Amazon Prime - Burn to Grow - Commoncog Case Library

Amazon’s growth story hit an unexpected snag in Q3 of 2004. Its online retail store was still growing, but at a slower pace compared to Q3 of the previous year. As an example, the year-over-year growth of its largest product segment, US Media (books, music, video) went from 15% to 12% growth. Not a huge slowdown, but this was a company that had set its sights on becoming an ‘everything store’. By comparison, Best Buy posted an annual sales growth rate of 17% that year. Furthermore, the shift from offline to online commerce was accelerating. This meant that Amazon could lose its head start amidst online retail’s growth if the trend continued.

This is a companion discussion topic for the original entry at

One thing I’ve started to notice from all of these “Idea Maze” posts is that many ideas start from a thesis of how the world works:

  • ByteDance - the core thesis was that mobile internet would fragment user time into shorter chunks, making in-depth/long-form content inconvenient to consume and increasing the prevalence of light-hearted/short-form content. Despite numerous setbacks, Bytedance stuck to this core thesis
  • Apple’s iPhone Keyboard - the core thesis was that a dynamic app interface was preferable over a static one (ie. using a physical keyboard) because it allowed for a broader range of apps to offer high usability and native performance on the device. As a result, the virtual keyboard was a non-negotiable, core feature for the iPhone that needed to be perfected. Despite numerous setbacks, the team never contemplated switching to a physical keyboard due to this core thesis.
  • Amazon - Bezos’s core thesis was that greater convenience and satisfaction for customers would drive sales growth. His core approach to addressing this thesis, in the case of Prime, was to dramatically improve the shipping experience, bringing convenience closer inline with the in-store shopping experience. He stuck to this thesis despite 2 years of negative results for Prime Shipping.

In all 3 cases, the core thing I’ve noticed is that the thesis is a non-rational idea. It does not typically rely on data, but instead on the expert intuition of the founder. Data and rational decision-making is used to help guide the company through the Idea Maze, but the maze itself is defined by the broader thesis. That is, all rational thought is in service of this larger, intuitive view of where the world is heading.

Given this observation, my question is this - how should one go about learning to arrive at better theses of how the world works? My first instinct is to resort to analogical reasoning, much as we’re doing in this case library - read histories of industries similar to your own, as well as histories and biographies of companies and individuals that operate within these industries. These case studies should serve to help intuitively identify trends as they arise.

However, this brings up another question - if the thought process to generate these theses is generally intuitive and we claim that this intuition can be improved through analogical reasoning, how do we evaluate whether or not we are actually improving? The theses tend to unfold over a timescale of years, so the feedback loop is much too delayed to actually evaluate your performance and iterate on your approach.

Perhaps the solution is to read histories of trends that you are unfamilair with and guess how things will unfold before you reach the end of the book or case study, evaluating how you did in your assessment and analyzing where you went wrong. If analogical reasoning really does work well here, then as you accumulate more case studies in your head, you should get more and more of these “unseen” trends correct during your study.


Reading your comment reminded me a little of Marc Andreesen’s ‘Four Kinds of Luck’. To summarise crudely:

  1. Luck Type I: is absolutely random and unasked for.
  2. Luck Type II: by acting and generating lots of motion, you create an environment for serendipity.
  3. Luck Type III: when serendipitous things happen, you have a prepared mind (due to past experiences, etc) that allows you to notice certain opportunities … especially where others do not.
  4. Luck Type IV (which, to be honest, I don’t fully understand): when a unique individual, with a unique perspective and approach, takes on a particularly knotty problem. The opportunities generated by this unique individual are often very different than if done by any other person; so the luck that this person experiences is qualitatively different from the kind of luck generated in Luck Type II.

In many of the cases in the Idea Maze series, it does seem like they all got lucky, but in many cases the luck they experienced were instances of Type III.

All the usual complaints about ‘luck’ and ‘selection bias’ apply to these cases, of course. But what is valuable about these cases isn’t “does luck play a part?” — it’s clear that luck always plays a part. The question worth asking is “given luck always plays a part, how should you think about getting lucky when navigating the idea maze?’”


Yes I definitely agree! The way I like to think of it is the following: luck is essentially the “surprise” of succeeding, which we can model as the probability of success, p(success). A positive outcome that had a low inital p(success) value is seen as inherently luckier than one with a high initial p(success) (eg. hitting a fullcourt shot in basketball vs making a layup). There is always some possibility of failure since p(success), in the vast majority of situations, can approach but never reach 1, so “luck” in some sense is that p(failure) not occuring due to random chance.

I think the key here is that p(success) is not fixed and can be maximized through better thinking, skills, intuition, etc. Yet many people still see highly skilled people with a high p(success) value as being “lucky” (I think this relates to the Base rates are a hell of a drug thread elsewhere on the forum). I think it’s clear from reading these case studies that, while there is an element of luck involved, the founders in question developed a process that maximized their p(success) - their theses of where the world was heading directed experimentation in a way that maximized the probability that they would reach a positive outcome.


I hadn’t seen that luck classification before - I like it!

My mom used to think I was so lucky that I found great job opportunities every time I needed a new job. She had a mental model that all luck was type 1 luck.

My theory was more along type 3 luck. Great opportunities were always coming by, but I didn’t see them when I wasn’t looking for them. You only see what you’re paying attention to.

I’ve come to appreciate type 2 luck over the last few years. Doing something, anything, creates forward momentum, especially if combined with a feedback loop so you can figure out what’s working. I often use the analogy of inertia for this one: an object at rest tends to stay at rest, an object in motion tends to stay in motion. So if you want things to change, get in motion.

Type 4 is interesting. I am connecting this to the theory mentioned in your Steve Jobs Apple case, where he took over Apple, simplified the Mac product line, but didn’t have a path to creating new growth, and said “I’ll wait for the next wave”. I feel like type 4 luck is impactful when somebody’s individual journey intersects with a big societal trend. Actually the original Mac might be a better example where Jobs’s interest in calligraphy and aesthetics intersected with computer tech capabilities to create a new possibility.

It also reminds me of Cal Newport’s phrase “rare and valuable skills” from his book So Good They Can’t Ignore You, where he claims that gaining control of your time depends on creating a niche for yourself where you have skills that nobody else has. This then creates a combination of type 2 and type 3 luck, where you see opportunities nobody else does because nobody else has your unique combination of skills.