GPM First
Chapter 1 of Project Risk Analysis (978-1-4094-4241-7) by Derek Salkeld

The Case for Risk Analysis


Why is So Little Known about Project Costs and Timescales?

In 2008, the British economy turned over just under £1,450 billion.1 [1] The measure of this, the gross domestic product (GDP), comes in three forms,2 [2] one of which is the value of the goods and services its people produce (Figure 1.1).

All of our efforts to keep the wolf from the door are in there somewhere, but when I ask people in which part of this spectrum they work, it quickly becomes apparent that there are quite a few people for whom the answer is a ‘kind of neither, probably both’ portmanteau expression that German speakers would have little problem expressing in a single word. They do not work in manufacturing, but similarly they do provide relief to those in need. They are neither producers nor providers of service. Instead, they say they produce a good, or perhaps just a small number of them, a specific product to a specific client. They say they work in teams that deliver motorways, bridges, computer systems, production lines, power grids, buildings and so on. They work in what I suggest is a hitherto unsung sector of the economy called ‘Projects’ (Figure 1.2). How big this sector of the economy is I do not know, but I would not be surprised to learn that it is both large, and that a small number (tens of percent) of the UK workforce consider themselves to be part of it.

Since working on projects is economically productive, there must be a band in this spectrum equal to the value of the projects completed annually. Where it sits, towards the goods or services end of the GDP spectrum, is not clear, but it has to be in there somewhere.

It almost goes without saying that the value of the goods the UK produces has declined in recent years. However, since the GDP has grown, it is reasonable to assume that the value of UK services must have increased

Among the goods we can no longer buy are Raleigh bicycles (started in 1887 and closed in 2003), Rover cars (1885 to 1967, when the company became part of British Leyland), and Norton motorbikes (1898 to 1975). Among the services that have replaced them are English-language entertainments (BBC TV, the West End), tourism (easyJet) and finance (providers’ interests, first pensions, money invested in money, bets placed against the economy, private equity purchases of successful companies, etc.). Indeed, we may have been too successful at the last. Though the profits made by these companies may have generated taxes for the Exchequer to expend (or invest of course), being privately owned means those profits have not been accessible to the masses due to the equity no longer being publically traded on the stock exchange. I digress, but perhaps even for a capitalist like me, life is a progression from the political left to the political right and then slightly back again.

Figure 1.1 The UK gross domestic product spectrum I


Figure 1.2 The UK gross domestic product spectrum II


Somewhere in the spectrum of goods and services lies the projects sector. Projects have provided most of the objects and products around us that we would consider evidence of a modern, civilised society. Many of these such as roads, railways and airports are clear, apparent to the masses, while others such as broadband, air traffic control and customer information databases are less so.

Projects are distinguishable from goods in that generally each project produces only one good before it is closed, and even though some projects are commissioned to set up facilities to produce goods, once the products start to roll out of a new production facility, the project is finished. A subsequent requirement for a further good is usually the cause for a further project.

Projects are also distinguishable from services. The idea behind a services-based business model is to reduce costs and improve quality through economies of scale and repetition. If I can work out a way of doing something you do but for which you have neither the time nor the inclination, and if I can persuade you it would be better value, then I ought to be able sell that service repeatedly to you (and to others). I would make my profit from multiple sales, probably accumulating it bit by bit as the business year rolls by.

If what you need however is a project, then you are probably only going to commission one from me. I need to make my annual profit, or rather, earn your contribution to it, from a single sale, normally taking it at the end of our business deal when the project is finished. Moreover, when it comes to building the next thing you want – power distribution network, ship, or satellite – I will probably have to persuade you to use me all over again, usually through some form of competitive tender. It would be very unwise of me to base my business plan on a presumption of a number of multiple sales to you previously.

Of course there are many companies that deliver major projects throughout their business year, receiving repeat business (without competition) from satisfied clients. Goods and services are generally typified by lots of smaller projects and sales.

I may have given an impression that projects is the harder business sector to be a part of, but this is not entirely the case. There are some compensatory aspects that are extremely attractive. First, projects is arguably a much less risky business than goods or services because project costs, staff, facilities, materials and such like, are often reimbursed by the project funder within a few weeks of them being incurred by the project delivery company. Secondly, if project-delivered goods do not work as intended, remedying the situation can possibly be paid for by commissioning a further project. This sounds slightly scandalous, but my perception is that projects sometimes deliver things that are markedly different from the original designs, and that possibly the newer items may have attracted additional funding. You must also contrast this with the money-back guarantees and refunds – necessary nowadays to secure sales in the goods sector. Finally, if the project good does not provide the benefit foreseen, say perhaps the number of drivers across the toll bridge is fewer than forecast, then that is not the fault of the project-delivery company, nor is it a loss-making scenario. The project delivery company only does what it is commissioned to do and doesn’t generally take responsibility for the level of use of the good.

So somewhere in the GDP spectrum of goods and services is a ‘neither one nor the other’ band, a group of projects that spend other peoples’ money and accrues its profit on single trades, usually some time after the work has been commissioned. There is evidence that this band exists everywhere we look: telecommunications systems, water supplies, power generation and distribution, transportation, information systems. But there is no sight of it as a sector in our economic statistics (though there is evidence in the services end of the spectrum of its use). Perhaps this economic data is available to economists. It would let them quantify the parameters of the sector, and though I am not suggesting for one moment that there is a conspiracy on the part of those involved in projects not to publish this data, I can see an economic rationale for its absence. I will illustrate this using four case histories from projects that have made the front pages and the prime time broadcasts in their own countries. They are:

  • The Australian Federated Health Care Records (IT system)
  • RAPID, The European Heavy Logistics Aircraft (Defence)
  • The Buenos Aires Urban Light Rail Transport System
  • The Sino-Canadian Polar Air Traffic Control System


I spoke at a university conference recently and asked the audience of academics and students to speculate why these projects were in the news headlines. I suspect the students took their lead from their professors but regardless, they all agreed it must be because the projects were over budget and late. No one suggested that they just opened then closed. No one suggested they were white elephants, or even posited that they were associated with infamous accidents.

