How Big Things Get Done: The Surprising Factors That Determine the Fate of Every Project, from Home Renovations to Space Exploration and Everything In Between

by Bent Flyvbjerg

In fact, in my experience and in line with the results of behavioral science, people not only tend to naturally look at their projects this way, they tend to exaggerate just how unusual their specific project really is. This is the “uniqueness bias” we encountered in the previous chapter.[9] We all have it. It makes us love our kids. But it’s unfortunate in some circumstances because it blinds us from seeing our project in the second way.

“You’re absolutely unique, just like everyone else.”

Kahneman and Tversky dubbed these two perspectives the “inside view” (looking at the individual project in its singularity) and the “outside view” (looking at a project as part of a class of projects, as “one of those”). Both are valuable. But they’re very different. Although there’s little danger that a forecaster will ignore the inside view, overlooking the outside view is routine. That’s a fatal error. To produce a reliable forecast, you need the outside view.

But there’s a way around them. You just need to start over with a different perspective: See your project as one in a class of similar projects already done, as “one of those.” Use data from that class—about cost, time, benefits, or whatever else you want to forecast—as your anchor. Then adjust up or down, if necessary, to reflect how your specific project differs from the mean in the class. That’s it. It couldn’t be simpler.

It takes from five to ten years to complete transportation projects, and people have been building them for centuries. How can it be less risky to decommission nuclear power when it takes much longer and we have almost no experience doing it? I agreed. That made no sense.

But as noted in chapter 1, my analysis revealed that only a minority of the many project types in my database are “normally” distributed. The rest—from the Olympic Games to IT projects to nuclear power plants and big dams—have more extreme outcomes in the tails of their distributions. With these fat-tailed distributions, the mean is not representative of the distribution and therefore is not a good estimator for forecasts. For the most fat-tailed distributions, there isn’t even a stable mean that you can expect outcomes to cluster around because an even more extreme outcome can (and will) come along and push the mean further out, into the tail toward infinity. So instead of good old regression to the mean, you get what I call “regression to the tail.”[22]

you’re a professional at a large organization, you should do better than this rough-and-ready approach. You need to get serious about gathering enough data to allow you to statistically analyze the distribution and determine if it’s normal or fat-tailed. If it’s normal or near normal, do a reference-class forecast using the mean. This would still give you an approximately 50 percent risk of a small cost overrun. If you want to reduce this risk further, add a 10 to 15 percent contingency (reserve), and you’re done.[24]

Providing such contingencies would not be budgeting; it would be blowing up the budget. So what can you do about the tail? Cut it off. You can do that with risk mitigation. I call it “black swan management.”

Sure enough, the distribution had a fat tail. High-speed rail is risky business, as we saw in Hong Kong. So we zeroed in on the projects in the tail and investigated what exactly had made each project blow up. The answers were surprisingly simple. The causes had not been “catastrophic” risks such as terrorism, strike actions, or other surprises; they had been standard risks that every project already has on its risk register. We identified roughly a dozen of those and found that projects were undone by the compound effects of these on a project already under stress. We found that projects seldom nosedive for a single reason.

managers know this and keep an archaeologist on speed dial.

There just aren’t that many archaeologists, and, unlike plumbers or electricians, responding to emergency calls isn’t a normal part of their work. So when multiple discoveries overlap, the delays can become severe. And those delays can in turn delay other work. The result is a chain reaction of setbacks, like a line of cars sliding into one another on an icy street. In this way, what starts as one minor fender bender becomes gridlock capable of derailing the whole project.

Interestingly, early delays are not seen as a big deal by most project leaders. They figure they have time to catch up, precisely because the delays happen early. That sounds reasonable. But it’s dead wrong. Early delays cause chain reactions throughout the delivery process. The later a delay comes, the less remaining work there is and the less the risk and impact of a chain reaction.

If they had, they would have spent time thinking about live events. How do they fail? One common way is equipment failure. Mics don’t work. Computers crash. How is that risk mitigated? Simple: Identify essential equipment, get backups, and make contingency plans. That kind of analysis is dead easy—but only after you have shifted to the outside view.

We also noted that procurement and delivery were often accompanied by delays as lower-level employees at MTR contacted lower-level employees at the supplier. We advised that such decisions be pushed up the ladder, so that the CEO of MTR would contact the CEO of the supplier—a remarkably effective way to accelerate response time, we found.

But if a project is falling behind, managers don’t want to wait until the next milestone arrives before they are alerted to the delay. They need to know and act as quickly as possible. Our data were so detailed that we could make a further set of subforecasts, so we invented “inchstones.” And we specified in detail beforehand who would be responsible for what. If MTR started falling behind under the new schedule, managers would know immediately, and they would know who must act, so no time would be wasted. With the Hong Kong government, we developed the inchstone approach into a general methodology based on artificial intelligence, which today is used on other Hong Kong projects and can be used on any project anywhere.[32]

we can put reference-class forecasting and risk management into the toolbox, along with experience, Pixar planning, and thinking from right to left. These are the essential tools for thinking slow in planning before acting fast in delivery.

