How Big Things Get Done

Miles: A couple in Brooklyn decide to redo their kitchen. One small room in a townhouse. They hire an architect, who draws them something plain and details it carefully before anyone lifts a hammer.

Miles: Then, with the walls open, an idea arrives. The house is a building site already, so why not do the jobs they'd been saving for later while the dust is up anyway? One job pulls in the next, and before long the whole place is torn apart.

Miles: It finishes eight times later than planned and nearly four hundred percent over budget. Two determined people, a detailed plan, one small room.

Miles: Today: How Big Things Get Done, by Bent Flyvbjerg and Dan Gardner. In two sentences: nearly every big project runs over, and most of the damage is done before anyone breaks ground, in a planning stage that got rushed so the digging could start. Plan slowly, deliver quickly, and later Nora's going to argue that plenty of neurodivergent people have worked that way all along, whenever somebody let them.

Miles: Sixty Books. Presented by Stoar Consulting and Coaching.

Miles: Book forty one. You're listening to Sixty Books, and I'm Miles. One management title an episode, sixty drawn from the best of the last hundred years, and each one put to a pair of tests: what it gives leaders and teams who think differently, and what still stands once AI sits in on the work. Nora has the research, and this one arrives with sixteen thousand projects behind it.

Miles: Nora, we closed Noise on the idea that a scattered estimate becomes a scattered budget. You owe me a bill.

Nora: I do, and the bill corrects us slightly. If cost estimates were only scattered, they'd miss high about as often as they miss low. Across this database they don't. They miss in one direction, toward too cheap, project after project. So in last episode's vocabulary, this book is mostly about bias.

Miles: Then where did the scatter go?

Nora: Out to the far end. Most projects overrun by a modest amount, and a handful overrun by sums nobody's range of guesses would ever have reached. That's chapter six.

Miles: And a returning name on the cover.

Nora: Dan Gardner, who wrote Superforecasting with Tetlock and comes back here beside an Oxford economist. Same writer, and this time a database instead of a tournament.

Miles: Give me the story the argument leans on.

Nora: Sydney, January of nineteen fifty seven. A Danish architect, Jørn Utzon, thirty eight years old, wins an international competition to design an opera house. The estimate is seven million Australian dollars, with the doors opening in January of nineteen sixty three.

Miles: And it opened when?

Nora: Nineteen seventy three. Ten years late, at a hundred and two million. In Flyvbjerg's data that's fourteen hundred percent over budget, and Guinness lists it as the largest proportional cost increase on record for a building.

Miles: How does a building get that far away from its plan?

Nora: By starting before there was a plan. The state government pushed for work to begin early, worried that funding or public opinion might turn against it, so construction began in nineteen fifty nine, before Utzon had finished the final designs. The podium went in first, and its columns turned out to be too weak to carry the roof, so they had to be rebuilt. Utzon left the project in February of nineteen sixty six, partway through, and never went back to Australia to see it finished.

Miles: And yet it's on every postcard.

Nora: Which the book concedes. Flyvbjerg's point is that the bill landed elsewhere, and he's written that the overrun and the row that followed it destroyed Utzon's career. Then the book sets it beside a museum in Bilbao, by Frank Gehry, finished on schedule and a little under budget. The difference, on the book's account, is where the thinking happened.

Miles: So the kitchen and the opera house tell one story.

Nora: One story at two sizes. Though, in fairness to the couple in Brooklyn, nobody had to dynamite any part of their kitchen.

Miles: A low bar.

Nora: Yeah. A low bar, and the opera house didn't clear it. The authors write that part of what was built in Sydney had to be blown up and done again.

Miles: The authors.

Nora: Bent Flyvbjerg is a Danish economic geographer, with a doctorate in urban geography and planning. He was a professor of planning at Aalborg, then the first BT Professor at Oxford's business school and the first to hold its chair in major programme management, a post he retired from in twenty twenty one. He now holds a professorship in Copenhagen, at the city's university for information technology, and the publisher notes a knighthood from the Queen of Denmark and advisory work on more than a hundred projects of a billion dollars or more.

Miles: And the database.

Nora: Started small. In two thousand two he and two colleagues published a study of two hundred and fifty eight transport projects, worth about ninety billion dollars, in the Journal of the American Planning Association. They concluded that the estimates used to decide whether to build were systematically misleading, and that error couldn't explain the gap. That collection has grown to more than sixteen thousand projects, in a hundred and thirty six countries.

Miles: And Gardner's part?

Nora: A former newspaper columnist and investigative journalist, now an honorary senior fellow at the University of Ottawa's public policy school. The book came out from Currency in February of twenty twenty three. An introduction, nine chapters, a coda of eleven rules of thumb, and an appendix that's nothing but base rates for cost risk, sorted by type of project.

Miles: Walk the chapters with me, in order.