My presentation was entitled ‘What we know and what we don’t about projects’, and I think that I was successful in convincing my audience that what all of us know about projects, even those who are not involved with them, is that they cost more and finish late.

I have carried out this unscientific vox populi experiment a number of times over the last ten years, enough for me to be confident of the outcome. With one exception I have never had any response other than that the projects must be over budget and late. The one exception was right though: none of the projects exist. I made them up and so they are not famous at all, but in the minds of the audience projects must be newsworthy because of their flaws. It is what we have come to expect.

If I had presented the names of four fake goods in my lecture and asked the audience why they thought those might be newsworthy, I suspect I would not find a consensus in the answers. I also suspect the same would apply to four fake services. If there was a consensus among my audience then it would probably be a deduction made on the basis that, since all successful goods and services attain such a strongly branded recognition in our shopping society, these ones must be the flops. It would be the reason they’re newsworthy. Fuzzier, more nuanced business-speak might be that the goods were poorly positioned in the market, there was a lack of quality control of the service, the underlying business model was inflexible, there was a failure to adapt to emerging markets. In reality, ‘over budget’ and ‘late’ would be suitable responses. If there is an ingrained perception of failure, both financially and in relation to time, what does one do if they are working in the projects sector of the economy in order to thrive?

This is not, metaphorically speaking, a theological debate about what people believe to be the truth. Though it may not be supported by evidence, there are justifications for this public perception. In the 1997 edition of his book The Management of Projects,3 [3] Professor Peter Morris explains how he could find the final out-turn costs for only 1449 projects in the public domain and that of these, only 12 had been delivered on or below budget. I could see that number from numerous office windows in any big city. Professor Morris also describes a later analysis of 3000 projects that found a similar result.

How then do companies that deliver projects survive when there is evidence the sector fails to deliver on time and to budget? Sorry to sound cynical, but ‘on time, on budget, on occasion’ is not going to be a business-winning pitch for a company, no matter how bravely honest it may be. Perhaps it would be wiser, all things considered, to keep quiet about overspends and overruns in case admission of their existence may be seen as evidence of lack of grip and drive (essential prerequisites for employment as a project manager). It is vital that prospective clients are assured that past performance is not a guide to the future.

It is also prudent for those that deliver the project not to let their performance (delivering on time and to budget) be subject to analysis by economists. Those researchers would need to identify their sources if their work were to have credibility when peer reviewed for publication. This is obviously not the case with goods and services, both of which offer researchers rich seams of data on supply, demand and pricing to mine for econometric understanding and insight.

I suspect the reason people and companies involved in the projects business survive overspends and overruns is the same as the reason why projects does not feature in the GDP spectrum of goods and services – information regarding the frequency and extent of failures to deliver on time and budget is never released. Though not completely true, published data about the actual timescales and eventual costs of projects are scarce.4 [4] The projects sector keeps quiet and gives out nothing by which it may be delineated and measured. One is therefore led to conclude that it is not in the economic interests of any one whose livelihood is earned in the projects sector to publish information about their performance. There is no commercial advantage to be gained from doing so, and I suggest this is why the sector is not recognised in the GDP spectrum.

The Case for a Better Understanding of the Projects Sector of the Economy

Delivering projects as a business venture, in which someone provides a good to someone else for a fee, has probably been around since public works began. A Eurocentric view would choose the forums of the Mediterranean city states or the pyramids of Egypt as early example. This led to the introduction of grids of one form or another: roads, canals, rail, mass housing and so on, upon which our current daily lives have very much come to depend for their relative ease and comfort.

I estimate that approximately 250 years’ worth of projects have been designed and built by project delivery businesses. Inevitably, a Web search of project management companies finds thousands, so I started to draw up a list of companies I know of that are currently practicing in the sector and looked up when they started: 1920s, 1880s, 1944, 1978, 1848, 1901, 1885. Clearly this is a large and a persistent business sector. The questions that then arose were:

  • If these the companies have all been around for a long time, how could they have avoided overspending and overrunning?
  • If they have all been trading for longer time frames than any of the projects they have worked on, then in spite of probably enduring their share of overspends and overruns, they must have been able to continue efficiently trading to attract further business.


My suggestion in relation to the first point is that they keep quiet about it. To the second – that like railroading in the nineteenth-century United States – dominant positions for the design and delivery of certain kinds of projects may be capable of being established through company size, experience, expertise, connections to political power and an ability to provide the funding required. One can see that access to these would make a compelling case for the appointment of a project delivery company, even if its track record of delivering on time and to budget were no better than any of the others.

Where the market demand will be for the supply of project delivery services, and what types of projects will be commissioned therein over the coming years, I do not know. The projects sector of an economy is not as well understood in terms of the benefits or encumbrance it brings to societal development as it should be. Certain projects in certain circumstances may be worth supporting politically in order to stimulate and develop expertise that could then bring economic benefits to the country through its export. Not the projects of science fiction: monorails everywhere and hologram holiday experiences, but the deliberate nurturing of a capacity and an ability to deliver reliable, useable and appreciated assets in markets that needed them. If you want a good car, turn to the Japanese who have plenty to choose from. But if you want, say, a potable water network or an electricity distribution grid then why not turn to the British? However, the establishment of a good reputation for delivery of projects in a particular field will always be difficult – impossible – if they are persistently delivered over budget and late. The UK has already exported ‘lines and a ball’ style games – cricket, football etc., and with systems of tolerable government around the world and I can see no reason why the world demand for projects should not automatically turn to UK PLC. One has only to look at the presence of US project management companies in the UK, doing what I thought we could do perfectly well for ourselves, to deduce that there is an export market for project delivery expertise. It does not matter what the types of projects are, but it does matter that we avoid overspending and overrunning.

Some Attempts to Compensate for Potential Overspends

The political interest in projects is considerable because government departments and agencies are major funders of them. As a result there have been several initiatives to develop processes that will avoid potential cost overruns on projects.

