible changes.
  • Gain agreement on changes to the situation that the different perspectives or worldviews could accommodate.

It works better if much of the activity is actually carried out by the participants in the situation, with the O.R. practitioner acting as a facilitator, as they are the ones who have the detailed knowledge, and it is they who must eventually commit themselves to taking action.

How Effective are Soft O.R. Methods?

Many new methods were developed to deal with “wicked” problems, but were they actually successful? Clearly many projects carried out by practitioners are never written up and published, so the evidence is to some extent only the tip of the iceberg.

The first published surveys of Soft O.R. use was probably Mingers and Taylor [22], who surveyed O.R. and systems practitioners (some of whom were also academics) about their practical use of SSM. More than 90 users of SSM responded to the survey, and 66 percent had used SSM more than once. The most common benefit was that SSM provided a coherent structure both for managing the project and for dealing with the complexity of the situation. Sixty-three percent evaluated their success with SSM as “good” or “very good.” This study was replicated in Australia [23] with similar results. More recently, van der Water et al [24] produced a classification scheme for applications of SSM based on published articles. They discovered more than 110 papers on SSM alone. The main areas of application were ecology and environment, information and communication technology and action research.

Soft O.R. in the United States

Unfortunately, Soft O.R. has been marginalized, if not completely ignored, within the O.R. community in the United States. If Soft O.R. is alive and well and successful throughout much of the world, why hasn’t it found a home in the U.S.?

It is certainly one of the main philosophical differences between Soft and Hard O.R. that Soft O.R. tends to take an “internal” view of the problem situation, recognizing and valuing the viewpoints of those most closely involved. In its developed form, Soft O.R. sees its role as one of informed facilitation of key participants using rigorous and structured methods to elicit and debate differing worldviews. In this sense it does pay heed to the preconceptions and prejudices, not of the investigator but of the stakeholders.

From the viewpoint of traditional O.R. (and the U.S.-based journals that focus on traditional O.R.), this could be seen as a weakness for O.R. always claimed its legitimacy from its scientific approach, its mathematical models and its supposed external objectivity. I would argue that this view of the validity of O.R. became untenable many years ago, and is neatly summarized in Ackoff’s paper “Optimization + Objectivity = Opt Out.” There is not the space here to go over these arguments again but I would refer the reader to two papers by Maurice Kirby about the history of O.R. [25, 26] which, together with a selection of classic papers [27-38], should make the case more eloquently than I can.

My own view is that Soft and Hard O.R. are not alternatives but are complements to each other [39]. Every complex real-world situation has aspects that are amenable to quantitative analysis, and other aspects — such as culture, power and politics — that are simply not. The two approaches can therefore usefully be combined together, although in my own personal experiences of real projects the non-quantifiable aspects often dominate.

An alternative view might be that O.R. in the U.S. does tackle these kinds of problems but with its own methods or with “softened” versions of traditional methods. An example of the former might be Saaty’s Analytic Hierarchy Process (AHP) [40]. This is a well-known and widely used process for helping decision-makers make choices between alternatives where there are multiple criteria and so is seen as an example of multi-criteria decision analysis (MCDA) [41]. In terms of the characteristics of PSMs stated in the introduction, AHP recognizes different criteria but is essentially a method for combining them all into one. It does involve decision-makers but only at one point and has to generate a consensus between them; it is clearly quantitative, the whole point being to force subjective and often fuzzy preferences into ratio-scale numbers; and some parts of the process are reasonably transparent, but the algorithms are clearly not.

An example of the latter could be Ralph Keeney and his value-focused thinking approach [42]. This is essentially a form of decision analysis, but it involves eliciting a whole range of objectives and values through discussions with key stakeholders, structuring these into means and ends and investigating measures of performance and stakeholder views on tradeoffs where they conflicted. Methods used included workshops, workbooks and influence-type diagrams. This certainly has the hallmarks of Soft O.R. and to that extent I would say “great,” but given that many other methodologies are available and tested why not use them?

However, if we look at the situation from a more historical and sociological perspective we can see that the discipline has developed in a particular way in the U.S. that reinforces the dominance of mathematically based modeling at the expense of Soft O.R. For example:

  • The top U.S. journals only publish mathematical papers (see the mission statement of Operations Research, for example), and gaining tenure and advancement requires publications in such journals.
  • O.R. courses in the United States only cover traditional O.R., so new researchers and academics do not get exposed to Soft O.R.
  • In U.S. practice, O.R. confines itself to tackling the kind of problems it knows it can solve (see Brenda Dietrich’s paper mentioned earlier), and the really “hard” problems are left on the shelf.

I hope this article generates an informed discussion and debate, carried out in a good spirit, which may lead to a healing of the unfortunate rift between the O.R. community in the United States and proponents of Soft O.R. as practiced throughout much of the rest of the world.

John Mingers ( is a Professor of Operational Research and Systems at the Kent Business School, University of Kent, Canterbury, United Kingdom.

Acknowledgments, Further Reading and References

The author acknowledges the helpful comments of Fran Ackermann, Peter Checkland, Colin Eden, Alberto Franco, Mike Jackson and Jonathan Rosenhead.

For readers interested in finding out more about Soft O.R. methods, see “Rational Analysis for a Problematic World Revisited,” edited by J. Rosenhead and J. Mingers (Wiley, 2001).

For a complete listing of references appearing in this article, see



Analytics Blog

Electoral College put to the math test

With the campaign two months behind us and the inauguration of Donald Trump two days away, isn’t it time to put the 2016 U.S. presidential election to bed and focus on issues that have yet to be decided? Of course not.


2017 Tech Trends: the kinetic enterprise

From dark analytics to mixed reality, machine intelligence and blockchain, Deloitte’s annual Technology Trends report analyzes the trends that could disrupt businesses in the next 18-24 months. CIOs who can harness the possibilities of these technologies will be better positioned to shape the future of their business. Read more →

FICO: Financial crime trends for 2017

If the world seemed dangerous from a financial crime perspective last year, FICO experts predict an even more challenging 2017. In a new paper, four of the leaders in the company’s fraud and financial crime group laid out 17 predictions, ranging from killer devices in the home to hacked fingerprints. Read more →




2017 INFORMS Healthcare Conference
July 26-28, 2017, Rotterdam, the Netherlands


CAP® Exam computer-based testing sites are available in 700 locations worldwide. Take the exam close to home and on your schedule:

For more information, go to