How big things get done

Bent Flyvbjerg and Dan Gardner’s guide to the factors behind every successful project, from home renovations to space exploration.

‘Getting Big Things Done’ highlights that our own behavioural and cognitive biases are the greatest threat to our projects. A few key themes emerge:

1. Hire a master-builder and their people, to build a great team Employ leaders with a proven track record in managing similar projects. Their main job is to pick the team (likely one they’ve assembled before). Big projects routinely do not make maximum use of experience due to "uniqueness bias." Projects are (incorrectly) seen as unique, one-off ventures that have little to learn from earlier projects. Ambition urges our project to be the first, the biggest, the tallest, the longest, the fastest. Each of these superlatives makes a project ‘unique’, and so undermines its ability to learn from experience. Humans tend to underappreciate how deeply experience can enrich judgement, planning and leadership. Like knowing how to ride a bike, this ‘tacit knowledge’ often cannot be put into words; it can only be felt, and learnt by trying, failing, and trying again - the process of learning by experience. Heathrow’s Terminal 5 expansion was successful because it paid the high cost needed to put together a team that had worked together before many times. It had a clear identity and purpose, with shared financial risks / incentives, and an environment of psychological safety.

   
   

   

   
   

2. Ask ‘why?’ and think right to left

   To prevent the big picture being lost in detail or crises, always start with the fundamental question of why.

   
   

   ‘Backcasting’ is a practice which implements this - working back from the end goal to identify the necessary interim steps. One example of backcasting is Amazon’s practice of creating press releases and FAQs as the first step of a new product, not the final one. Writing the end result in plain language, meeting to read it individually and then discussing it line by line, forces clarity of thought from the beginning.

   Backcasting also helps you to ‘Say no and walk away’ when required. This means recognising - as early as possible - when a project is not viable and having the courage to abandon it. This prevents wasting resources, and allows for better allocation of efforts. In a less drastic way, backcasting brings clarity which enables you to say ‘no’ to everything that won’t accomplish the end goal.
   

   

   
   

3. Think slow, act fast ‘Planning’ is cheap, delivery is expensive. Engage in prototyping to limit the amount of time spent in executing the project. Limiting execution time limits the chances of an unexpected risk materialising, and limits the extent of its cost.
   

   Good planning includes avoiding the wrong kinds of optimism, and avoids rushing to conclusions.

   ‘Optimism is useful for many endeavours, but it is widespread, pervasive and costly’ as Daniel Kahneman observed. Optimism is not only dangerous for pilots(!), but projects. Glossing over risks and inflating benefits may secure initial buy-in to a project, but will likely doom its outcomes. Equally, rushing to start work on a preferred option often means planners do not stop to consider better alternatives. ‘Bias for action’ should only apply to reversible decisions!
   

   Instead of jumping to apparently obvious - and wrong - conclusions, the mark of good planning is “The range and depth of the questions it asks and the imagination and the rigour of the answers it delivers”. Good planners ‘commit to not commit’ until enough analysis has been done.
   

   ‘Planning’ can include rapid, low-cost prototyping: ‘Pixar planning’, so-called after the cartoon studio’s extensive use of storyboarding prior to committing to expensive production. While this planning is somewhat labour intensive, its cost pales in comparison to the expense of making a bad film. The aim is to provide just enough of a prototype (the ‘maximum virtual product’ i.e. still in ‘planning’) to prove the crucial assumptions about the project. In Pixar’s case, the storyboards help prove whether the audience will be carried by the characters and storyline of a proposed movie. This method taps into the reality that humans are poor at getting things right the first time, but great at tinkering. It is summed up by experiri - a Latin word for ‘to try’, ‘to test’ or ‘to prove’. In other words, this iterative approach is built around ‘experimenting’ and ‘experiencing’. Whether using an architect's model or a project simulation, prototyping helps us learn from the mistakes we might have made. Prototypes allow us to ask questions that uncover ‘the illusion of explanatory depth.’ In other words, if you can’t explain the answers to key questions in a simulation, you don’t understand the project you are planning sufficiently.

   The final advantage of remaining in planning mode for an extended period is the creativity benefit. Psychologists have shown that stress has a largely negative effect on creativity, particularly in two circumstances: when we feel that the situation is mostly beyond our control and when we feel that others are judging our competence. As such, a project in trouble is exactly the sort of situation in which we can expect stress to hamper creativity. Imaginative leaps belong in planning, not delivery, where stakes and stress are lower, allowing more freedom to wonder, try, and experiment.
   

   

   
   

4. Find your Lego Identify a small, manageable, repeatable component that can become the building block toward your end goal. These get better, faster and cheaper over time, and allow for flexibility, easier management, and the ability to scale up efficiently. This mirrors how nature itself functions.