Perhaps the project with the highest public profile is the Private Finance Initiative (PFI) in which the public sector contracts the private sector to deliver an asset it then pays to use. The private sector designs and constructs, and often operates, the asset using its own sources of funding. As a result, some form of contractual arrangement is put in place to ensure the private sector accrues enough usage charge over time to cover its costs and make a profit. One key reason for procuring projects in this way has been a belief that the private sector will be more successful at delivering the projects on time and to budget than the equivalent public sector organisations.

Since the private sector tends to borrow the funds it needs to design and construct the asset, there is an argument that because the government could borrow those funds at a lower rate of interest than the private sector, its own project teams could deliver the project for less. This argument is predicated on an assumption that the public sector has project delivery organisations similar to those at the private sector’s disposal. However, one has only to read the daily government procurement notices over a period of time to become aware that this may not be the case, for what was once done in-house by the public sector must now be contracted out because the public sector no longer has the wherewithal. To illustrate the point in an unscientific way, here is a typical procurement notice (Table 1.1) from the Official Journal of the European Union for the kind of project management services the local authority concerned, whose name I have redacted, would at one time have had the capability to do for itself (own in-house resources). Now the capability and capacity has to be bought in.

Local authorities once had departments who did the kind of work in the notice, and by inference, the requisite skill base was maintained in-house. Now that it is contracted out, however, the public sector expertise is in the buying of the skill, rather than the practice of the skill itself.

Table 1.1 A procurement notice for the contracting out of project management services from the public to the private sector

Notice Details



UK: business and management consultancy and related services

Name of awarding authority:

xxxxxxx Borough Council

Award criteria:

The most economically advantageous tender

type of awarding authority:

Local authorities

CPV product code:

45210000 (Building construction work.)

45214000 (Construction work for buildings relating to education and research.)

45314000 (Installation of telecommunications equipment.)

50700000 (Repair and maintenance services of building installations.)

51610000 (Installation services of computers and information-processing equipment.)

70000000 (Real estate services.)

71000000 (Architectural, construction, engineering and inspection services.)

71240000 (Architectural, engineering and planning services.)

71315000 (Building services.)

71541000 (Construction project management services.)

72222300 (Information technology services.)

72500000 (Computer-related services.)

72514300 (Facilities management services for computer systems maintenance.)

Nature of contract:

Service contract

Type of Procedure:

Competitive Dialogue

Publication Date:


There is nothing in the strategy of contracting-out project delivery services that specifically mitigates overspending and overrunning, other than the assumption that the private sector can manage projects better than the public sector. I am not sure about this, because the work of delivering projects is the same be it done in the public or private sector, and I am certain there are (or were) just as many talented project people in the public sector as in the private. The public sector is at least as able as any other employer to attract project management talent, not least of which is its ability to offer what I think must be a deeply attractive prospect: using a career to build an entire system, a city say, instead of being a peripatetic builder of, for example, office blocks. The rebuilding of places like London and Plymouth in the 1950s and 60s, and of building Milton Keynes in the 1970s must have made for many a satisfying public sector career in projects.

The reason project delivery services are gravitating from the public to the private sector may be that when projects in the private sector are over budget and late this no longer emerges into the public record, but instead into a private one. Those who see it emerging may well take a commercial view that perhaps with some judicious renegotiations of the loans to reduce the interest rate, or some astute adjustment of operational and maintenance regime, any overspends and overruns can be quietly dealt with over time, well away from the public gaze and inquiries by a fourth estate.

Not all projects naturally cleave to a PFI model. Some are simply too complicated. They contain too many systems that need to be integrated flawlessly and sustainably for the delivered asset to work reliably and consistently to secure the initial funding required. A high-speed motorway project with interfaces to slower major roads may be possible to fund through PFI, but I doubt a high-speed railway with interfaces to slower lines could be. There would be too many component systems: rail, rolling stock, traction power, signalling, telecommunications and civil engineering structures. It would not be impossible of course – but it would probably be a scheme that would be too difficult to arrange, hindered by the complexities of resolving liability for train delays and cancellations among individual system providers. Warships and other military system projects come into the same category, with the additional complication of the possibility of the delivered asset to be lost when in use.

Other projects will have too many external parties whose competing needs have to be accommodated, moderated or confronted in order for a private sector project company to be able to get on with the job without disruption. The design and building of a modern Olympic Games is an example of this.

And so, though the public sector has contracted out a lot of project delivery to the private sector, in the end not all projects are suitable for such an approach. In these cases the public sector has little option but to make a presumption that the funding and timescale estimates are sufficient to get on with the project itself.

Unfortunately there is evidence of concern in central government that its devolved agencies and branches frequently fail to get these sums right. In 2002, Her Majesty’s Treasury received a report entitled A review of large public procurement in the UK, prepared by the consulting engineers Mott MacDonald.5 [5] It ‘demonstrates the existing high level of optimism bias in project estimates arising from underestimating project costs and duration or overestimating project benefits’. Mott analysts were given data regarding overspends on 50 projects. They split them into five categories and reached the not entirely statistically robust conclusion that projects of certain types tended to become overspent by certain percentages. These overspends were attributed to an optimism in the original estimates prepared by the project sponsors: thus the term ‘optimism bias’ (OB) has become a quite well-known percentage mark-up, so much so that is added to many proposals submitted to the Government for funding (Table 1.2).

Table 1.2 Optimism bias percentage adjustments for duration and capital expenditure

Project type

Duration Upper % Lower %

Capital expenditure % Upper % Lower %

Non-standard buildings





Standard buildings





Non-standard civil engineering





Standard civil engineering










One may wince at the confidence with which the results of this analysis were presented, but no one really doubted the basic truth that overspends were a persistent phenomenon, and certainly no one would bet their careers and investment funds that the next project along would be underspent and delivered early. However, the data set was statistically too small to support the inferences made. Such is the reticence that surrounds the public disclosure of project accounts, it may well have been all the data there was.