“What you call the costs are not the full costs,” I continued. “Yes, the Sydney Opera House cost a large amount of money, far more than it should have. But the full cost of that building includes all the other architectural treasures that Jørn Utzon never built. Sydney got its masterpiece, but cities around the world were robbed of theirs.”

For him, careful planning doesn’t obstruct creativity; it enables it.

“what I’m constantly trying to do is slow things down,” he says. Take time to develop ideas. Take time to spot and correct problems. Do it on the drafting table, not the construction site. “If you slow things down sometimes and you take a second and a third look, you end up making less mistakes,” he says. “And that means [the project gets done] faster.”

Heathrow was, and is, one of the world’s busiest airports, and the new terminal—Terminal 5 (T5)—would be an immense addition. The main building would be the largest freestanding structure in the United Kingdom. Add two more buildings, and T5 would have fifty-three gates and a total footprint of 3.8 million square feet.

When we think of airports, we imagine runways and large buildings like these. In reality, however, airports are complex agglomerations of infrastructure and services, like little cities. So T5 also required a long list of other systems—tunnels, roads, parking facilities, rail connections, stations, electronic systems, baggage handling, catering, safety systems, and a new air traffic control tower for the whole airport—that had to work together seamlessly. All that would be built between two runways with the existing central terminal area at one end and a busy freeway at the other. And the airport could never shut down.

The word deadline comes from the American Civil War, when prison camps set boundaries and any prisoner who crossed a line was shot.[5]

You will recall that most project types are not only at risk of coming in late, going over budget, and generating fewer benefits than expected. They are at risk of going disastrously wrong. That means you may not wind up 10 percent over budget; you may go 100 percent over. Or 400 percent. Or worse. These are black swan outcomes, and the project types at risk of them are called “fat-tailed.” They include nuclear power plants, hydroelectric dams, information technology, tunnels, major buildings, aerospace, and many more. In fact, almost all the project types in my database are fat-tailed. But not quite all.

five project types that are not fat-tailed. That means they may come in somewhat late or over budget but it’s very unlikely that they will go disastrously wrong. The fortunate five? They are solar power, wind power, fossil thermal power (power plants that generate electricity by burning fossil fuels), electricity transmission, and roads. In fact, the best-performing project types in my entire database, by a comfortable margin, are wind and solar power.

Operations experts call this “negative learning”: The more you learn, the more difficult and costly it gets.

It is a giant understatement to say that that turn of events could not have been predicted in 1983. But when delivery takes decades, the unpredictable becomes inevitable.

Modularity is a clunky word for the elegant idea of big things made from small things. A block of Lego is a small thing, but by assembling more than nine thousand of them, you can build one of the biggest sets Lego makes, a scale model of the Colosseum in Rome. That’s modularity. Look for it in the world, and you’ll see it everywhere. A brick wall is made of hundreds of bricks. A flock of starlings, which moves as if it were a unitary organism, may be composed of hundreds or thousands of birds. Even our bodies are modular, composed of trillions of cells that are themselves modular. There’s an evolutionary reason for this ubiquity: In survival of the fittest, the “fittest” is often a module that is particularly successful in reproducing itself.[9] The core of modularity is repetition. Put down one Lego block. Snap on another. And another. And another. Repeat, repeat, repeat. Click, click, click. Repetition is the genius of modularity; it enables experimentation. If something works, you keep it in the plan. If it doesn’t, you “fail fast,” to use the famous Silicon Valley term, and adjust the plan. You get smarter. Designs improve. Repetition also generates experience, making your performance better. This is called “positive learning,” as we saw earlier. Repetition rockets you up the learning curve, making each new iteration better, easier, cheaper, and faster.

Manufacturing in a factory and assembling on-site is far more efficient than traditional construction because a factory is a controlled environment designed to be as efficient, linear, and predictable as possible.

As I mentioned in the previous chapter, this process—properly known as “design for manufacture and assembly”—is a big part of the explanation of the success of Heathrow’s Terminal 5.

The technical term for this property is “scale free,” meaning that the thing is basically the same no matter what size it is. This gives you the magic of what I call “scale-free scalability,” meaning you can scale up or down following the same principles independently of where you are scalewise, which is exactly what you want in order to build something huge with ease.

