The outside view
🧩EXPLANATION
Estimating from how things like this usually go, before looking at the details of this one. The method is sometimes called reference class forecasting:
- Name the reference class: projects of this kind, at this size, in this place.
- Find its record: how long they took, how often they failed.
- Start from that record. Adjust for the particular case only where there is a reason you could defend to a stranger.
It is the correction for the 🧩 Planning fallacy, and it comes from the same chapters of 📕 Thinking, Fast and Slow. My notes: 📖 Kahneman on the planning fallacy.
Often confused with: pessimism. It is not a mood. It is a different source of data. It also has a real weakness: choosing the reference class is a judgment, and a careless choice returns a confident wrong answer.
Where I have seen it
Halden's last eight launches. On 23 March I listed them against their first announced dates. I had been at the company for all of them.
- Average: six weeks late.
- Mine: three weeks late, and I had spent February feeling like the first person it had ever happened to.
The conclusion is its own note, 💎 My launch estimates ignore every launch I have seen.
Since then every date I give has the record of similar work beside it. When there is no record I call the estimate a guess, in that word, which is less popular than a date.