In addition to the limited data set, there is another problem. I was once commissioned to calculate the size of the optimism bias mark-up that needed to be added to a hybrid project comprising a mix of civil engineering and IT works. A colleague6 [6] pointed out that whereas the final cost of a project should be just the one, and its value known, the other side of the ratio, the initial cost, could have had several candidates, each reflecting the different scopes, specifications, baselines and development maturities that are the typical outputs of the early stages of projects. This raised the question: in the data set used to fix the optimism bias, what had been the maturity of the initial estimates? And, to take it a step further, after the initial estimate, had the projects been successfully or badly managed? These important measures of the appropriateness of applying OB to other projects were simply not available.

I might easily have applied an OB adjustment to a very mature and well-understood project intended for delivery by an experienced team that had been derived from a small set of estimates from a number of poorly managed, overspent projects and their early days, back of a fag packet, estimations. The opposite may equally have been the case: an OB adjustment based on well-performing projects could be applied to one that was going to perform badly.

The potential for the inefficient use of funds is obvious. OB gave project managers money but the reward came with no reason why it was the amount it was. Optimism bias was meant to cover everything without explaining why and what for. It further struck me that if OB mark-ups were always applied to project estimates then it would not be long before design and contracting companies would quickly see them as an opportunity to increase their prices. Like C. Northcote-Parkinson’s Law that work expands to fill the time available, prices would rise to absorb the additional funds that OB brought. Furthermore, without any communication from the past to the present about what the underlying difficulties were for which the treatments are an OB mark-up, I would not be surprised to find that after a few rounds of funding projects with OB applied, projects were still overspending and overrunning because managers would not be obliged to spend OB funds mitigating those difficulties. The funds may simply be expended on the works generally, for example, by the acceptance of higher tender prices. It is very rare for a project that has stepped out into the public spotlight of issuing invitations to tender to be stopped because the quotations received are slightly higher than estimated. One can imagine commercial belts being loosened all round with sighs of relief on learning that the calculations of the funding required for a project include a non-specific OB mark-up. Overall OB provides a layer of financial cover, whether or not any underlying problem is likely to pertain to a particular project or not. This may not only be inefficient, it may also be ineffective because the funds are not aimed at anything specific. OB mark-ups can also be very high for projects that must use a mix of engineering technologies, sufficient perhaps to overturn the case for doing them if they were blindly applied without consideration of appropriateness.

A more sophisticated approach was needed so the Department for Transport commissioned a more statistically valid analysis from Bent Flyvbjerg, then working in the Department of Development and Planning at the University of Aalborg, Denmark.

Professor Flyvbjerg, in association with the consultants COWI, reported previous work in which 260 transport projects had been studied.7 [7] Professor Flyvbjerg noticed and described a behavioural phenomenon in which people working on projects, their roles ranging from design and construction staff to commissioning agencies and funders, and even politicians at both local and national level, seem to collude unconsciously in an underestimation of the costs and timescales needed by a project, as well as an overestimation of the benefits that will be obtained once it has been completed.

There is no suggestion that there are conspiracies to get a project approved and started, or even connivances based on a nod, a wink and a tap of the finger on the side of the nose. But the sense of there being an unspoken understanding between those whose livelihoods and place in history may depend on a project going ahead can be inferred from the very readable book Professor Flyvbjerg wrote about his researches.8 [8]

The mathematician John Forbes Nash described an eponymous equilibrium in which apparently competing players in a market were discovered not to be doing so, each of them coming to a conclusion that the rewards they were getting were sufficient and that competing against each other for more was neither worth the risk of losing market share nor the additional effort required. Perhaps something similar exists in the project world between the stakeholders? Everyone is accruing something they value from the current phase of the project development, and they subconsciously understand that if they behave in ways that fulfil the assumptions and needs of the others, then the next stage of the project will more likely gain authorisation to proceed. Consequently, they may stand a decent chance of continuing to be party to it and thus sustained in business. All it may take for a project to continue in development is that no one should rock the boat unnecessarily, and one way to do that would be not to show that the costs and timescales are intolerable but instead to suggest that with further diligent research, a way to make them acceptable will be found.

Professor Flyvbjerg improved on the Mott MacDonald analysis by devising a percentage mark-up that should be added to the estimated costs and timescales for a project if it were to have a given level of confidence in the final costs and completion time scales before it was authorised to start (see Footnote 8). The greater the level of confidence required, the greater the mark-up to be applied. The curve in Figure 1.3 shows the percentage uplift that should have been added to the estimates for projects in Professor Flyvbjerg’s data to deliver roads for an acceptable chance of a cost overrun.

From the figure, if only a 10 per cent chance of overrun is tolerable, i.e. there must be a 90 per cent chance a road project must be delivered within budget, then the estimate must be increased by approximately 65 per cent.

Though this was a far more robust analysis, and one of considerable value to financial analysts, it is still susceptible to the same drawbacks as the Mott McDonald one when it comes to commercial and managerial practicalities: the mark-up does not cover anything specific so there is no need to spend it on the treatment of the causes of overspends and overruns. Whether it would be better to spend the money on hedging against demand-led cost increases or setting up a systems integration and test facility separate from the main operational facility is not known. There is simply a pot of money waiting to be used for something unspecific that may or may not emerge. Indeed, because the mark-up is derived from an analysis of many projects, when it comes to a particular case of a project in difficulty it may be more than sufficient. Of course it may also be insufficient, and so the potential for inefficient and ineffective use of funds remains. Further, given a non-specific fund to expend, how should the project manager use it? Do they keep the fund as a contingency measure, for use in the event of something deleterious? Or should it be used to punt on a pre-emptive act intended to avoid it? If the former and the event happens, how can the manager be confident it is not the first of many, and thus able to judge how much of the fund to use and how much to reserve?

Figure 1.3 A Flyvbjerg curve


I think the efforts to evaluate a compensating mark-up to cover project overspends are rational, a good thing and academically sound, but a mistake from the standpoints of project funding and management because of the possibility of the inefficient and ineffective use of funds and because managers are given little intelligence about the what, why and when of overspends that they can assimilate and perhaps mitigate.