Nora: The introduction is called California Dreamin', after the state's high speed rail line, which the publisher describes going from thirty three billion dollars to a hundred billion without reaching all the places it promised. Then the headline numbers, from the database. Just under forty eight percent of big projects finish on budget. Eight and a half percent finish on budget and on time. And on budget, on time and delivering the benefits promised: half of one percent.

Miles: One in two hundred.

Nora: Yeah. One in two hundred. Flyvbjerg calls it the iron law of megaprojects: over budget, over time, over and over again.

Miles: Chapter one is the title phrase.

Nora: Think slow, act fast, a deliberate turn on Kahneman's Thinking, Fast and Slow. Planning is cheap: you can try something, scrap it and try again for the price of a few hours. Delivery is expensive, and every day of it is another day for something nobody planned for to go wrong. So stretch the cheap part and squeeze the dear part. The failure mode, in the authors' image: a project that begins as an excited sprint turns into a long, desperate slog.

Miles: Then why does anyone sprint?

Nora: Chapter two, the commitment fallacy. People commit to a project before they understand it, and once delivery starts, they're locked in. The book also takes on why first numbers come in low, and it calls one cause strategic misrepresentation, a budget pitched to win approval. It quotes a former mayor of San Francisco, who wrote in a newspaper column in twenty thirteen that for civic projects the first budget is really just a down payment.

Miles: In print.

Nora: In print, under his own name. Chapter three turns the direction of planning around. Think from right to left: start at the far end, with the result you want and why you want it, and work back toward the means. That's the kitchen's lesson. The authors argue that asking what the project was for, at the very start, would have surfaced the couple's goals before the walls came down.

Miles: And Pixar gets a chapter.

Nora: Pixar Planning. A film there goes through outlines, scripts and storyboards, again and again, with screenings and notes in between, and the book says the average film is storyboarded about eight times before the expensive animation starts. Gehry does the same thing for buildings with detailed digital models, so the arguments happen on a screen instead of on a site. Iterate where mistakes cost little.

Miles: Chapter five.

Nora: Experience, which the book thinks gets badly undervalued. The Empire State Building is its showcase. Construction began in March of nineteen thirty and the building opened on the first of May, nineteen thirty one, about four hundred and ten days. It came in at just under forty one million dollars against a fifty million budget. The authors credit seasoned builders, proven designs and technology that had already done the job elsewhere.

Miles: And chapter six is your far end.

Nora: It is. So You Think Your Project Is Unique? The answer's no. Your project is one of those, a member of a class, and the class has a record. That's reference class forecasting, an idea from Kahneman and Tversky that Flyvbjerg turned into a method. Instead of building your estimate up from your own plan, you start with the recorded outcomes of comparable projects and adjust from there.

Miles: Why would my own plan be worse than strangers' results?

Nora: Because of a study I'd show anyone who manages a calendar. Buehler, Griffin and Ross, Journal of Personality and Social Psychology, nineteen ninety four. Thirty seven psychology honours students estimated when they'd finish their theses. On average they said about thirty four days. It took about fifty five, and only about three in ten finished by their own date. They were also asked for a date if everything went as badly as it possibly could. That averaged about forty nine days.

Miles: So even the worst case was too hopeful.

Nora: By about a week, on average.

Miles: Huh.

Nora: And then the far end, which is the noise from our callback. The book argues that many kinds of project don't overrun on a bell curve, they overrun on a fat tail. Flyvbjerg and colleagues, Journal of Management Information Systems, twenty twenty two, looked at five thousand three hundred and ninety two information technology projects and found the overruns followed a power law: lots of small misses and a long tail of enormous ones. A review in the Wall Street Journal gives the book's figure as eighteen percent of those projects running more than fifty percent over.

Miles: So the average hides the disaster.

Nora: It hides the ones that sink you. Chapter seven asks whether ignorance can be your friend, and I'm saving it for the evidence. Chapter eight is the team: Heathrow's Terminal Five, which the authors say came in on budget and on time because contracts didn't reward blaming the other firm, the workers were listened to, and people could speak up. And chapter nine is the one I'd hand a manager first. What's your Lego?

Miles: Meaning?

Nora: Modularity. Find the small unit you can repeat, and build by repeating it, so the project learns as it goes. The base rates in the appendix make the case on their own. Average cost overrun for solar power, about one percent. Transmission lines, eight. Wind, thirteen. Fossil fuel plants, sixteen. Hydroelectric dams, seventy five. Nuclear power, a hundred and twenty. The more a project is made of one repeatable piece, the less it tends to run over.

Miles: And the eleven rules?

Nora: Hire a masterbuilder, ask why, build with Lego, say no and walk away, among others. And the hardest of its principles to hear: your biggest risk is you.

Miles: Now the evidence check, and you've got two arguments for me.

