“Bad companies are all alike; every Good company is Good in its own way” – Peter Thiel

This essay is based on two ideas:

1. Complex Adaptive Systems cannot be modelled, but the properties of these systems can be measured. Benchmarking the measurements to other systems helps us to see if we’re on the right path.
2. Good businesses are complex business, as Peter Thiel says “Competition is for losers”. A good business, is by definition a business that is hard to copy or catch up to. Therefore, by measuring the complexity of a business system, we can truly get a sense of it’s moat.

I had this idea a long time ago for picking good PhD topics. As I can no longer call myself a researcher, I’ve shifted the focus to businesses. Regardless, I’m quite confident that the same principles apply.

By the end of this essay, my goal is to convince myself of my own ideas. I’ll cover the following properties of complex systems, but for the sake of brevity will not define them as they’re well documented:

• Emergent Properties
• Adaptation
• Interactions
• Non-Linearity

Emergent Properties

Simple emergent properties are the source of some of the most difficult to cross moats in all of Silicon Valley. The most popular emergent property that most people are aware of is Network Effects. This is a classic example of having interactions between your users can result in a value that’s not created from any of the individual components. However network effects purely dictate that the performance of the core product improves per users. I would argue that brand new emergent products arise through emergent properties which business can leverage to further their moat.

Some of these other emergent properties are; ad-hoc marketplaces and emergent machine learning models (e.x. anomaly detection). These properties arise not because any specific individual, but instead because a group of user gather and start to use the platform:

• Facebook: Classic Network Effect of the platform being more valuable the more your friends and family use it.
• Github: Github’s platform doesn’t really get any better the more people use it. It’s quite rare that we would browse the top repositories instead of going to them directly from twitter. In fact, Github allows anyone to download a repository without even signing in. Instead, they’ve built a marketplace to help sponsor open source developers on their platform. Now a platform that was built to share code can help to support infrastructure that the entire world runs on.
• Stripe: While initially just building a payments infrastructure company, aggregation of a large set of users allows Stripe to easily build and deploy machine learning models. Stripe can easily identify outliers through these emergent ml models which help build anomaly detection algorithms to combat fraud.

Adaptation

The hallmark of a good business is that it’s users continue to learn and improve over time. The hallmark of a great business is that users adapt to other users of the same platform. A perfect example is Uber pricing, as someone who knows Uber prices are sky high at 9am, I’ll leave my house at 7am to avoid the surge pricing. This specific example doesn’t lock me into Uber’s platform, but instead helps to build a steady stream of demand for Uber drivers resulting in great revenue in the long run. It’s obvious to see that a 100 drivers can over a 100 users in one hour, but a 100 drivers can cover 200 users in two hours leading to greater revenue without needing greater investment in driver sign-ups.

A contrasting example is where user’s don’t need to adapt to other users who are using the platform. These are examples of businesses such as information systems such as weather apps. Due to the lack of adaptation required on weather apps, we can all be satisfied with surface level information but don’t need to investigate deeper.

Meanwhile, a Bloomberg Terminal’s and Weather App infrastructure is nearly identical, but Bloomberg Terminals are significantly more valuable as a business. I know that other traders are trading on surface level information on the terminal, which means to out compete, I investigate a layer deeper. My nemesis knows I’m going a layer deeper, and therefore needs to adapt to my behavior to go two layers deeper. Despite having the same purpose (gathering information) the adaptation required in the complex system of the stock market makes Bloomberg a better company.

Interactions

During the recruiting season at INSEAD, I observed a psychopathic relationship between my friends and their Linkedin accounts. It seems that in order to get a job, they needed to have; the perfect headshot, the perfect bio, the perfect location (God forbid you apply to a job in London from Paris), the perfect experiences with a detailed blurb on each company, the perfect education highlighting their GMAT score, the perfect posts about what they learnt, the perfect likes on their Linkedin history (Sorry Gianluca, I can’t follow your marijuana business while I interview for McKinsey).

Fascinatingly, all of these small components interacted and were interconnected to their ability to network and find a job. Despite the absurdity, they were all correct in their approach. This leads to why Linkedin as a social network for professionals is so powerful. Each of these small components have dependencies that drastically affect your outcome.