Projects overspend and overrun. They have done so enough for academics to have noted it and for the UK government to have tried to do something about it: but the solutions I think are unsatisfactory. I want to propose a better one.

It may not, in its early stages of development, be any more accurate (in its predictions of outturn costs and timescales) than the mark-up solutions, but I hope that its practice will stimulate challenge and improvement so that ultimately it is seen as a much more accurate and insightful approach.

I suggest projects overspend and overrun because they were not estimated correctly in the first place. If I were a scientist seeking to advance human knowledge and I devised a scientific hypothesis that was subsequently invalidated, I would try to think of a better one, one I hope would not be confounded by the data. It seems to me that the experimental evidence tells us that projects cost more and take longer than intended, and accordingly that this must invalidate the ideas, and hence the methods used to estimate their costs and timescales. I don’t think the calculations used are entirely the right ones. In the same way that Newtonian mechanics begins to fail experimental tests at the subatomic level, resulting in the need for quantum mechanics, I suggest the basic methods of estimating the money and time needed for a project begin to fail when a project becomes big in some way or ways: physical size, number of components, geographical spread, length of time needed etc.

What I believe would make the estimating methods more accurate is risk analysis.

An Approach to Forecasting the Potential Overspends in Project Costs

I would like to describe a small experiment which, like the fake projects I listed earlier, is a bit of a cheat. Mathematicians will spot the deception in an instant, but even so, it has worked so well with students that I am told one of them has adopted it for use in pitches to potential clients. I hope you will tolerate its naivety because a point needs to be made.

Imagine I want to commission a project to build a garage with a drive leading up to it. I approach a builder and he gives me a price of £3 for the garage. The numbers are implausible but make what follows easier to assimilate. The builder only erects structures and so I have to hire a ground works contractor to lay the drive. He too gives me a price of £3. I then apply to the bank for a loan of £6 to fund my project.

The bank are concerned about this scheme and call in a risk analyst (it is a fantasy story) who then researches the supply and demand of garage components in the marketplace, as well as what can go wrong with their assembly and commissioning. They find that the supply of components outweighs the demand. As a result they can be readily purchased for £2. Some suppliers even want to offload stock and are prepared to sell them for £1.

Research into problems with existing garages finds that if the slope to the base on which they have been erected is not level, it can cause the embrasures for the door and window openings to be out of true, necessitating the planing of the doors and windows to fit the hole. These extra works can add up to another £3 of costs for the garage.

We now have a scenario in which the garage could cost between £1 and £6, the former having no problems, the latter a full set of commissioning problems. The £3 estimate has, in risk analyst argot, a risk–opportunity spread of £1 to £6 (please ignore reversed order of the numbers vis-á-vis their description. It is easier to say, that is all).

Research also finds that the concrete has an estimated cost of £3, but demand is weak so the order may be available for £2 (or even £1 if a cancelled order for elsewhere can be taken). The quantity needed however is variable because of undulations in the ground, and an additional smooth mix may be needed to give an acceptable finish to the top surface. Again the analyst finds a £1 to £6 risk–opportunity spread on the £3 drive.

The problem is how to add that up? Adding £3 for the garage and £3 for the drive is easy enough, but £1 to £6 added to £1 to £6 is counter-intuitive. Thankfully, it can easily be calculated. Imagine two dice from a board game, one to six; one red, one white. The red one is the cost of the garage, the white the cost of the drive. I have two here that I will roll as I write:

Throw 1:

Red 2

White 4 &hellips; cost of the project = £2 + £4 = £6


but that is not necessarily the cost because

Throw 2:

Red 3

White 6 … cost of the project = £3 + £6 =£9


which is not necessarily the cost either because

Throw 3:

Red 2

White 1 … cost of the project = £1 + £2 = £3


And so on. Each throw generates a perfectly valid cost for the project. Any number on a dice is as equally likely to occur as any other number.

Some more throws:

Throw 4

total £7

Throw 5

total £8

Throw 6

total £7

Throw 7

total £10

Throw 8

total £4

Throw 9

total £7

Throw 10

total £3


and so on.

If these different but valid costs are plotted on a histogram to show the number of times each total has occurred, and the dice are thrown another 100 times, a pattern begins to emerge (Figure 1.4).

Obviously the minimum total cost of the project will be £2 and the maximum £12. But what also becomes apparent is that £7 occurs more often than any other total cost. I can conjecture that if I incorporate any risk–opportunity spreads I can find the estimate for a project. The most likely total cost seems to be more than the original estimate without them.9 [9]

If I had borrowed only £6 with which to build the garage, the most likely outcome of the project would be that I would run out of money and have to seek a further £1 to make up the £7 needed. By not researching the risks and opportunities, as well as not including those identified in the funding calculations, I would have booked myself a place among Professor Morrison’s set of 1437 rather than the 12. My project would most likely be overspent in due course.

Figure 1.4 Histogram of valid project costs


I think there will be some irritation over the above triviality – I acknowledge it is a long way from a nonsense garage project to a real-world major infrastructure project, so let me close this part with five serious and worthy points that I hope my example showed.

  1.  When we take into account the risk and opportunities, our understanding of the funding required alters. In gaining this insight we also discovered that the risks and opportunities data underlying this result were accessible and capable of being analysed quantitatively.

  2. If the risks are larger than the opportunities, and they usually are in real-world projects, then the funding required will be more than the estimate originally indicated.

  3. If we had proceeded with the project on the basis of the estimate alone, in due course we would probably have found the project overspent.

  4. Not only did the calculation show the minimum, maximum and most likely cost of the project: £2, £7 and £12, it also revealed that if we were to accept a contractor’s offer to deliver the garage for £6, we would know that we really ought to set aside a contingency fund of up to £6 more to accommodate the potential for final costs to increase up to £12. We now know the size of the overall financial reserve needed to cover the total of the risk exposure of the project and the estimate of its cost. In gaining this insight we also found what was needed to avoid incurring the £12 cost – make sure the base is level.

