If this book was written about doctors, it would recommend lollipops to improve healthcare outcomes. Consider: the evidence for implementing checklists to improve outcomes is flimsy, and hard to replicate.

But doctors who give lollipops to patients get a much higher survival rate than those who do not. This seems strange, until you consider that many kids go to doctors for regular check-ups, and are in general healthy. In contrast, many people go to doctors with serious conditions.

While lollipops are predictive of patient outcomes, in that you can make a good guess that if a patient gets a lollipop they will survive, it is not a good intervention.

Sadly, much of this book’s evidence is no stronger. It completely fits my biases: I am a big fan of automated testing, continuous integration, monitoring, trunk-based development, and many of the other practices they tout.

Because of this, I want to scrutinize evidence that agrees me with me hard. This book did not survive the scrutiny.

It is is easy to complain that measuring interventions is hard and expensive. It is! But economists do it using natural experiments. People who have to obey laws and codes about ethical constraints manage to measure health-care interventions. This book is laziness touting something I already believe in.