If part of your job is saying yes or no to other people's plans, the most valuable skill you can learn is how to look at a proposal from the outside.
The planning fallacy
In 1979 the psychologists Daniel Kahneman and Amos Tversky named a pattern they kept seeing: people predicting how long a task will take tend to imagine how this project will go, step by step, and underestimate how long it will actually take. They called it the planning fallacy. Later work by Dan Lovallo and Kahneman widened it to include underestimating costs and risks and overestimating benefits.
It shows up even in small, familiar tasks. In a well-known 1994 study by Roger Buehler and colleagues, students estimated when they'd finish their senior thesis. Their average guess was about 34 days. Even their worst-case guess averaged about 49 days. The actual average was about 56 days, and only around 30% finished within their own estimate.
The people writing the plan are the ones most likely to believe it. That's not dishonesty. It's what being close to a project does to everyone.
Inside view vs. outside view
Kahneman described two ways to forecast:
- The inside view: start from this project's own plan (its tasks, its team, its schedule) and add it up. This is how nearly all proposals are written. It feels rigorous, and it systematically leaves out the things nobody has thought of yet.
- The outside view: ignore the details for a moment and ask, how did similar projects actually turn out? What did the last five software migrations, product launches, or office moves really cost and take?
The outside view usually wins, because the reference projects include all the surprises the planners couldn't foresee. Formalized as reference class forecasting, the method was developed for practical use by the planning researcher Bent Flyvbjerg and adopted as official guidance by the UK Department for Transport in 2004.
One worked case: Edinburgh's tram line was forecast at £320 million, including a 25% contingency. A reference-class analysis put a more cautious estimate at £400 million. The line opened three years late at a final cost of £776 million. The outside view wasn't perfect, but it was much closer.
Seven questions to ask about any proposal
You don't need a database of megaprojects to use this. You need better questions. Next time a plan lands on your desk:
| Question | What it's testing |
|---|---|
| "What happened the last few times we, or someone like us, tried this?" | Pulls in the outside view. If nobody knows, that's a finding. |
| "What's this estimate based on?" | Separates measured numbers from hopeful ones. |
| "What would have to be true for this to work?" | Turns opinions into checkable claims (see The Mental Models Top Operators Use). |
| "What's the smallest version we could test first?" | Shrinks a big bet into a cheap experiment. |
| "Imagine it's a year from now and this failed. Why?" | The pre-mortem, from psychologist Gary Klein. It gives people permission to name the risks they're privately worried about. |
| "What result would make us stop?" | Agreeing a stop condition in advance protects you from sinking more money in just because you've already spent some. |
| "Is this reversible?" | Easily undone decisions deserve speed; hard-to-undo ones deserve this whole list. |
Notice what isn't on the list: "How confident are you?" Confidence is the one thing the planning fallacy guarantees in large supply.
How to say it without crushing people
Leaders who interrogate every plan can teach a team to stop bringing ideas. The goal is a better plan, not a defeated planner. A few habits help:
- Ask questions, not verdicts. "What did the last migration take?" lands better than "This will never be done by March."
- Separate the idea from the estimate. You can love the idea and still double the timeline.
- Reward honest numbers. If people who give realistic estimates get turned down while optimistic ones get funded, you will get optimism. Your reactions set the incentive.
- Approve in stages. "Yes to a six-week pilot, and we'll decide on the rest with real data" is often better than either yes or no.
The honest limits
The outside view is a correction, not an oracle. Sometimes there's no good reference class, because the project really is new. Sometimes the past is a poor guide because the world has changed. And ambitious projects do sometimes beat the odds, which is part of why people keep attempting them. Flyvbjerg and his co-author Dan Gardner argue that the answer isn't to stop building big things, but to spend more time planning and testing before committing, when changes are still cheap. The aim is not to say no more often. It's to say yes with your eyes open.
The takeaway
Almost every proposal is built from the inside view, and the inside view is reliably too optimistic about time, cost and benefits: the planning fallacy. As the person deciding, your job is to supply the outside view by asking how similar efforts actually turned out, what the numbers are based on, what would have to be true, what a small first test looks like, what would make you stop, and whether the decision can be undone. Ask in a way that rewards honest estimates, and approve in stages where you can.
Try this: pick one project currently being planned around you and ask just one question: "How long did the last similar project actually take, compared with its plan?" Then compare that ratio with the current plan.
Sources & Further Reading
Sources
- Sydney Opera House: original 1957 estimate £3.5 million with a January 1963 target; opened 20 October 1973 at A$102 million (see the Sydney Opera House's documented history, summarized in Wikipedia's entry and its sources).
- Planning fallacy: Kahneman & Tversky (1979); broadened by Lovallo & Kahneman (2003). Thesis study: Buehler, Griffin & Ross (1994), average predicted 33.9 days, worst case 48.6, actual 55.5, about 30% on time.
- Reference class forecasting: developed for practice by Flyvbjerg and COWI (2004); UK Department for Transport guidance (June 2004); Edinburgh Tram Line 2 figures (£320m forecast; £400m reference-class P80; £776m outturn).
- Pre-mortem: Gary Klein. Flyvbjerg & Gardner on longer planning: How Big Things Get Done (2023).
Go deeper
- 📘 Thinking, Fast and Slow — Daniel Kahneman (the inside and outside views, in Kahneman's own words)
- 📘 How Big Things Get Done — Bent Flyvbjerg & Dan Gardner (why projects overrun, and how the successful ones plan)
- 🌐 Companion pieces in this library: The Mental Models Top Operators Use and How to Make a Hard Decision