Uncertainty

Aleatory

Chaos, randomness you cannot predict nor control - and no amount of information gathering will help you mitigate this uncertainty.

Zoom had been the unquestionable winner of the COVID-19 pandemic. The Venture Mindset by Ilya Strebulaev and Alex Dang details how Qualcomm Ventures led Zoom’s $6m Series A and owned 2 percent of it when it went public in 2019 at a $9B valuation. With a mixture of belief in the founder and extraordinary luck, some parts of Zoom were aleatory uncertainty with the best possible outcome.

Qualcomm is unique in the sense that, having a startup succeed because of the pandemic, they had one that failed spectacularly as well. OneWeb, a satellite internet startup, took in $3.4B of funding before it imploded when a bridge financing agreement with their biggest backer, SoftBank, broke down. Would you have predicted that one of the larger satellite companies would cease to exist in a few short months due to a global shutdown that brought whole industries to a standstill? Yeah, me neither.

So, now what?

• Price it into your returns.
• Structure deals across your portfolio to protect against downside - key person terms, tranching, and so on. 
• Diversify across risk profiles, geographies, and sectors.
• Be open and honest with LPs: this risk is just a fun part of life.

Epistemic

Epistemic Type 1 is the gap in knowledge that can be researched and diligenced away - your common “can we have access to your data room?” type beat.

Type 2 is more sinister. The answer already exists; somebody already knows it, and you specifically not knowing is just an access issue. It comes across as an ordinary knowledge gap, an invitation to “learn more,” but it is not information you can have no matter how long you wait. To illustrate - 

A newborn baby was diagnosed with a rare liver disease that had a fifty-fifty chance of killing him before he had lived half a life, and the absurdidty of it all, it only effected 1 in 1.3m people. One option was a liver transplant, but it was not possible at that stage and waiting would result in irreversible brain damage. Researchers at the Children’s Hospital of Philadelphia and Penn Medicine, in record-breaking time and efficiency, were able to design a base-editing therapy for his particular mutation. He survived.

It was truly wonderfully crazy stuff - making a cure in a span of six months to treat such a low-incidence disease. It brought the same warm, fuzzy feeling I got when I read about Emily Whitehead and her impact on the field of CAR T-cell therapies. Imagine my delight when I saw that Menlo Ventures had partnered with one of the lead researchers on this case, Fyodor Urnov, to establish Aurora Therapeutics. To make gene-editing accessible even to those with rare gene mutations, so it could be done at a scale to leave no child behind. To say I was bullish was an understatement; if asked, I would have been more than happy to mail my $2 bank balance to anyone at Menlo Ventures if it meant supporting the vision. Aurora launched with $16m in seed funding - $16,000,002 if they had called me.

Now, if it ended in a great company that did great things and venture did what it is supposed to, I would not be waffling on about this. Unfortunately, Epistemic Type 2 happened.

A few weeks in, Aurora announced staff layoffs and deprioritised the cure they had based the company’s founding on. The reason given by the company itself is vague, and I do not think I will be privy to that information any time soon. But I do know one thing: Beam Therapeutics launched BEAM-304, its own editor for the same disease, just 40-odd days later. A much larger competitor with access to most of the patents had undercut them to the point that it made no economic sense to continue on the same path. There is a chance that Beam had been sitting on this, waiting for a clear FDA pathway - which Aurora had helped achieve - and there was no way for Menlo to diligence this away. It was information they did not have access to, and could not.

I still have faith in the remaining team and what they set out to do. I highly recommend reading Menlo’s stance on the startup and the research paper that spurred this; both are linked below.

Ontological

I like to think of this uncertainty as the inherent unknowableness of humans and all things in general. If your founder suddenly pivots into space technology when all the space experience he has is looking up at the night sky, and he simply will not be moved by your pleas of “your company is 60 percent of my NAV,” then that is ontological uncertainty at play.

Do not mix it up with the other two. Aleatory is a process playing out (COVID in this instance); epistemic is what can be known and what was not yours to know. Ontological is more woo-woo adjacent. To get a yes on ontological, you need to say no to both of these: “Is there a single knowable party who holds the determinate facts today, and is the question itself well-defined?”

The simplest way I can explain it is with the moat versus first-mover advantage. If you are in a space early, as you would like to be as a VC, there is no one-size-fits-all metric you can use to test that. The market will prove you right or wrong on several test runs. Being very particular about why and how you went about backing a company - as industries, people, and times change - is addressing this uncertainty head on.

Ok, so how do I use any of this

The way I go about it is tracking how each company moves from stage to stage, and you can apply it to all stages of diligence. Take a team and its ability to execute:

• Aleatory - Do you honestly, heart of hearts, believe this founder will hold up in a genuine shit storm: a currency collapse, a regulatory freeze, a cofounder split? Untestable at present, but you can price it into vesting, key-person terms, and so on.
• Epistemic Type 1 - You already know the drill: track record, how they behaved in their last ventures, customer checks. It is important to be honest if you tripped up here; maybe do not invest in your wife’s cousin’s janky startup next time.
• Epistemic Type 2 - Diligence here will not help. An unresolved co-founder equity or role dispute being managed quietly while they are raising, or a departure of a key hire despite still being listed in the informational materials.
• Ontological - A founder decision that contradicts every pattern of behaviour you have observed to date. Not drawn from their past choices, so no amount of pattern-matching predicts it.

It works best for me when I have a clear question or area I want to do a deep dive on, and a clear sense of what I can control and cannot. It has been humbling to see that most of my experiences have been ontologically driven - i.e., maybe not understanding a founder’s motivations as well as I could have. Again, no model is ever 100 percent, so if you do decide to riff on these concepts, I hope you make them your own.

Credit where it is due: the framework was inspired by Probabilistic Machine Learning for Finance and Investing by Deepak K. Kanungo, chapter 2 in particular. If you would like a more technical and ML understanding of these concepts and how he applied them, I sincerely recommend this book.