Good to Great

Version 0.2

A Critical Interpretation of Good to Great by Jim Collins

The core of my reaction is that Good to Great is not even good, but rather deeply flawed. Perhaps I should criticize it based on my own hedgehog concept that time is much more important than money? That is actually linked to the book's extremely narrow definition of "great"in narrow monetary terms, and that narrow focus made me increasingly suspicious as I read the book.

However, in the end I concluded that the premise was completely misleading, and my largest curiosity is whether or not any of the research team members voiced similar concerns. The book purports to be a comparison of great companies to merely good ones, but it is actually a comparison of lucky gamblers to non-gamblers, and the losing gamblers are completely ignored. Near as I can tell, they never considered the possibility that some companies could use ALL of their recommended techniques and still fail to become great or even avoid bankruptcy. The reason I'm certain that is the case is because of the scale of his thinking. There are only so many things that are available to try to be #1 at, and even fewer when defined in the broad fashion of his examples, and there are just far too many companies for all of them to become #1, no matter how long and how closely they follow his recommendations.

My interpretation of this study is that he defined a specific profit profile and then used it in almost exactly the same way they sometimes compare a large number of hedge funds. He started with over a thousand companies, and was able to find 11 winners in the top fraction. In the comparison of hedge funds, they usually use the example of starting with some power of 2 and picking the best one after each assessment period, cutting the sample in half each time until you get the best one--but then it turns out that gambling on this lucky winning hedge fund is going to win next time is no better than chance because the entire procedure is merely a statistical game. Think of it another way: Out of 100 companies, one of them has to be #1 and the top 1% on any dimension you use, and he merely used a dimension defined by stock performance. Of course he found the lucky winners, which is kind of a punchline since his winners mostly appreciated and gave credit to their own luck--and he basically avoided looking at the unlucky losers. (Actually there was one example in his reported data of a company that did everything right until there was an unlucky break, whereupon the company quickly lost it's 'great' eligibility. However, he didn't follow up on that hint to look for other similar cases.)

The author's conclusion is that the principles he discovered can be applied by any organization, and if applied for a sufficiently long time, then success will be assured, but that's NOT the way lotteries work. Actually, the main difference from lotteries is that the payoff date is known in advance, whereas the profit schedules of the companies are not known. His research essentially shows that these companies gambled everything they had on their hedgehog idea, and if they didn't lose, if they were betting on the right hedgehog and outlasted any other gamblers betting on the same hedgehog, then of course they outperformed their competitors who didn't gamble to the same degree. And what about all the companies that gambled everything and lost? Well, they just disappeared without notice or comment. The companies he focused on for comparisons were simply competent companies that didn't gamble. This book is comparing apples to oranges, not comparing great apples to good apples.

The part before this was basically my more considered opinion after a few days. The part that follows is based on my earlier dictated comments using the diary approach recommended by the same guy (google employee #107) who gave Good to Great his highest recommendation. I still haven't decided on the efficacy of this kind of verbal stream of consciousness writing, but I do think the contributed to the clarity of my conclusions in the first part... So here are the cleaned up results of the dictation session:

The hedgehog versus fox thing is the old idea about the fox knowing a lot of things fairly well while the the hedgehog knows one big thing, and knows it extremely well, which supposedly gives the advantage to the hedgehog. My personal hedgehog (and the underlying basis of my critique of this book) is that time is much more important than money. The hypothesis of this book is predicated on measuring goodness purely by the simple metric of stock price. In other words, he essentially ignores time and considers money alone (essentially stock cap growth) as a metric to determine which companies are good or great (and expends almost no thought on companies below those levels). I strongly disagree with this kind of simplistic reduction, no matter how much economists and MBAs like it.

The fundamental problem with the author's approach is that it it's essentially a manifestation of the cancer model of growth uber alles. At his starting point, whatever a company does doesn't matter unless the company grows enough to be extremely valuable (in relative terms), as measured by the performance of its stock price against other companies and the market averages. He didn't select the dates in advance, but pivoted his data to focus around the most convenient dates for the desired performance profiles, which is another kind of criticism, though at least he was clear about what he was doing and his lack of any theoretical basis or hypothesis for the procedure. Given a large enough pool of companies (and he started with more than a thousand), some of them were going to be the best matches and therefore be defined as 'great', and they they used similar financial criteria to select essentially random comparison companies. However, once they had found the supposed differences that were correlated to the great companies (and remember that correlation is NOT causation), they never looked for the companies with their differences that somehow failed to become great successes.

Actually those two paragraphs are highly modified from about twice as much stuff that was too poorly recognized to reconstruct clearly. I wish there were an option to keep the recorded voice until I can clean up the transcription, and obviously use that data to improve the recognition for the next time.