Nora: Two exchanges, both in journals, and I'll give each side's reading of the record and stop there. The first is about the database itself. In twenty eighteen, Peter Love and Dominic Ahiaga-Dagbui published a critique of that two thousand two paper in Transportation Research Part A. They argued that framing underestimation as either error or lie is a false choice that doesn't fit how infrastructure is bought, and, as the reply to them quotes it, that ideally the data should be made publicly available so the results can be evaluated and reproduced.

Miles: And the reply?

Nora: Later that year, in the same journal: Flyvbjerg and thirteen co-authors, a paper called Five Things You Should Know About Cost Overrun. They argued their critics had measured from the wrong starting point. If you're studying contractors, you measure from the contract. If you're studying decisions, you measure from the budget at the moment of deciding to build, which is what they'd done. They also argued that overrun is fat tailed and rooted in behavioural bias. On the data, they wrote that in practice sources often insist on non disclosure agreements as a condition of letting researchers study their projects.

Miles: So it can't be opened.

Nora: Not all of it, on their account. There was a second round in twenty nineteen: a further critique from Love, Ika and Ahiaga-Dagbui, and a reply from Flyvbjerg's group arguing that the critique contained serious statistical errors. And in twenty twenty four an economist, Timo Valila, in the Journal of Benefit-Cost Analysis, called it the most extensive database of its kind and not openly accessible, and argued that large forecast errors in it don't necessarily mean a project lost society money.

Miles: Where does that leave the base rates?

Nora: Published, in peer reviewed papers by Flyvbjerg's teams, with their methods described. Not yet rerun by outsiders on the same projects, because outsiders can't get all of them. Both of those are true, and neither is a charge against anyone.

Miles: And the second exchange.

Nora: That's chapter seven. In nineteen sixty seven the economist Albert Hirschman proposed the Hiding Hand: we take on hard projects because we underestimate the difficulty, and then human ingenuity rises to meet the problems, so ignorance turns out to be a friend. He drew it from a study of eleven development projects. In twenty sixteen, in World Development, Flyvbjerg tested it on two thousand and sixty two projects and found the Hiding Hand's pattern in about one in five, and the opposite pattern, the planning fallacy, about four times as common. He and Cass Sunstein, from last episode's book, also wrote a paper naming the reverse case, the Malevolent Hiding Hand.

Miles: And who pushed back?

Nora: Lavagnon Ika, same journal, twenty eighteen. He called the test a straw man, argued that a project can fail on budget and still succeed overall, and in a hundred and sixty one World Bank projects found the Hiding Hand more common. Flyvbjerg replied in the same volume that Ika's data were supervisors' recalled ratings on a seven point scale rather than measured costs and benefits, and that his cut point leaned toward finding success. The book takes Flyvbjerg's side. A reader should know a published dissent exists.

Miles: Anything from outside that supports him?

Nora: On the method, yes. Batselier and Vanhoucke, Project Management Journal, twenty sixteen, two researchers outside Flyvbjerg's group, compared reference class forecasting with Monte Carlo simulation and with earned value methods on real project data, and it gave the best forecasts of both cost and time.

Miles: Tier statement.

Nora: How Big Things Get Done is practitioner literature, a book for general readers rather than peer reviewed research. Underneath it sits a larger body of peer reviewed work than most books on this shelf can point to, much of it by the lead author, built on a database outsiders can't fully inspect. Trust the direction, that projects lean toward over and that a few blow up enormously, and treat the precise percentages as the best published estimates rather than settled counts. None of that is a verdict on the people on either side of the argument.

Miles: Our half.

Nora: And here's the claim, which is an argument rather than a finding. Think slow, act fast describes how a lot of neurodivergent people already work, when the setting allows it. A long lead-in that can look from outside like nothing's happening, then a short, intense stretch where most of it gets done. Workplaces usually ask for the reverse: start at once, show steady progress, and keep that up for months. That's the excited sprint into the slog.

Miles: Evidence for the intense stretch?

Nora: Some. Hupfeld, Abagis and Shah, in a journal on attention deficit and hyperactivity disorders, twenty nineteen. A questionnaire built for the study, a preregistered hypothesis, a pilot sample of two hundred and fifty one adults and a replication sample of three hundred and seventy two, with and without ADHD. People with more ADHD symptoms reported more hyperfocus, long stretches of highly concentrated attention, across schoolwork, hobbies and screens. Self report, and a symptom scale rather than diagnoses.

Miles: And autism?

Nora: Two interview studies, so these are accounts, not measurements. Rapaport and colleagues, Autism in Adulthood, twenty twenty four, spoke with twenty four autistic adults about starting and stopping tasks. Their accounts of being immersed matched what researchers call flow and hyperfocus, and they described how hard it could be to balance that absorption with the rest of life. Russell and colleagues, in the same journal in twenty nineteen, ran twenty eight interviews with autistic adults, and hyperfocus was among the strengths they described most often. Their participants were clear that the same trait helped or hurt depending on the setting.

