Useful to distinguish between pre-Bayesian thinking "what we are doing here can't be reduced to cold hard numbers" and post-Bayesian "even the most sophisticated Bayesian model is a simplified abstraction, and treating its output like a fact can lead you badly astray"
Any successful sales organisation operates on the model that if the numbers and the anecdotes disagree, believe the anecdotes, unless the number is a dollar value. Metrics and KPIs are means to an end, not ends to be fetishized.
Useful to distinguish between pre-Bayesian thinking "what we are doing here can't be reduced to cold hard numbers" and post-Bayesian "even the most sophisticated Bayesian model is a simplified abstraction, and treating its output like a fact can lead you badly astray"
Interestingly, Savage was very clear on the limits of applicability of Bayesian models to "small worlds," where you can always "look before you leap."
Yes. Simon Grant and i wrote a paper on this very point, but it never got published, and the working paper archive is now gone. I should rescue ot
https://www.semanticscholar.org/paper/Bounded-rationality-and-small-worlds-Grant-Quiggin/b53f6c3dcf81d4f4102e169c5ded9750ac7590c3
Please do!
Any successful sales organisation operates on the model that if the numbers and the anecdotes disagree, believe the anecdotes, unless the number is a dollar value. Metrics and KPIs are means to an end, not ends to be fetishized.
Really great article.
Is there a recording of the event with Fourcade?
https://www.youtube.com/watch?v=ehI41DPsGo0
Thanks!
The ordinal society is a brilliant book that every citizen needs to read 😎 Literally!