Drinking from the Firehose #129: 🔭 An ode to physics 🔭


I spent July 4th eating hamburgers and watching fireworks with my wife's side of the family in Michigan. At one point, I struck up a conversation with a relative who had recently finished his junior year of high school. We discussed his favorite classes, and, to my delight, one of them was AP Physics.

I recommended that he pursue a major in physics in college, even if his goal was ultimately something else. I recognize that this advice is somewhat self-serving. Hunter S. Thompson once wrote, "All advice can only be a product of the man who gives it." Well, I myself am a product of an education in physics.

On the car ride home, I reflected on how useful I've found physics to be in many respects in business, and in particular in venture capital. I thought I'd share three examples this week in a longer than usual "big thought."

1. Everything in the universe has a root cause.
Life is complex. Working from "first principles" can help focus the mind on root causes, instead of confusing symptoms of a deeper problem. To see why this framing may be useful, take a common occurrence in venture capital: a portfolio company misses its quarterly plan.

Several surface layer issues may explain a miss. A few deals slipped into the next quarter; a new customer acquisition channel didn't perform as expected; a new product feature, which was expected to increase conversion rates, was unsuccessful. Superficial issues like these can be fixed within a quarter with better processes, more resources, or different team members.

What if the company missed its plan for a deeper reason? All of the above could be explained by a simpler, scarier reason: market saturation. If that's the case, changing your VP of Sales, achieving greater scale in an inefficient channel, or conducting more A/B tests is not going to solve the underlying problem. If you're running out of market, the actions you should take are dramatically different.

The first principles thinking I learned in physics has helped me be a better board member to founders. Immersed in the day-to-day rigor of operations, founders can often miss the proverbial forest for the trees. Physicists, and good board members, are trained to visualize the forest too.

2. Curtail complexity with the right approximations.
Physicists know that very few real world problems have analytic solutions, meaning they can be written down in closed form with common mathematical operations like addition, subtraction, etc. Quantum mechanics, for example, is governed by Schrödinger's famous equation, which describes the evolution of a particle's so-called "wave function" over time and space.

College physics students spend an entire semester deriving results from this equation. Surprisingly, only two non-trivial analytic solutions exist: the particle in a box and the harmonic oscillator. Additionally, you can solve Schrödinger's equation for the orbit of an electron around a Hydrogen nucleus, but even Hydrogen -- the simplest of atoms -- produces equations with exotic special functions. Moving to two electrons (Helium) is more complicated, and once you get to 3+ electrons (Lithium), the solution has no analytic, closed form. Everything from there needs to be solved by a computer numerically.

Other branches of physics are similarly frustrating. The orbit of a 3-body system in space quickly becomes chaotic, with no straightforward solution to Newton's kinematic equations. Fluid flowing over an airfoil, described by the Navier-Stokes equation, can easily break down into turbulent conditions near the surface. Turbulent flow also resists an analytic solution.

I walk through these esoteric examples to illustrate that, even when you're confident in the underlying rules, complexity requires a certain degree of approximation to get a manageable result that can yield intuitive insight. To this point, Howard Georgi, who helped pioneer the Grand Unified Theory at Harvard, once told me that the most step in solving a physics problem is knowing what approximation to apply.

A good example in venture capital is how we think about unit economics. As early stage investors, the concept of a 3-year LTV is laughable at face value, since we frequently invest in companies younger than 3 years old. Even if a company is more mature, the early cohorts tend not to be indicative of recent or future cohort behavior.

In reality, we solve this conundrum with a set of approximations. First, we examine the behavior of recent cohorts. Then, we try to find subsets of users in older cohorts who look like the users in recent cohorts. We stitch together the aggregate behavior from both sets and use them as the baseline for future cohort performance. If we see momentum along a few variables that drives meaningful changes in performance, we may continue to improve those variables to underwrite more upside. We use data from comparable companies to assess the accuracy of our estimations.

When we do this type of analysis, I am often reminded of how physicists choose which approximations to apply in certain scenarios. The physicists I admire most tend to have a "gut feel" for when certain approximations produce better results.

3. Tough problems must be solved by teams.
Gone are the days of the solo physics practitioner. Einstein's generation, and those who came before him, picked the low hanging fruit in theoretical physics. Most of the major discoveries now are the product of teams of researchers collaborating across disciplines. In experimental physics, which requires machines of massive scale, teams often span geographies and languages. The world's largest particle accelerator (The Large Hadron Collider) took a decade to build in collaboration with over 10,000 scientists, hundreds of labs, and over 100 countries. In physics, if you don't collaborate, you won't get very far.