Interactions are required to build a valuable business. Long term businesses require a system where small interactions can have an outsized impact on the final result to capture any of the value they create. Any business with a straightforward linear structure is asking to be disrupted by a by a CS freshman’s final project.

Imagine a world, where there is only one tax rate for everybody with no rebates, dependants, or subsidies. Just a flat, 30% rate. Would you still use TurboTax to help you with your taxes?

Non-Linearity

Traditional investors already look for non-linearity in returns, for example the first lesson of pitching to VC’s is that you better have some sort of exponential or hockey stick curve in your platform. Instead, I would like to look at businesses from another lens. The non-linearity of outcomes on the platform.

As strange as it sounds, I think a unassailable (not good) product has exponentially more returns based on small increases in the user’s input. A extreme example is sports cars. The more effort and skill you put into driving a sports car, the significantly better you’re able to perform compared to novices. Small changes in accelerator pressure and wheel angles can results in vastly different lap times.

These types of products lend themselves well to human creativity and skill expression. While making a product complex and non-linear flies in the face of good product development, I don’t consider it a factor in a good product. Instead I view it as a consequence of a good business. Any business that aims to tackle a complex field which has enormous moats, by definition has a product that’s extremely non-linear.

A small difference in Airbnb photos, cleanliness, or furnishing will have an extremely outsized effect on your earning on their platform. Not because Airbnb’s product is complex, but the system of vacation rentals is exceedingly complex. The same can be said about a variety of other companies such as Shopify (Design/Merchandising) and Instagram (A influencers personality/appearance/photos vs their influence). A great influencer can impact millions of people, which means that Instagram only needs recruit one Kim Kardashian for 300M users to follow vs needing to recruit 1M influencers which 300 people each follow.

To iterate again, good businesses chose complex markets, therefore their product’s try to solve complex problems, leading to a drastic non-linearity on their platform.

Measuring TCV

TCV has recently invested in the series A of Nourish 7 days ago. To quote from their press release:

[Nourish] provides clinical-grade nutrition care powered by a virtual network of dietitians, all covered by insurance.

“For patients, we(Nourish) unlock access to personalized nutrition care that has historically been limited to a subset that can afford to spend thousands of dollars out-of-pocket. For [Registered Dietitians], we remove the barriers to accepting insurance, enabling them to focus on patients and building their practices rather than administrative tasks and paperwork. For payers, we’re providing access to a high-quality dietitian network and virtual nutrition platform to measure and improve clinical outcomes.”

Democratizing Nutrition: Our Investment in Nourish

Let’s evaluate Nourish as a Complex Adaptive System:

• Emergent Properties: As more and more patients subscribe to Nourish’s platform, it can better connect patients to dieticians, through network effects. Over time, Nourish can also build ML algorithms to better detect and optimize food delivery routes and patient conditions. [8/10]
• Adaptation: User’s don’t need to adapt to other users using the platform. I get the same use from the platform no matter what other people do. Perhaps, I have to act quickly to book an in demand nutritionist, but that seems to the limit. [1/10]
• Interactions: There are major interactions between each health issue you’re facing the resulting impact on food delivery, dietician support, and insurance coverage. For myself, as a vegetarian worried about Heart health and cholesterol in my family, along with limited US insurance, Nourish’s platform makes it very easy for me to be paired with the best dietician. [10/10]
• Non-Linearity: There is no non-linearity on the platform. Each dietician can only support a limited number of users, and regardless of the quality of their profile, it’s unlikely to increase the number of clients they can procure to the limited amount of time they have. [0/10]

Based on the criteria, I would evaluate Nourish a good investment given the core impact of their platform around Emergent Properties and Interactions. Long term, I would worry about Nourish’s ability to build a defensible business that other companies cannot easily replicate or for insurance companies to build their own systems inhouse. I would argue that Nourish could drastically increase it’s scores in Non-Linearity by signing exclusivity contracts with insurance platforms, forcing large number of user’s to use their platform dietician support or have non at all.

I’ve set an alarm in my calendar to check on this in 3 years (2027). I’ll update this page then.