On some level, Loonshots seems like an overly long and convoluted way of saying something very simple and obvious: that finding solutions and scaling solutions are fundamentally different problems. There, that’s my one sentence summary of the book, which can comfortably fit on the cover, as it’s indeed shorter than the actual subtitle of the book (How to Nurture the Crazy Ideas That Win Wars, Cure Diseases, and Transform Industries). Okay, maybe that’s me being a bit too disingenuously reductionistic. Here’s a more serious high level summary of the themes of the book.

• Finding new solutions to problems (the Loonshots) is fundamentally different from scaling existing known solutions, each requiring its own unique set of processes, skills, and environments
• To be successful, you need to do both. However, as organizations grow, inevitably, structural incentives shift from favoring finding new solutions (riskier Loonshots) to favoring scaling whatever already works (safer franchises), similar to the phase transition of a liquid to solid
• To solve this problem, you must separate the two types of problems into distinct groups, run them differently (befitting to how each of them should be run), while facilitating a good flow of information between them. The physics analogy is to create phase separation with dynamic equilibrium of exchange between two differing phases
• The most important job of the organization leader is to help create, then continuously improve the structures and processes that allow both groups to flourish, rather than managing what the groups are actually doing

Of course, the book expands upon each of these points with much more details and nuances, but the general ideas are more or less all here. What the author spends most of the 368 pages on are stories that showcase and validate these ideas, as is typical of such organizational philosophy books. From the WWII research and development efforts of the US military under Vannevar Bush to the creation of the animation giant Pixar under Jobs and Catmull, these stories span across a wide range of industries and time periods. The stories are fascinating enough on their own to warrant a read through the book. But as far as validating the ideas that the author has put forward, they leave much to be desired.

Don’t get me wrong, the examples from history used in the book perfectly illustrate all the main points listed above. Vannevar Bush was able to successfully lead the American OSRD to the position of technological dominance during WWII by properly separating the scientists (solution finders) and the military (solution scalers). He sheltered the researchers, who, by the nature of their work, failed often and required nurturing and patience. He helped the military, who only wanted more of what was already there, from technological stagnation, by allowing the successes from the R&D group to flow through when they were ready for mass deployment. But how this story of success (and others in the book) help validate the abstract ideas depends as much on what’s in it as what’s not: the failures. After all, it’s fairly easy to cherrypick examples from the past to demonstrate almost any reasonable sounding hypothesis, especially when one is free to retroactively attribute explanations to events. How many others did what Vannevar Bush managed to do, but still failed? How many others, under similar circumstances, were successful, but did not follow the principles outlines by the author? Loonshots did not adequately address these questions in my opinion, and the stories in it left many impressions of survivorship bias in my head.

Perhaps I’m demanding too much rigor in the justifications for the ideas in the book. But a comparison can be made with the approach that Jim Collins employed in his seminal work Good to Great. In Good to Great, the team of researchers examined many cases of successes and failures, in pairs of organizations in similar situations that chose differing strategies. In each of these cases, the decisions and actions of both organizations (usually competitors in the same industries) are compared and contrasted. As the number of examples piled up, the common themes of the successful cases started to emerge, forming the conclusions of the book. This is fundamentally the opposite way of writing than what Safi Bahcall did in Loonshots. Collins began with no hypothesis, and let the data form patterns in an atheoretical way, distilling out the principles naturally. Bahcall on the other hand, seems to have started with a strongly held theory, and gathered examples from the past to justify his position. I think that the approach used in Good to Great, though still prone to human error, is usually able to remove more biases, reduce intellectual narrative manipulation, and form a more convincing argument overall.

This criticism applies broadly to many similar books on organizational philosophy. In a sense, Jim Collins’ book, with his method, is the exception. But this is a review of Loonshots, and it probably doesn’t look too good when I just spent an entire paragraph praising the approach of a different book. Despite the critique above though, I do still think that Loonshots is a worthwhile book to read. The ideas presented, even though some are obvious, while others could’ve been supported more rigorously, are still valid and valuable enough to be pondered. Here are a few keys ones not mentioned above that I thought were especially interesting.

There are two types of Loonshots, “P” and “S”. Remember that Loonshots are the attempts at “finding new solutions”, not scaling, so this is saying that there are, in general, two types of innovations or discoveries. “P” is for Product, the “no way that could ever work” ideas. These are the technological innovations that disrupt entire industries (e.g. the invention of the jet engine), causing dramatic, sudden changes. Missing it usually means a quick death for the organization. “S” is for Strategy, the “no way that could ever make money” ideas. These are hard to spot even in hindsight, and typically exploit some complex interactions of market forces (e.g. the invention of the frequent-flyer program).

Outcome vs system mindsets for analyzing failures. The outcome mindset focuses on questioning why a decision was bad. The system mindset on the other hand, questions the process that caused or was behind the bad decision. The system mindset is one level above the outcome mindset, in the meta sense, and is the one that should be encouraged in the organization. Using the system mindset on successful outcomes is also critical, not just on failure. Bad decision processes that result in good outcomes (due to flukes or luck) are still bad, and need to be fixed.

The threshold organization size at which point the structural incentive transitions can and should be manipulated. Small organizations tend to favor Loonshots, while larger organizations tend to favor franchises (i.e. the groups responsible for scaling known solutions). The threshold size, in number of people, that the balance tips over from one side to the other, varies based on a number of factors. The book actually gives an equation for computing this, but the exact mathematics behind this is not that important. What’s more important is how these factors affect the threshold size, and the fact that by tweaking these factors, one can increase the threshold size, allowing the organization to grow larger without disincentivizing Loonshots. The factor of “return on politics”, for example, decreases the threshold size, while the factor of “project-skill fit” increases it. One thing to note is that even though the threshold can be increased, it will only delay the inevitable. Once an organization is too large, the only thing that will prevent the complete death of Loonshots is by separating them out and protecting them explicitly.

Are the ideas in Loonshots supported by solid evidence? Not as convincingly as I’d like. Can the complexities of organization dynamics really be captured in a mathematical equation, like Bahcall had done? I don’t think so (though the equation is still useful for illustrating concepts). Is reading this book summary enough to substitute reading the book? Maybe for the high level concepts, but not for the history lessons. Does the author try a bit too hard at times to jam physics and network theory analogies into his ideas? Absolutely yes. But underneath all the flaws and quirks, the core philosophy behind the ideas of Loonshots is brilliant enough to shine through. The stories included to help illustrate these principles are never dull, and if nothing else, serve as a fun journey through some very interesting periods of history.