I told my team the project would take two weeks. It took six. I wasn’t lying. I genuinely believed two weeks. I’d estimated using what Kahneman calls the “inside view” — and like everyone who uses the inside view, I was catastrophically wrong [1].
Two Ways to Predict the Future
Inside view: “Let me think about the specific details of MY project, MY skills, MY timeline, and estimate from there.”
Outside view: “Let me look at how long SIMILAR projects took for OTHER people, and use that as my starting point.”
The inside view feels more informed. It isn’t. Kahneman and Tversky documented this as the planning fallacy — our systematic tendency to underestimate time, cost, and risk for our own projects while accurately predicting failure rates for others’ projects [1].
The Data Is Brutal
Buehler, Griffin, and Ross (1994) asked students to predict when they’d finish their thesis. The average prediction was 34 days. The average actual completion time was 55 days. Even when explicitly asked for a “worst case” estimate, students still underestimated [2].
The same pattern appears everywhere:
- Home renovations: average 50% over budget [3]
- Software projects: average 66% over schedule (Standish Group)
- The Sydney Opera House: estimated 7 years, took 16. Estimated $7M, cost $102M
Why We Keep Getting It Wrong
The inside view activates the narrative fallacy. You construct a story of how things will unfold — step by step, everything going roughly according to plan. The story is vivid and feels real. But it only contains one scenario out of thousands of possible ones.
The outside view bypasses your story entirely. It asks: “Of all projects like this, what’s the distribution of outcomes?” That distribution already includes delays, surprises, scope creep, and bad luck — because those things happen in most projects.
How to Use the Outside View
- Identify the reference class. “Projects like this one” — similar scope, similar team, similar domain.
- Find the base rate. How long did they actually take? Not how long was planned — how long did they take?
- Adjust only for strong evidence. If you have a genuinely unique advantage, adjust modestly. But the burden of proof is on you to explain why you’re different from the base rate.
Fermi Estimation: The Outside View’s Practical Cousin
Enrico Fermi could estimate almost anything — the number of piano tuners in Chicago, the yield of an atomic bomb — by breaking problems into components with known base rates [4]. You can do the same:
Instead of: “How long will this app take to build?” (inside view → “maybe 3 months”)
Try: “How many features? (~12) How long does one feature take on average? (~1 week for simple, ~3 weeks for complex) What’s the ratio of simple to complex? (~60/40) Total: 12 × (0.6 × 1 + 0.4 × 3) = ~22 weeks. Plus 30% buffer for unknowns = ~28 weeks.”
That 28 weeks might feel uncomfortable. It should. Comfort is the inside view talking.
The Personal Application
I now multiply every time estimate by 2x. Not because I’m slow — because the data says everyone is 2x optimistic about their own plans. After a year of this, my accuracy went from ~40% to ~80%.
The most rational thing you can do about your plans is distrust your own judgment about them.
References
[1] Kahneman D, Tversky A. “Intuitive Prediction: Biases and Corrective Procedures.” TIMS Studies in Management Science, 12, 313-327, 1979.
[2] Buehler R, Griffin D, Ross M. “Exploring the planning fallacy.” Journal of Personality and Social Psychology, 67(3), 366-381, 1994. DOI: 10.1037/0022-3514.67.3.366
[3] Flyvbjerg B. “What you should know about megaprojects and why.” Project Management Journal, 45(2), 6-19, 2014.
[4] Weinstein L, Adam JA. Guesstimation: Solving the World’s Problems on the Back of a Cocktail Napkin. Princeton University Press, 2008.