  5. If we wanted to give the contractor an incentive to deliver the project for less in exchange for giving them an opportunity to increase their profit, we can see to what extent it would be reasonable to do so. Setting a target of £2 for the garage might be somewhat over-ambitious, but £4 or £5 would be fair, in exchange for a bonus of half of the £1 or £2 that may then be saved.


By taking into account the risk opportunity spreads, we come to know the minimum, the most likely and the maximum cost of the project. We also have an analysis-based idea of the size of the contingency fund we will need to cover our exposure to the risks of the costs increasing. What’s more, we know the possibilities for cost savings that could be reasonably incentivised. This set of five insets has always struck me as a revelation, and moreover it is one that provides pointers to where management action can be deployed to good effect. I think the addition of the risk and opportunity analysis gives an overall result that is so much more insightful than simply adding a percentage mark-up to £6 in order to compensate for the historically optimistic estimates of garage projects.

When presented, this simple example above has all too often been immediately challenged by students who point out that there is must be more than just two risk– opportunity spreads on a real project, and that surely they are unlikely to be conveniently valued at £1 and £6. This is perfectly true.

On the first point, the number of risk–opportunity spreads on a real project is almost certainly more than two. I have never known it to be less than ten, and I am really only beginning to be confident I have an understanding of a real project when I have found at least twenty. It could indeed be any number: 10, 20, 100. The most I have seen was about 1500, a total that would prompt me to check whether or not the identification of the risks and opportunities had not been somewhat overdone and become a form of self-serving cottage industry, the sanctity of which was a heresy to protest against. Imagine then having a jug of dice, each one representing one risk–opportunity spread of £1 to £6 and repeating the experiment of throwing them, adding them up and plotting the totals (Figure 1.5).

It is easy to see that for every dice to show a one as a result of a shake and throw would be as unlikely as everyone of them showing a six; and that there would generally be as many high values (four, five, six) showing as there would be low ones (one, two, three). Therefore, we would expect the total to be somewhere in the middle, between the maximum number of ones and the maximum number of sixes. The minimum, most likely and maximum pattern of the two dice case would therefore emerge with the many-dice case, albeit stretched over a wider range but still with a peak in the middle.

And if the dice were not all one to six, but designed like those in Figure 1.6?

Figure 1.5 Many dice


Figure 1.6 Many multi-faced dice


Unless the dice are loaded, the chances of any one face on a dice being shown is the same as for any other face. The idea that all dice should show their minimum values on a single throw, or equally, their maximum values, is increasingly unlikely in proportion to the total number of dice in the bucket. The more dice there are, the less likely this is, just as is the case for a bucket full of one to six dice. The general outcome is not altered by having multi-faced dice.

However, the most likely value is not as readily apparent. It all depends on what numbers are written on the faces of the multi-faced dice. It is usually there, however, and a few hundred ‘shake, throw and add-up’ cycles will reveal it.

Generally, no matter what the risk–opportunity spreads are in a project, either in terms of their number or their values, when we make an attempt to measure their combined effect we will find a minimum, maximum and most likely cost. We will be able to assess the overall risk exposure and the potential for cost reduction.

If you will permit another smoothing over of the actuality, the final result looks something like this (Figure 1.7). I have fitted a smooth curve over a typical histogram and labelled it with the five pieces of information given in the previous paragraph.

Now let me put a question to you. If you were going to manage or fund a project, a new hospital say, upon which your personal reputation and well-being (your self-esteem, your choice of house, your options for your children’s education, where you take your holidays and so on) depended, and you could only buy one set of information to help you secure a successful outcome to the project, which would you choose, the cost estimate or the above curve?

To me the answer is obvious: the curve. The next figure (Figure 1.8) shows a situation where the single valued estimate of the project, the spot cost, lies within the spread of the curve. To choose the spot cost as the project budget would be to take a significant chance (shown by the shaded area to the right of the figure) that at some point in the future it would become overspent. Though it is a contrived scenario, in this case the probability of an overspend would be about 75 per cent – the portion of the area under the curve to the right of the spot cost. Further, because the analysis that had gone into researching the curve would not be available to you had you chosen the spot cost analysis instead, you would probably have little evidence available to help put together a strategy to preempting it.

Figure 1.7 Spread cost curve


If, however, you would choose the curve over the spot cost, and were consistent in this choice over a career’s worth of projects, then there is a slightly startling implication: you and I would be accepting that there is no such thing as the cost of a project. There is instead its cost function (cost curve if you like).

To be more precise, there is no such thing as the cost of a project that is not qualified by the probability of its occurrence. Presuming that your choice would be the same as mine, the curve over the spot, we will be far better informed. As a result we will know what funding we need, and just as importantly why we may need it.

Professor Flyvbjerg was right in his retrospective findings. His curves are the cost curves for different types of projects. They show the effect of the risk opportunity spreads a posteriori, which is to say after the event. I would now like to explain how to calculate the curve for any specific project a priori, in advance.

I hope I have made a case for risk analysis to be used to estimate the funding a project will need. I would now like to propose how to do it.

Figure 1.8 Spread cost curve with spot cost


Using Risk Analysis to Determine the Funding Needed for a Project

I think it important I propose a theory for the use of risk analysis to calculate the funding needed by a project so that it is capable of falsification and thus open to improvement or replacement.

The basic theory for estimating the cost of a project is simple and intuitively obvious to begin with. A conception of what the project has to make is broken down by the engineers and managers involved into its constituent parts, down to a practicable level usually set by the rational units of purchase of these parts. The quantity required of each is then calculated. If the component is bulk material, say sand, and the unit of purchase is cubic metres then it is the numbers of these that is totted up. If it is computer hardware then is probably circuit boards. The unit cost of each item, and the quantity required, are multiplied together to give an overall item cost. All the item costs are added up to give a total cost. So far so trivial, and I feel slightly embarrassed writing it down.