Miles: Which is where your point about being allowed comes in.

Nora: It is. A deep focus with no runway to start it and no clear moment to stop is a liability. Given a long, cheap period to explore and then a short, bounded window to deliver, it starts to look like the book's best practice.

Miles: And the slow half?

Nora: That's where the outside view comes in, and I'd call it an accommodation nobody labels as one. Marx and colleagues, Journal of the American Academy of Child and Adolescent Psychiatry, twenty twenty one, pooled fifty five studies of timing tasks and found people with ADHD less accurate and more variable at judging durations. Those are lab tasks of milliseconds to seconds, not months, so the stretch to project estimates is mine. But a reference class takes the estimate out of anyone's head and puts it on a table of what happened last time. A base rate lets you plan without trusting your own feel for duration.

Miles: Cautions.

Nora: Two. First, nobody has tested Flyvbjerg's method with neurodivergent planners or teams as the outcome. I've joined a book about megaprojects to studies of attention, mostly self report and interviews, and the join is mine.

Miles: And the second.

Nora: Slow in this book doesn't mean waiting. It means active, cheap iteration, the Pixar loop, with drafts and feedback. A long runway with nothing on it is just a delay, and Rapaport's interviews came out of a study of autistic inertia, difficulty starting and stopping tasks. So the slow phase needs visible small loops, and the fast phase needs a fixed edge, an end date, so an intense stretch doesn't run on until it costs the person doing it.

Miles: Fair enough.

Miles: Sixty seconds, read our way. How Big Things Get Done says nearly every big project runs over, that the overruns lean one way and sometimes run to catastrophe, and that the cure is slow, cheap planning followed by short, fast delivery. For anybody leading a mind that works on its own timetable, or working inside one, here's Monday. Before anyone commits, write down why the project exists and what finished looks like, and plan backwards from there. Make the planning a loop of rough drafts and feedback rather than one long document. Estimate from what comparable projects took, not from how long the work feels like it'll take. Then compress delivery into a bounded window with a date at the end of it. And if somebody on your team does their thinking slowly and their finishing fast, judge them on what landed, not on how busy the first weeks looked.

Miles: Now the tools.

Nora: The standing caution on these tools comes at the very end, unchanged, so this time goes to what Flyvbjerg's book suggests.

Miles: Go on.

Nora: The book's whole logic is that planning should be cheap enough to repeat. These tools make a rough draft nearly free, so use them for the loop. Ask for three different versions of the plan, then a critique of each, and run that several times before anyone signs anything. That's the Pixar process on a smaller budget.

Miles: And right to left?

Nora: Have it question you before it writes anything. What's this for, who benefits, what would make it a success a year out. Then draft the announcement of the finished result, and plan back from that page. For the reference class, ask it to list comparable projects and what happened to them, and then check every one against a source you can open, because a base rate nobody can trace is a guess with a decimal point. And for Lego, ask it to break the delivery into the smallest unit you could repeat.

Miles: And the line to close on.

Nora: Spend the machine on the slow half. Every loop it makes cheaper is a loop you can afford before the digging starts, and the book's argument is that this is where projects are won or lost.

Miles: That's How Big Things Get Done.

Miles: Next time: Cal Newport's Deep Work, from twenty sixteen. Six books on deciding when you can't know, and every one of them assumes you've got the attention to do the planning. So we turn to the attention itself, and to who gets to have it.

Nora: Which is the resource this whole stretch spent without counting.

Miles: You brought the receipts on this one, Nora. Thank you.

Nora: Worth every minute.

Miles: On to the next.

Miles: This has been Sixty Books, a production of Stoar Consulting and Coaching. Logan Williamson is a Licensed Professional Counselor in Texas and Louisiana. You heard one book read for the people who run teams and the people who work on them. It wasn't coaching, it wasn't counseling, and it wasn't a word about you.

Nora: The sources behind every claim, and the transcript, are on the episode page. The next book is two days out.

Miles: Two things before you go. The pair of voices you heard are synthetic: there's no Miles and no Nora, neither holds a licence, and nothing said here was about whoever's listening. Scripts are drafted with AI assistance and then edited and approved by Logan Williamson, a Licensed Professional Counselor in Texas and Louisiana, at Stoar Consulting and Coaching.

Nora: And one note about the AI segment. The evidence that these tools help neurodivergent people at work is early. There are controlled trials now, but they measure things like conversation skill rather than how a job actually goes, and for work itself most of what exists is individual providers reporting success. So take it as promising and not proven. The tools also get rebuilt every few months, so whatever you set up today is built on something that moves.