Decision-Making Under Uncertainty and Entrepreneurship (with Miles Lasater)

Jan 20, 2021 Episode Page ↗
Overview

Spencer Greenberg and Miles Lasseter discuss leveraging intuition in quantitative analysis, managing uncertainty in decision-making, entrepreneurial risk and skill, and the nature of innovation, drawing from Miles's experience as a founder and VC.

At a Glance
34 Insights
1h 19m Duration
14 Topics
9 Concepts

Deep Dive Analysis

Introduction to Venture Patterns

Decision Analysis and Probabilistic Thinking

Calibrating Intuition and Judgment

Arguments Against Probabilistic Approaches

Combining AI and Human Intuition

Decomposing Problems and Value of Information

Understanding Base Rates in Prediction

Power Laws in Startup Outcomes

Implications of Power Laws for Investors and Founders

The Nature of Innovation

Innovation's Role in Improving the World

The Role of Risk in Philanthropic Innovation

Entrepreneurship: Luck, Skill, or Inherited Trait?

Mindset and Worldview for Founders

Venture Patterns

These are reoccurring patterns, mental models, and ways of thinking or doing things in venture capital and startups. The goal is to document timeless, enduring patterns rather than tactical details or current news, allowing people to learn from others' experiences.

Decision Analysis

This involves using tools of probability, such as Bayesian thinking and Monte Carlo analysis, to make thoughtful decisions. It aims to bring data and quantification into venture capital investing, moving beyond reliance on magic intuition or gut insight.

Calibration of Judgment

This is the process of training oneself to translate expert intuition into numerical probabilities. Through repeated practice and feedback, individuals can become calibrated, making their numerical predictions more accurate and meaningful, so that a 90% likelihood actually corresponds to an event happening 90% of the time.

Monte Carlo Simulation

This technique involves modeling uncertainty by drawing each variable from a probability distribution based on its inherent uncertainty. By repeatedly combining these random draws, it builds a distribution of all potential outcomes, quantifying uncertainty alongside expected value, similar to ensemble weather forecasting.

Value of Information

This concept helps prioritize due diligence or research by identifying which variables, if their uncertainty were reduced, would most significantly impact a decision or valuation. It's about finding key leverage points where the cost of acquiring information is outweighed by its potential to refine estimates.

Base Rates

These are average probabilities or frequencies of events in a given category, which are crucial for informing estimates in specific cases. Paying attention to base rates helps to avoid miscalibration and anchors predictions in broader statistical reality before adjusting for case-specific details.

Power Laws (in Startups)

This describes a fat-tail distribution of outcomes in startups, where most ventures fail or produce little value, but a very small number achieve massive success that accounts for the majority of returns. This distribution implies that outliers drive the average, making it crucial to capture these rare, huge successes.

Simultaneous Invention

This phenomenon describes how multiple teams or individuals often discover the same invention or idea around the same time. It suggests that innovation often involves recombining existing ideas that have become possible due to prior developments, rather than singular, isolated breakthroughs.

Fox vs. Hedgehog (Forecasting)

A 'hedgehog' forecaster has one big worldview and applies it to everything, while a 'fox' knows many things, using diverse models and lenses to view the world and combine them for predictions. Research suggests foxes tend to outperform hedgehogs because they can cancel out biases and weaknesses by integrating multiple perspectives.

?
What are venture patterns?

Venture patterns are documented reoccurring patterns, mental models, and ways of thinking or doing things in venture capital and startups, aiming to provide timeless, enduring wisdom rather than tactical details or news.

?
How can intuition be translated into numbers for decision-making?

Intuition can be translated into numbers by training oneself to state convictions and predictions in numerical probabilities. Through repeated practice and feedback, individuals can become calibrated, making their numerical estimates more meaningful and predictive.

?
What are some arguments against taking a probabilistic approach to decision-making?

Arguments against this approach include leading to overconfidence, the potential for analysis paralysis due to the time it takes, and the belief that certain qualitative factors (like team chemistry or future market size) are inherently resistant to quantification.

?
How can AI and human intuition be combined for better decision-making?

AI and human intuition can be combined in several ways: human input can be a variable for AI prediction, AI predictions can inform human decision-makers, or human and AI predictions can be combined (e.g., weighted averaging) to leverage both strengths.

?
How does decomposing a problem help with estimation?

Decomposing an intractable problem into a series of simpler, more tractable parts allows for much more accurate predictions. By making estimates for each smaller part, and then reassembling them, the overall problem becomes more manageable and estimable.

?
What are the implications of power laws in startup returns?