The next stages of basic theory are more difficult to apply and call for more experience and intellect in order to produce valid answers. Validation needs to be considered as something distinct from verification. The latter is assurance that the calculations were correct, whereas as validation is assurance that they were the right ones to do. They are not the same thing, and it is validity and not verification that is the issue here. First, it needs to be determined what types of labour are required, as well as how long it will take to complete such work to an acceptable standard. Secondly, what facilities and services have to be provided for them, be it an office or a production plant? Thirdly, the costs associated with the logistics of getting everything to the client’s desired location need to be determined. Finally, it is a matter of commissioning, testing, training and trialling as what has been delivered is brought into service.

The steps in the basic theory of estimating are as follows:

  • Specify and cost the materials.
  • Specify and cost the labour.
  • Specify and cost the facilities and supporting services.
  • Specify and cost the logistics.
  • Specify and cost the commissioning.


Experienced estimators will almost certainly feel the total cost calculated thus far may not be quite enough. They will add additional money to cover the costs of overcoming emerging adverse situations which they know their client would reasonably expect them to have foreseen, for example, that another project happening at the same time has cornered the market in a type of material. The estimators will probably have caveated their calculations with the assumptions and exclusions to forewarn the client that not everything is covered by them.

This basic theory seems intuitively right. If all the costs of all the items and services needed are known, then what else is there? The theory seems not to have worked for projects of the scale analysed by Flyvbjerg and Morris. It might work for simple projects, but it does not seem to me to scale-up awfully well to projects that are large, complex and protracted in terms of timescales. That they are paid for by one team of people but delivered by another also seems to be a persistent characteristic. Somewhere along the path of increasing project complexity, the basic theory seemingly fails. Further, no additions to the costs of most projects with published historical records would seem to have been adequate (presuming that estimators have always added mark-ups to cover unforeseen eventualities, and that the Flyvbjerg and Morris data are not the result of a fluke from projects that had no additions made to their estimates by incautious estimators, an endangered species if it not already extinct).

Something is missing from the basic theory if it is to be used to inform the funding decision and this is the risk. If we were a contractor engaged to design and build a shed, and we found a disused well when digging the footings, we would go to the sponsor of the garage project and ask for an increase in our funding to fill in the well, or bridge it, or perhaps relocate to another site. We would do this because it is not our fault the well is there. But it is not the sponsor’s fault either. What is certain however is that if the sponsor wants the garage built, they must pay for something to be done about the well. In commissioning the garage project, the sponsor was therefore taking the risk that there would be no well. But what if the sponsor had looked into what might go wrong with the project by asking the neighbours or even ourselves, the contractors, about our previous experience of carrying out similar works in the locality and so had come to realise there might be wells in the area? Then it would be prudent of the sponsor to obtain not only sufficient funding needed to cover the cost of the works, but also enough to cover the risk there might be a well.

One of the attractive features of the basic theory is that to use it requires only a knowledge of arithmetic, and thus estimates can always be computed by anyone capable of determining what the other information needed may be: which materials, which labour, what facilities, the how of the logistics and the means of introducing into service.

One of the unattractive features of the expanded theory I am going to propose is that it is mathematical in nature, not arithmetical. It requires more than plus, minus, multiply and divide. It uses calculus, statistics and equations and as a result is not as simple without appropriate training. By way of developing the argument, take the economy of the UK. This is surely a much more complex thing to model in terms of numbers and formulae than the business of a corner shop. If we happened to discover that the former was modelled in Her Majesty’s Treasury using nothing more than arithmetic, I think we should all be rather surprised. And so, I have always been mildly incredulous to find that the calculations used for estimating small projects no bigger than a station roof repair worth a few thousand pounds are the same as those used to estimate an entire railway requiring billions of pounds to be spent over many years. The cost of a complex project is surely not just the sum of the products, the quantities and rates of its parts? Surely it is more sophisticated? The method used to calculate the cost function for the simple garage project used not only quantities and rates, it also used formulae (the risk–opportunity spreads), statistics (the probability of things happening, which in that example was 100 per cent but could easily have been another value), calculus (the act of choosing a budget from the curve at a desired level of confidence is equivalent to mathematical integration of an area under the curve up to a boundary), and simulation (the Monte Carlo method of repeatedly putting random numbers into a set of equations that together define a model of a system in order to infer the statistical behaviour of that system – its mean, most likely and maximum costs in the garage project). This is the level of sophistication necessary to estimate complex projects.

The cost of a project is a valid concept only in the context of a completed project. It is one of those things that can only be known after the event, and this can be inferred from the careful avoidance of the use of the term by estimators everywhere in favour of the term ‘estimate’ when talking about projects that have yet to be started or completed. As I hope the trivial garage example in the previous section showed, what can be known about a project prior to its completion is its spread cost curve, one that has yet to start.

Hiding behind the insights of the cost function curve is a difficult decision that someone has to make: how much money ought to be allocated to a project, since there is clearly a range of possibilities with associated levels of confidence in them. Choosing a value from the curve is to decide the funding that will be made available to the project team for delivery of the project. The amount of funding is not the amount desired by a project team to deliver the project. I am sure the team would rather have something more towards the right-hand side of a curve while the funder would rather have something more to the left. The pre-project funding requirement is shown diagrammatically in Figure 1.9.

The height of the stack is the funding required for the project. It has two parts, expenditure and exposure. The expenditure is the part of the funding calculated using cost estimating techniques, and assuming a perfect world in which all of the material and labour components required for the project can be determined in advance, the expenditure is a fixed sum. A funder will probably set a budget for the manager against the expenditure. If they are generous it could be 100 per cent, though usually some is retained and so the manager is ‘incentivised’ by receiving, for example, only 90 per cent.

The other part of the funding is the exposure. This is the element calculated using the risk-analysis techniques seen in the garage project. Assuming an imperfect world, not all of the risks to the project will be known in advance and so the exposure is open-ended, as Figure 1.8 indicates with the increasingly fading blue shading.

The practical solution to the question of how many risks does one add into the exposure calculation is best answered by experience. Risks should be added to until the risk analyst is confident that the exposure is realistic and rationally complete. The only risks that remain to be added are those that are irrational (Mars may invade); or that are trivial matters covered by ordinary business (it may rain tomorrow); or that require incompetence on the part of competent people, processes and systems. This is not to say that people, processes and systems will never go wrong, but rather that they are not deliberately guided to fail. Risk analysts are therefore allowed to assume good intentions from all those who are contributing to the success of a project.