   Of all the types of projects researched by the author, the best-performing types - by a considerable margin - are wind and solar power. They may come in somewhat late, or over budget, but it's very unlikely that they will go disastrously wrong. This is because they rely on modularity. Instead of complex, bespoke parts, they focus on many small, simple parts built up (or reassembled) quickly over time, maximising experimentation and experience, delivering value much earlier (from the completion of the very first module), and with greater resilience to ‘black swan’ events which might occur during the delivery period.
   

   The author uses the example of building a major part of an entire national school system in Nepal, by focusing on the creation of many simple, individual classrooms: “In some instances, a single classroom was the whole school. In others, putting a couple of classrooms together made a school. In still others, three or more classrooms made a school. Assemble enough classrooms into enough schools, and you have the schools for a district. Do that for all districts, and you have a national school system.”
   

   These were functional, high quality, easily transportable, and earthquake proof, but with only three main designs. They also demonstrated the property of being "scale free", a ‘fractal’, meaning that the thing is basically the same no matter what size it is, or how many modular components it has.
   

   

   
   

5. Take the Outside View and Watch Your Downside: Use reference class forecasting to compare your project with similar ones that have been completed in order to set realistic expectations, and focus on proactively mitigating ‘black swan’ risks.

   
   

   It's often assumed that if delivery fails, the problem must lie with delivery. In fact it frequently lies with the expectations set by poor forecasting.

   
   

   Basing forecasting on a natural ‘anchoring and adjustment’ approach - where an initial estimate (the anchor) is refined based on additional information - is problematic. As psychologists have shown in countless experiments, final estimates made this way are biased toward the anchor. That means the quality of the anchor is critical, but it’s easy to choose a bad one (see Kahneman and Tversky using a blatantly irrelevant "wheel of fortune" to generate a random number which incorrectly anchored participants’ future estimates). Furthermore, spotting potential risks to adjust your forecast won't get you a foolproof estimate - there are always "unknown unknowns" that cannot be foreseen or used for adjustment purposes.

   People tend to exaggerate just how unusual their specific project really is - the "uniqueness bias" above. However whatever sets it apart, a project apart shares characteristics with others. Using cost / time / benefits data from those other projects - your project’s reference class - allows you to create a more reliable anchor. This is then adjusted to reflect how your specific project differs from the average in that reference class, although this should be done sparingly to avoid bias creeping back in.

   This allows ‘unknown unknowns’ to be forecasted too: the data for the projects in the reference class reflect everything that happened to those projects, including any unknown surprises. We may not know precisely what those events - or their effects - were, but, because the data for the reference class reflect how common and how big the ‘unknown unknowns’ really were for those projects, your forecast will also reflect those facts.

   You can also use a reference class from another type of project to inform your estimate if you don’t have a specific reference class you can draw on. The author used the example of leveraging a reference class for transportation infrastructure to help inform a nuclear decommissioning project: whatever the right estimate was for (unusual, complex) nuclear decommissioning, it was definitely going to be higher than the reliable figures presented for more routine transportation projects.
   

   The authors’ analysis showed that only a minority of project types have 'normally' distributed risk. The rest have more extreme outcomes in the 'tails' of their distributions. With these ‘fat-tailed’ distributions, the average is not representative of the distribution and therefore is not a good estimator for forecasts. As such forecasts should move from reflecting a single outcome ("The project will cost X") to forecasting a spread of outcomes ("The project is X percent likely to cost more than Y"), using the full range of the distribution. In a typical fat-tailed distribution in project management, about 80 percent of the outcomes will make up the body of the distribution. For that portion of the distribution, you can protect yourself with affordable contingencies that will fit into a budget. But the tail outcomes — the "black swans" — cover about 20 percent of the distribution. Contingency might have to be 300, 400, or 500 percent over the average cost: budgeting for this is prohibitive.
   

   Instead, “cut the tail off”, mitigating these ‘black swan’ risks, by:
   

   1. Conducting exhaustive planning that enables swift delivery, narrowing the time window in which ‘black swans’ could materialise,
      

   2. ‘Building bridges before you need them’ with wider partners and other stakeholders, maintaining good relationships and collaboration which can be relied on in the event of something going wrong,
      

   3. Studying what the black swans typically are for your kind of project. Projects seldom nosedive for a single reason, but are undone by the compound effects of ‘black swan’ events on a project already under stress,
      

   4. Strenuously avoiding early delays, as - far from being easily caught up - these cause chain reactions throughout the delivery process, and
      

   5. Measuring progress in ‘inchstones’ not ‘milestones, so that, if a project is falling behind, managers don't have to wait until the next major progress marker before they are alerted to problems.
      

Close

‘How Big Things Get Done’ is partly written as a ‘call to arms’, with its closing chapters encouraging project managers of all kinds to ‘build climate mitigation in’ to their approaches, as ‘there are few bigger projects we need to accomplish’. Flyvbjerg and Gardner’s work is compellingly written, and practically written, to innumerable areas of work and personal life - it's well worth a read.