How Big Things Get Done

The model

Bent Flyvbjerg is a Danish economic geographer who has spent much of his career at the University of Oxford studying very large projects: bridges, tunnels, rail lines, dams, Olympic Games, IT systems and power plants. Over three decades he built a database of more than sixteen thousand projects in over a hundred countries. How Big Things Get Done, written with the journalist Dan Gardner and published in 2023, distils what that database shows.

The headline finding is stark. Of all the projects in the database, fewer than one in ten came in on budget and on schedule, and only about one in two hundred delivered on budget, on time and with the benefits promised. Flyvbjerg calls this the iron law of projects: over budget, over time, under benefits, over and over again. And the overruns are not normally distributed. In many kinds of project they have fat tails, meaning that a substantial minority go catastrophically wrong, by hundreds of per cent. Information technology projects are among the worst.

The model explaining this has two parts.

Why projects fail. Projects get into trouble during the long middle, when the work is being done, but the cause usually lies earlier, in planning that was too fast and too shallow. Flyvbjerg identifies two forces. One is psychological: the planning fallacy, the tendency to base forecasts on the specific case and the best-case scenario rather than on how similar projects have gone. The other is political: strategic misrepresentation, the deliberate understatement of costs and overstatement of benefits to win approval. Both push in the same direction. And once a project is committed, the pressure to start building leaves problems undiscovered until they are expensive to fix. The longer a project runs, the more exposed it is to unexpected events, so slow delivery is itself a cause of overruns.

How projects succeed. Flyvbjerg’s slogan is “think slow, act fast”. Spend time on planning, when change is cheap, and then deliver quickly, when change is expensive. Start from the end goal, asking why the project exists, and work backwards. Plan by iterating, testing ideas with simulations, models and prototypes until the design is proven. Forecast using the outside view, specifically reference class forecasting, which takes the actual distribution of costs and schedules from similar past projects and adjusts from there. Hire an experienced master builder and a team that has worked together. And wherever possible, make the project modular, built from many small, identical, repeatable units, like solar panels or wind turbines, so that each unit is learned from and improved.

Seeing with it

The book contrasts famous cases. The Sydney Opera House became an architectural icon, but it was completed a decade late at many times its original budget, and its architect, Jørn Utzon, left before it was finished. He had begun building before the design was resolved. Frank Gehry’s Guggenheim Museum in Bilbao, by contrast, opened in 1997 on time and on budget, because Gehry’s team had used detailed digital models to resolve the design before construction. Pixar, whose film-making Flyvbjerg describes, works the same way: its films are drafted and redrafted many times as cheap storyboards before expensive animation begins.

The model also explains the success of modular projects. Large solar and wind farms, built from thousands of identical units, rank among the best-performing project types in his data, while nuclear power plants and the Olympic Games, which are unique and must be completed as a whole, rank among the worst.

The site’s idea pages The Inside View and the Outside View and Sunk Costs and the Escalation of Commitment explain the psychology behind these patterns.

Where the model stops

Flyvbjerg’s work has been debated by researchers who argue that some overruns reflect changes in scope that made projects better, and that benefits are harder to measure than costs. The economist Albert Hirschman argued in the 1960s that optimism can be useful, because if people knew how hard projects would be, they would never start them, and creativity under pressure often rescues them. He called this the hiding hand. Flyvbjerg, with Cass Sunstein, examined Hirschman’s argument against data and concluded that the hiding hand more often harms than helps. The debate turns partly on whether one values the projects that would never have been started.

The model is also strongest on projects with clear objectives. Many large undertakings, such as transformations of organisations, have vaguer goals, and the techniques apply to them less directly.

Since it was published

The book became a bestseller and brought Flyvbjerg’s research to a general audience. Its arguments have been taken up in debates about infrastructure, about the costs of building in countries such as the United States and Britain, and about the energy transition, where modular technologies have become cheaper much faster than large, bespoke ones. The site’s essay on enterprise software, The Company Inside the Machine, describes several IT projects whose overruns follow the pattern he documents.

Questions it lets you ask

Why are we doing this project, and what does success look like at the end? What happened to the last ten projects most like this one? Are our numbers based on that record, or on our hopes? Have we tested the plan cheaply before committing to build it? Who has done this before, and are they on the team? Can the work be broken into small, repeatable pieces? And how long will we be exposed to things going wrong?