Figure 1.9 The project funding requirement stack


This is a different level of confidence to the one that says the funding is sufficient for the project. It is down to the analyst to make the judgement as to whether the calculations are all in there, and as such it is equivalent to an estimator deciding that the materials and labour items have all been listed in the estimate. My advice is that if the unwanted outcome of a risk has a clear possibility of materialising during the project which is not necessarily confined to the period of its design and construction, but can include the period of use, then funding needs to be provided for the risk.

Once the exposure has been computed, the funder will probably set a contingency fund at a desired level of confidence against it. This fund may be shared with the manager in some form of bottom-tier, top-tier manner, or it may be retained entirely by the funder. It depends how often the funder wishes to be bothered by the manager asking for withdrawals from the contingency fund.

Exposure to risk in practice only becomes a liability when commitment to expenditure is made – no commitment, no risk. However, the value of the exposure is independent of the value of the expenditure. For example, consider systems integration risk – design faults will emerge during the integration of diverse systems into a single entity, such that it will not work as expected. Though the liability for the risk is taken on once a decision to design and build the system is accepted, exposure to it only occurs during the later stages of the project, when the subsystems are assembled into a whole, and commissioning is attempted. The size of the exposure is low to begin with but rises later.

In contrast, expenditure on component subsystems will be higher during the early stages of the project, probably falling to zero in the integration phase. Even though both expenditure and exposure are initiated simultaneously, they vary independently of each other thereafter. Consequently, exposure is not a proportion of expenditure, even though they are often both measured in money.10 [10] To put it bluntly: exposure is not expenditure.

Making it so is a common error in business-case modelling, and one with potentially serious consequences. Which manager would like to discover the size of their available contingency was lower than their current exposure when problems began to emerge, say, during systems integration?

Exposure is also spread over time and has a profile like expenditure, though not one with the same shape (Figure 1.10).

Other analysts and managers will use the terms estimate, cost, budget, assessment, exposure and contingency differently. Often, risk is used to purport the contribution to the cost curve attributable to events that may or may not happen, such as things going wrong. Contingency meanwhile is sometimes seen as the contribution attributable to errors in measuring the quantities, and the variations in prices of things that make up the projects shopping list. Then again some will say risk is beyond what we may have analysed and modelled, that it is the ‘unknown unknowns’, to use the famous words of US Secretary of Defence Donald Rumsfeld. I used to get het up about others using terms I use in different ways to my own preferences, but I am more relaxed about it now. I happily take on board and use terminology that others are comfortable with and translate or interpret as is necessary. It seems wasteful (of money and energy) to debate the differences, as if the business of risk is what we’re being paid to manage, not the risks themselves.

Figure 1.10 The stack with profiles


I do however have one exception to this terminological laissez-faire, and that is the acceptance into the risk lexicon of ‘unknown unknowns’. I doubt they exist in projects that are being planned and developed. All of the project risks I have experienced were available, even if that was only obvious with hindsight. If I had been more diligent in my inquiries, someone somewhere could have told me about it in advance. For project people to use the phrase ‘unknown unknowns’ in their analysis is tantamount to an implicit acceptance of incomplete work.11 [11]

However, I do believe that ‘unknown unknowns’ exist in the external context in which a project is being developed, such as emerging changes in societal or political needs that may render a project superfluous, changes in demographics that will make it a bad investment to complete, or developments in a perceived military threat that would make the deliverables produced by the project inadequate. It would be irrational of the public and their tribunes in the press and political sector to describe a project as having failed when the reason is the emergence of these bigger issues. It is unfair to expect the funders and the project delivery teams to foresee the ‘unknown unknowns’, and it would be wasteful to expect project budgets and timescales to hold contingencies for them just in case. To adapt the words of John Maynard Keynes, if the facts change then I change my project. Projects should therefore be expected to be cancelled or have what they set out to deliver changed when they are part-way through.

What is fair though is a presumption on the part of the funders that the delivery team are competent and experienced enough to be capable of identifying everything that could go wrong with a project this side of flights of fantasy. They have allowed for these in their calculations of the cost curve. All that could go wrong on a project can be identified by due diligence on the part of the analysts researching a cost curve. I always challenge the existence of percentage mark-ups for ‘unknown unknowns’ in risk calculations, suspecting poor or incomplete research. The emergence of a Black Swan may well change the need for a project to dam a river to make a reservoir, but the discovery of porous bedrock should not invalidate its cost curve. As I am wont to say to my colleagues: there is no excuse for not knowing.

Should a funder wish to be more cautious than an exposure calculation indicates, perhaps because they have prior knowledge of changes to economic policy that will occur during the project lifetime, then it would be right for them to adjust the cost curve so that it reflects what they know but which was unknown to the analysts. Equally, they may be considering two or more scenarios, say a 10-year operating life with a change of ownership then, or a 20-year operating life closed by a decommissioning and removal stage. In these circumstances I would always produce a discrete cost curve for each and never conflate the two (Figure 1.11).

The cost estimate calculation is the summation of the products and prices needed by the project. It produces a number. The companion calculation of the risk exposure is a simulation of the overall behaviour of the risks, each of which is modelled mathematically by appropriate formulae. This calculation produces a curve. When the spot cost from the estimate is added to the curve of the assessment, the combined result is the cost curve.

Spread cost curves come in two types: with or without the estimate included. Most analysts calculate and report the latter first, then are asked to prepare and submit the former for inclusion in financial submissions to governments, public sector bodies, financiers and project sponsors. This is because it is more concise to describe the funding requirement as a single value number at a level of confidence than as an estimate and separate risk exposure.

Projects are perceived to have failed when they are over budget and late. It strikes me that a good strategy for success would be to avoid the causes of failure. To decide to fund a project on the basis of both its cost estimate and its risk assessment seems to me a good way to do this.

Figure 1.11 Exposure curve and spot cost giving the cost curve


Submit your own content for publication

Submit content