Power laws imply that most startups fail, and a very few massive successes generate the majority of returns. This suggests investors should aim to not miss the big winners, potentially by investing in a larger portfolio of startups to increase the chance of capturing one of these outliers.

?
Why might LPs prefer smaller venture portfolios despite the logic of power laws?

One explanation is that smaller portfolios inherently have higher variance, meaning they are more likely to appear in the top (or bottom) echelons by chance. This initial 'luck' can then become self-reinforcing as successful firms attract more capital and top entrepreneurs.

?
How can founders reduce risk in startups?

Reducing the risk of doing startups could involve a stronger social safety net (e.g., healthcare), immigration reform, or risk-sharing mechanisms among founders (e.g., having a small stake in other founders' companies) to create a portfolio effect.

?
What are key characteristics of innovation?

Key characteristics include the 'paradox of inevitability' where ideas seem to have their time and are discovered simultaneously, and the crucial role of trial and error, involving constant tinkering and numerous failures along the way.

?
How can investing in companies be a way of improving the world?

Investing in companies, particularly mission-driven ones, can improve the world by fostering innovation, which is a primary source of prosperity and material improvement. It provides resources to individuals taking risks to bring new inventions to scale and to customers.

?
What is the role of risk in philanthropic funding and innovation?

There's a concern that philanthropic funding may not adequately support early-stage, unproven ideas due to incentives to be associated with successful outcomes rather than taking risks. This might lead to an 'innovation system' that is broken for starting new nonprofits, hindering potentially high-impact but risky interventions.

?
Is entrepreneurship a learned skill, luck, or inherited trait?

Entrepreneurship is a mix of luck, skill, and unique attributes. While successful founders often possess intelligence, hard work, and specific insights, luck plays a significant, often underappreciated, role. Many entrepreneurial skills, like persuasion, can also be learned and improved.

?
What mindset is beneficial for entrepreneurs?

A beneficial mindset for entrepreneurs combines long-term optimism (belief that problems can be solved and the future is bright) with short-term pessimism (understanding that daily effort is crucial and inaction leads to failure). This also translates to being inflexible about long-term goals but very flexible about the path to achieve them.

1. Decompose Intractable Problems

When faced with an intractable problem, break it down into a series of smaller, simpler, and more tractable pieces, then solve those pieces and reassemble them for a solution to the whole problem.

2. Apply Probabilistic Thinking

Integrate tools like Bayesian thinking and Monte Carlo analysis into your decision-making process to think realistically about probabilities and improve aggregate outcomes.

3. Calibrate Your Intuition

Train yourself to translate your expert intuition into numerical probabilities and repeatedly practice stating them in numbers to become calibrated, making your gut feelings more predictive and useful for quantitative models.

4. Reconcile Intuition and Models

When your intuition and a quantitative model disagree, investigate the reasons for the contradiction to deepen your understanding and refine both your intuition and the model.

5. Combine Quantitative and Intuitive Judgments

Improve decision-making by combining quantitative analysis with intuitive gut feelings, as the process of detailed grading can make intuition more reliable, and averaging both approaches can lead to better outcomes.

6. Incorporate Base Rates in Predictions

When making predictions, start by considering the base rate (the average probability of an event in a relevant population) and then adjust based on specific case details, rather than solely focusing on individual case information.

7. Embrace a Learning Identity

Anchor your identity to the continuous process of learning, updating beliefs, and incorporating new information, rather than to specific beliefs, to remain adaptable and persistent on the path to truth and success.

8. Detach Ego from Predictions

To improve decision-making and forecasting, cultivate the ability to adjust your beliefs with new information by detaching your ego from past predictions or positions, prioritizing truth over being right.

9. Inflexible Goals, Flexible Path

Maintain unwavering commitment to your long-term goals, but remain highly flexible and adaptable in your strategy and methods to achieve them, adjusting based on data and feedback.

10. Cultivate Optimism, Short-Term Pessimism

Develop a mindset of long-term optimism (believing problems are solvable and the future is bright) combined with short-term pessimism (recognizing that daily effort is critical for success) to drive motivation and execution.

11. Be a ‘Fox’ in Your Thinking

Adopt a ‘fox’ mindset by cultivating many different mental models and lenses for viewing the world, combining them to make more accurate predictions and avoid biases inherent in a single worldview.

12. Embrace Trial and Error, Failure

Recognize that innovation is driven by constant tinkering, trial and error, and numerous failures; therefore, foster social, economic, and political structures that embrace failure as a learning opportunity rather than a career-ending event.

13. Learn from Others’ Patterns

Don’t reinvent the wheel; instead, learn from the recurring patterns, mental models, and ways of thinking used by others in your field, such as venture capital and startups.

