=
Out of the Crisis by Deming.
Everyone ought to read and absorb Deming. It's hands-down the most life-changing read I've ever come across, by a wide margin.
Dealing with life statistically rather than overreacting to randomness means it's possible to achieve much higher levels of effeciveness and productivity. Working with people to unleash their inner motivation is vastly superior to trying to carrot-and-whip them around.
Understanding that a specific process will yield a specific outcome, whatever outcome you wish for, is how to avoid getting the same result over and over while thinking you ought to see something different because gosh-darned if you aren't trying really hard!!
= Willful Ignorance by Weisberg.
This isn't that good a book, in that it doesn't cover very much content at all, but it really gave me a good intuition for what statistics and probability is about that has helped me understand more advanced things than I was able to before.
Both this book and Deming will take a few years to digest. You'll want to practise it in real life, then come back to read it again, then practise some more. Eventually the lessons will start to stick.
= Performance Modeling and Design of Computer Systems by Mor Halchor-Balter.
Despite the name, this is a book about queueing theory. You know, that branch of statistics that shows up literally everywhere, except nobody seems to understand that it does. I have been dabbling in queueing theory before this book, but the focus this book put on practical applications really made me understand it on a deeper level than before.
= Working Effectively with Legacy Code by Michael Feathers.
I didn't originally want to put this one in this list because it's so predictable, but I can't not. I keep coming back to it and re-reading portions of it every time I have to deal with tricky code. It's just that good at explaining clever ways to get things under test, so they can be worked on with confidence.
= How to Measure Anything by Douglas Hubbard.
What I really take away from this is how powerful it is to be able to estimate ranges that are true 90 % of the time. It's hard to convince others of what a superpower this is, but I hope to work with people who understand that some day.