The difference between cheap-and-ugly modular and these projects is imagination and technology. To fully unlock modularity’s potential, to see how astonishingly versatile it can be, we need to “think different,”

What is our basic building block, the thing we will repeatedly make, becoming smarter and better each time we do so? That’s the question every project leader should ask. What is the small thing we can assemble in large numbers into a big thing? Or a huge thing? What’s our Lego? Explore that question, and you may be surprised by what you discover.

Space has long been dominated by big, complex one-off projects, and priced accordingly, with NASA’s James Webb Space Telescope—$8.8 billion, 450 percent over budget—just the latest example.

Subways would seem to be an even harder case for modularization, but when Madrid Metro carried out one of the world’s largest subway expansions between 1995 and 2003, it leaned on modularity in two ways. First, the seventy-six stations required for the expansion were treated like Lego, with all sharing the same simple, clean, functional design. Costs plunged, and speed of delivery soared. To amplify those effects, Madrid Metro shunned new technologies. Only proven technologies—those with a high degree of “frozen experience”—were used.

But by turning cargo into Lego, it made shipping extremely modular and cost effective. The stacks on ships got taller. The ships got bigger. The transfer from one mode of transportation to another got quicker. The speed and ease of transporting goods soared, while costs plunged so steeply that the economics of production and distribution worldwide were changed.

Only five project types—solar power, wind power, fossil thermal power, electricity transmission, and roads—are not fat-tailed, meaning that they, unlike all the rest, do not have a considerable risk of going disastrously wrong. So what sets the fortunate five apart? They are all modular to a considerable degree, some extremely so.

Fossil thermal power? Look inside a coal-burning power plant, say, and you’ll find that they’re pretty simple, consisting of a few basic factory-built elements assembled to make a big pot of water boil and run a turbine. They’re modular, much as a modern truck is modular. The same goes for oil- and gas-fired plants.

The pattern is clear: Modular projects are in much less danger of turning into fat-tailed disasters. So modular is faster, cheaper, and less risky. That is a fact of immense importance.

In China, if the national government at the highest levels decides that a project is a priority, obstacles are eliminated and the project gets done.

Extreme weather events have always happened, but climate change is making them more frequent and more extreme. And they will keep getting more frequent and more extreme. The only question is how much more.

To halt climate change before it becomes catastrophic, most of the nations of the world have committed to a target of “net zero by 2050,” meaning that by midcentury they will emit no more greenhouse gases into the atmosphere than they take out.

It found that fossil fuels, which today account for four-fifths of the world’s energy production, could provide no more than one-fifth in 2050. Replacing them would require a vast increase in electrification—our grandchildren will encounter gas stations only in history books—and an explosion in the production of electricity by renewable energy sources.

Wind power must grow elevenfold. Solar power must grow a mind-boggling twentyfold.[37] Investment in renewable energy must triple by 2030, mostly delivered as hundreds, if not thousands, of large-scale, multibillion-dollar wind and solar farms. New nuclear and new hydro may have a role to play for the 2050 deadline, but for 2030 they have already proven too slow.

Denmark doesn’t have lots of uninhabited land and people don’t want to live in the shadow of wind turbines. In the late 1990s, a visionary Danish minister of the environment, Svend Auken, told companies seeking permission to build coal-powered generators that they could go ahead on the condition that they also build two of the world’s first offshore wind farms. They did. One worked, the other was a mess. Both gave the owners experiri. It was a start.

What they didn’t appreciate was the extreme modularity of offshore wind farms. Assemble four Lego—foundation, tower, head, blades; click, click, click—and you have a turbine that can start generating electricity immediately. Assemble eight to ten turbines and wire them together, and you have a “string” that can be connected to a substation that feeds into the national electricity grid. It, too, can start delivering as soon as it is assembled. Put together a few strings, and you have a wind farm that is operational on day one. Repeat, repeat, repeat. It can scale up as much as you like, with each iteration pushing everyone up the learning curve.

The most dramatic was growth in the size of turbines. Whereas a turbine in 2000 might have been a little taller than the Statue of Liberty and able to power 1,500 homes, a turbine in 2017 was almost double that height and capable of powering 7,100 homes.

Over the same ten years, the percentage of Denmark’s electricity generated by fossil fuels fell from 72 percent to 24 percent, while the share contributed by wind power soared from 18 percent to 56 percent.[43] Some days, Danish wind turbines produce more electricity than the country can consume. The surplus is exported to neighboring nations.

It’s what happened for film in Hollywood in the 1920s and for tech in Silicon Valley in the mid–twentieth century. Jutland is now the Silicon Valley of wind energy—which is striking for a country whose population is just over half that of Los Angeles County.

you want to start an avalanche big enough to change the world, government may have to help push the first boulder.

“Think slow, act fast” is an example of a heuristic. Experts and laypeople alike use them when making decisions under uncertainty.[2]