Kahneman on the planning fallacy
📚SOURCE
💎PERMANENT NOTES
🧵SYNTHESIS
Paraphrase, not quotation: chapters 23 and 24 of 📕 Thinking, Fast and Slow.
Kahneman tells a story against himself. He once led a team writing a school curriculum. After a year he asked everyone to estimate how long the rest would take; the answers clustered around two years. Then he asked a colleague who knew many such projects how they had gone. About four in ten had never finished, and the others had taken seven to ten years. The team heard this, felt deflated, and carried on. It took eight more years.
The inside view builds a forecast from the plan in hand: these tasks, this team. The outside view ignores the plan and asks what happened to projects like this one. The inside view is not stupid, it is detailed, and the detail is what makes it convincing. It simply has no place for the things nobody has thought of yet.
His remedy: start from the reference class, then adjust.
✨KEY IDEAS
The curriculum team had the statistics in the room, from one of its own members, and did not use them.
An inside-view forecast is a best-case scenario that feels like a realistic one.
The fix is to find a class of similar past projects, take its distribution as the starting estimate, and adjust for what is truly different.
Optimism is rewarded: the confident plan gets funded, and so organisations select for the planning fallacy.
A pre-mortem asks the team to assume the project has failed and write down why.
📝READING NOTES
Twelve days after the launch
I read these two chapters out of order, on purpose, because I wanted to know how wrong I had been and whether there was a name for it. There is: 🧩 Planning fallacy.
The facts of my own case
- Planned launch: 18 February.
- Recommended the slip: 12 February. See 📜 Slip the launch from 18 February to 11 March.
- Launched: 11 March. Three weeks late.
- My estimate in December was built entirely from the task list. I can reconstruct it. Every item on it was true.
Inside view, outside view
| Inside view | Outside view | |
|---|---|---|
| Question | What do we still have to do? | How did the last ones go? |
| Material | The plan | The record |
| Feels | Specific, responsible | Vague, slightly insulting |
| Misses | Unknown unknowns | What is special about this case |
🧩 The outside view has the method. This afternoon I did the exercise I should have done in December: listed Halden's last eight launches against their first announced dates. I had been at the company for all of them. I had watched them be late.
My three weeks were better than the house average. I had been treating the slip as a failure of my planning. Against the reference class it is an ordinary result that I announced earlier than usual.
Where I think he lets us off too lightly
He presents the fallacy as a cognitive error. In an organisation it is also a negotiation. Joachim needed a date for his quarter; an honest range would have been heard as weakness. An outside-view estimate is not only harder to think, it is harder to say out loud. The chapter on optimism nods at this and moves on.
And the outside view has its own trap: choosing the reference class. Eight launches at one company is a small sample, and I chose which ones counted.
What I will do
- Every estimate I give starts with the record and says so.
- A range, with the middle named second.
- The pre-mortem stays, but its findings need an owner, or it is theatre.
What it produced
💎 My launch estimates ignore every launch I have seen, written straight after the list of eight.
Added 26 September: used the outside view on myself for the promotion case: how long do these decisions usually take here, and how often is the first answer yes.
Questions it left me with
- What is the right reference class for discovery work, where there is no launch to be late for?
- If I give Joachim a range next time, what does he do with it? Is there a version he can put in a forecast?
- The curriculum team carried on after hearing the numbers. Would I have slipped the launch in January if I had made the list of eight then? I do not think so, and that bothers me.
Related notes
- 🧩 Planning fallacy
- 🧩 The outside view
- 📜 Slip the launch from 18 February to 11 March
- 📖 Kahneman on two systems
Reading record
- Section read: Chapters 23 and 24
🗣️IN SHORT
- In : Thinking, Fast and Slow, chapters 23 and 24
- Kahneman argues that : plans are forecast from the inside, as best cases, and that the record of similar projects is a better starting point than any amount of detail about this one
- by stating that : planners who knew the base rates still ignored them, that organisations reward optimistic plans, and that reference classes and pre-mortems are the practical correctives