4 ms·
We're getting quite off topic here, but the "duration of a human life" follows a narrow, predictable, bell-curve-shaped distribution. "Impact of this feature" d
by e1g 5y ago
We're getting quite off topic here, but the "duration of a human life" follows a narrow, predictable, bell-curve-shaped distribution. "Impact of this feature" does not - it follows the power-law distribution. Life insurance companies can meaningfully estimate when you'll die, but they cannot meaningfully estimate the value of attending any given party/meeting/trip/conversation. Most of those will be meaningless by comparison to the few that cause an inflection point and make your life worth living, often by accident.
- kqr 5y agoThe neat thing about these power law numbers is that you don't need to estimate whether we're talking 83 or 91. You just need to do enough work to tell apart 1000 from 10, which is often considerably easier. And note that I'm not saying you'll always know down to the cent, or even closest hundred dollar bill. Sometimes your best estimation is going to be "between zero and 999999" and that's fine. If you've determined that, you also know which of your inputs have the most uncertainly, and that might represent a blind spot of yours at an important aspect of the business. Or it represents a fundamentally unknowable thing. But then at least you know you're staking this part of the development on a fundamentally unknowable thing. That's more than many other know.