My experience of college physics was that it too fostered collaboration. My professors assigned weekly problem sets that were longer and harder than a reasonably hard-working student could expect to complete on her own. We were encouraged to form groups and discuss problems together. I worked with the same group of half a dozen students throughout my college education in physics. Some of these folks became close friends. I don't think I would have made it through the curriculum without them.

Venture capital is also a team sport, especially at the scale of a firm like Lightspeed. Every new investment we make is underwritten by a small group of investors with diverse skills and perspectives. The larger partnership weighs in on investment decisions and can often add new framing from an outside viewpoint (again, forest vs. trees). When we work with companies, the whole firm pitches in, not just a single partner who sits on the board.

When I first joined a venture capital firm, I lamented my lack of formal training in a directly "relevant" major like computer science. What I've learned, with some maturity now, is that the benefits of a physics education go far beyond learning the actual subject matter. The most useful lessons I learned were how to break down and solve messy problems, and how to do so as part of a team.



Digital couture.

Instagram has become the go-to channel for fashion influencers to reach their fans. Unless you're already famous, becoming a luxury influencer is expensive. No one's giving you pieces for free.

With digital rendering technology crossing the uncanny valley, how long until these influencers become famous for modeling clothing that doesn't exist physically? How long until their followers want to purchase that digital clothing, to show it off in their own photos? The idea isn't as crazy as it sounds. Digital goods are already commonplace in virtual worlds like Fortnite.*

A man, a plan, a canal - Panama! (video)

What happens in Panama affects the world. Its famous canal is "the world's shortcut." I visited one of the locks years ago and marveled as a massive container ship muscled its way forward with inches of clearance on each side.

Wendover produced a video that shows how the canal is interconnected to the rest of the world. The canal effectively sets the maximum width of all container ships. When the Panama government decided to expand the channels, the decision had global ripple effects. The Port of Newark spent $1.7 billion to raise a bridge by a few feet in order to fit these larger ships underneath. Ports in the Caribbean have popped into existence in order to break down large loads from these mega-ships and relay them to smaller ports in the Americas.


Silly as a service.

TikTok, owned by China-based ByteDance, is coming hard at the US. It has accumulated 1.2 billion cumulative downloads and has strong engagement: 8x opens per day, with 45 minutes of daily use.

The unusual thing about TikTok's entry strategy is how heavily they lean into paid marketing on their competitors channels. Spending has ramped to over $3 million a day in the U.S. I wouldn't be surprised if major platforms like Snap* and FB* tried to cut them off for strategic reasons, despite the big dollars flowing in at high margins.

You've got a friend in me.

Snap* recently published its first Friendship Report, and the data is fascinating. Across all major markets, the average person has 4.3 close friends, 7.2 good friends, and 20.4 acquaintances. Millennials are the most likely to share personal topics online; GenZ is more private.


A brief history of bubbles.

A somewhat rambly post on SeekingAlpha caught my eye since it put into perspective recent SaaS multiples, which are approaching record highs. There are now 20+ SaaS companies trading at 15x EV/sales and 15+ trading at 20x EV/sales. For the larger ones (CRM*, WDAY, NOW), multiples in 2019 vs. 2016 are 2-3x higher despite similar growth rates.

The post also goes back over prior times when valuations peaked, including several bubbles. One of the interesting lessons is that the bursting of a bubble isn't always bad for startups. The author writes:

"The burst of the [2000] tech bubble doesn’t get the credit it deserves for essentially creating some of today’s most successful ‘tech’ companies. Collapsing bandwidth prices, compute infrastructure hardware resources, and access to tech human capital costs played a huge part in creating today’s giants. Consider that up until 2005, nearly 85% of broadband capacity in the US was idle. Could Google have thrived or Netflix (NASDAQ:NFLX) have built a streaming business without this idle infrastructure? It’s like asking whether Uber could have ever scaled without AWS and Google Maps, both byproducts of the type of long-term differentiating investments tech bubble survivors were able to make."


The search for Einstein's brain (podcast).

When he died, Albert Einstein's brain was essentially stolen by a rogue autopsy technician. Radiolab found the reporter who tracked it down. The podcast also gets into how much of intelligence is wired into our hardware vs. learned throughout life.


Fast car (video).

My litmus test for a great singer/songwriter/performer is how well she performs live with an acoustic guitar. I have a running list of my favorite such performances.

I recently stumbled upon such a video of Tracy Chapman performing her hit "Fast Car" the year it dropped (1988). Her song and performance are so simple, yet effective. I read somewhere that she got her start performing at Club Passis and other small venues in Cambridge, MA while a student at Tufts. This video feels like you're at one of those intimate performances.