Blitzscaling: Abstractions are powerful
I am taking CS183C (Technology enabled Blitzscaling) this quarter at Stanford. Prompt for this week: Write about how the principles of blitzscaling change as you achieve real scale (e.g. Village, City). Focus on things which: a) resonated the most with you and b) surprised you. Explain how you plan to apply these principles in your own life/business.
—
In the past few weeks, we heard from some incredible speakers such as Reed Hastings (Netflix), Patrick Collison (Stripe), Diane Greene (VMWare) and Nirav Tolia (Nextdoor). One difference I noticed when listening to speakers in these past couple of weeks was that the lessons were richer in abstractions about how to think about various problems when building and scaling startups. A hypothesis for why this might be the case is increasing complexity. Complexity of problems such as efficient team communication, pace of innovation, maintaining a strong culture and maintaining hiring standards increases super-linearly in relation to company size. And optimal solutions to these problems seem counter-intuitive or unnatural at first. Therefore, in order to continue to do an effective job as the CEO, it might become necessary to form models or abstractions to understand markets, products and people, and reason from there as opposed to raw intuition. In the rest of this essay, I write about some themes from these talks that I found most insightful:
Flexibility trumps everything
In evolutionary biology, it is a well-understood fact that species that are adaptable to environmental conditions are the species that survive in the long run. Related to it is the concept of “generalization” error in Machine Learning. The idea is that when training a model on a given training dataset, we want to make sure the model does not overfit to the data i.e. it is robust enough that it makes accurate predictions on unseen data.
Something very similar is true in case of startups. They can be chaotic and less than perfect at times. However, every time something goes wrong, if you put a process in place to prevent it from recurring, then as Reed Hastings put it, “After a while, you have people who are just good at following rules but not good at adapting to a new market”. The debates within a company shift from substance to processes. The culture shifts from moonshots to guaranteeing lower bounds. Therefore, it is essential to optimize for flexibility and speed of learning, as opposed to overfitting for a particular market or problem.
History is underrated
Patrick Collison’s interview had many great points. Something that really resonated with me was that reading history (of anything really, but particularly, history of technology) is a way to “cheat” and steal good ideas! Startups often ignore all the good ideas that people thought of in another time, that are still very applicable today. Instead they go through this “brownian motion” through the space of all possible solutions when working on a problem, as Patrick put it.
Building Cars vs Building Roads
Something else that Patrick talked about was how internet and smartphones have enabled a massive flourishing of products/services that solve utility problems (Uber, Postmates, Doordash), communication problems (Facebook, Whatsapp) or entertainment problems (Netflix, Youtube, Instagram). However, products and services that realize the original “bicycle for the mind” (Software that augments human thinking and makes us smarter) vision of personal computing are surprisingly lacking. This has led to a systemic undervaluing of impact that such a tool can have. Patrick differentiated this by giving the analogy of “building cars” (building the end-product) vs “building roads” (being an enabling force) and that Stripe is one of those companies that is building roads.
Communication: Less is not more
Consider these facts — (a) The universe is expanding and the rate at which two given points in space move apart from each other is approximately directly proportional to the distance between them, (b) The “speed of light”, and hence the speed of causality is finite. Using these, we can conclude that there are parts of the universe that are permanently separated from us, i.e. a beam of light that starts from such a part will never reach us. This means, if for some reason, the laws of Physics were to change in that part of the universe, we’d never know since we cannot observe it.
Something very similar is true of startups — When a company is doubling in size every year, this exponential expansion can quickly outpace the rate at which diffusion of culture can take place via person-to-person interactions. This means, very quickly, individual teams can have separate cultures, separate priorities and goals. This is recipe for disaster and must be prevented. One of the most important things that founders can do during this stage of the company is clear, repeated, transparent and consistent communication. Patrick Collison talked about how it feels unnatural but it is absolutely essential (he also pointed out that in addition to speaking, as the company grows, it makes sense to also communicate in writing. Writing endures, forces communication to be precise, and can be updated).
Role of the CEO
“As the company grows, the CEO is the right person to make a decreasing fraction of decisions”. I thought this statement by Patrick was brilliant.
Org structures might be nearly optimal
I was surprised to hear Patrick Collison’s view on org structures. He basically said that the current solution that we have (some sort of hierarchy) may not be too far off from a global optimum, given some important empirical facts:
(i) whatever the shortcomings of the current organizational structures may be, they have been good enough to create the Facebooks, the Googles and the Intels of the world.
(ii) The timescale of evolution of technology (few generations) is much faster than evolution of people (few thousand generations). This means that solutions to technological problem will be reinvented much more frequently than solutions to “people” problem, such as how can you organize people into teams that work efficiently and towards a goal.
He also touched upon something important when talking about org structures — it is rarely the case that inventing a completely new way to organize people (versus not) will play a significant role in determining the success or failure of the company. Therefore, it might be a better investment of time to work on problems that matter more.
Cash flow businesses vs Growth businesses
I thought this differentiation that Nirav Tolia (and John Lilly) spoke about was very important (and new to me). Conceptually, venture funding makes sense for a startup in order to fund explosive growth to capture a large portion of a market. This may be desirable for a variety of reasons- network effects, economies of scale, outpace regulation etc. But none of these are applicable to cash-flow businesses. Getting venture funding for a cash-flow business is a bad idea, for then you start to think about growth ahead of revenue.
For a good part of 8 years or so starting in 2003, Netflix was locked in a competitive battle against Blockbuster. Talking about what the company learned from this, Reed Hastings said, “When attacked, we should have retreated to do our core better, rather than broadening the surface of attack.”. Fast forwarding to today, “When people ask, ‘Are you going to get into news and sports?’ we say No!”. Many people have echoed similar lessons — “Fewer things done better” (Jeff Weiner), “You can’t drive by looking in the rearview mirror to see what others are doing” (Dick Costolo). But the most effective exposition of this idea I found in Peter Thiel’s Zero to One: the secret (or at least one of the secrets) to success in technology is to do something so unique and so differentiated that in effect you don’t have to compete at all. This runs quite contrary to received wisdom, that the way to win (in sports, games, academics etc) is to compete as fiercely as possible!
—