14. Document and Communicate Ideas to Learn

To understand a subject more deeply, actively communicate and document your own ideas, as this process serves as a powerful method for learning in public.

15. Iterative Idea Refinement

Refine your ideas through a multi-platform feedback loop: start with low-cost platforms like Twitter for initial critiques, then move to platforms like Facebook for further refinement, and finally to a blog for a polished version.

16. Prioritize Research by Uncertainty Impact

Use quantitative models and simulations (like Monte Carlo) to identify which variables’ uncertainties most significantly impact your ultimate outcome, then prioritize your due diligence and research efforts on those high-leverage variables.

17. Evaluate Information Value vs. Cost

Before gathering more information, assess its potential value in reducing uncertainty relative to the cost of acquiring it, focusing on areas where information acquisition yields the greatest impact.

18. Seek Information in High Uncertainty Areas

When facing high uncertainty or a wide range of estimates, even minimal information gathering (e.g., talking to a few people) can significantly improve your understanding and narrow your estimate.

19. Use Predictive Proxies

When direct measurement is difficult or costly, identify and use sufficiently predictive proxy metrics (e.g., product usage for user happiness) that are cheaper and easier to obtain.

20. Adopt Superforecaster Habits

Improve your predictions by giving precise numerical probability estimates, being willing to make small adjustments to those numbers with new evidence, and continuously updating your estimates rather than fixing them.

21. Integrate AI with Human Judgment

When using AI for decision-making, consider three integration strategies: using human input as an AI variable, providing AI predictions as input to human decision-makers, or combining AI and human predictions with weighted averages.

22. Increase Portfolio for Power Law

As an investor, if you believe outcomes are power law distributed, increase your portfolio size to maximize your chances of capturing the few, massive winners that drive overall returns.

23. Invest in Mission-Driven Companies

As an investor or founder, focus on companies that are mission-driven and aim to improve the human condition, as this guiding force increases the likelihood of creating positive impact, even if not every outcome is perfectly predictable.

24. Identify Impact-Driven Founders

When investing in or collaborating with founders, prioritize those genuinely driven to create positive impact, as their commitment to the mission can ensure critical details are addressed, making the difference between perceived and actual good.

25. Allow Early Founder Payouts

As an investor, consider allowing founders to take some money out of the business at an earlier stage to create a financial floor for them, which can encourage them to continue pursuing aggressive growth and higher-risk, higher-reward strategies.

26. Advocate for Founder Risk Reduction

Support policies and mechanisms (e.g., better healthcare, shared risk models like owning small stakes in other founders’ companies) that reduce the inherent risk of starting a company, encouraging more innovation and problem-solving.

27. Practice Probabilistic Estimates

Engage in practice and iteration, especially with feedback, to improve your ability to make probabilistic estimates, aiming for your stated confidence levels (e.g., 90% likelihood) to accurately reflect actual outcomes.

28. Practice Confidence Intervals

Develop the skill of stating 90% confidence intervals for numerical estimates (e.g., ‘I’m 90% sure the real answer is between X and Y’) to improve your judgment on uncertain quantities.

29. Use Calibration Tools

Utilize free online tools like ‘Common Misconceptions Test’ and ‘Calibrate Your Judgment’ (on clearerthinking.org) to practice making probabilistic estimates and track your calibration.

30. Consult ‘How to Measure Anything’

Read ‘How to Measure Anything in Business’ to learn about making decisions under uncertainty, estimating the state of the world, and calculating the value of additional information.

31. Read Superforecasting for Predictions

To improve your forecasting abilities, read ‘Superforecasting’ by Philip Tetlock, which details techniques non-experts use to outperform experts.

32. Use Decomposed Monte Carlo Models

For investment decisions, use spreadsheet models that decompose a problem into dozens of specific estimates (e.g., market size, team likelihood of success) and run Monte Carlo analysis to understand probability-weighted return on invested capital.

33. Account for Variable Interaction in Simulations

When performing Monte Carlo simulations, ensure your models account for potential interactions between different variables (e.g., low sales impacting financing) to generate more realistic outcomes.

34. Listen to ‘Startups for Good’ Podcast

For more insights on mission-driven entrepreneurship, listen to Miles Lasseter’s podcast, ‘Startups for Good,’ available at startupsforgood.com or @startups4good on Twitter.

I am always struck at how trying to communicate and document your own ideas will help you understand a subject more deeply. It certainly works for me. It's part of how I learn in public.

Spencer Greenberg

I think that as time goes on, the field will become more oriented towards data, quantification, and thinking realistically about probabilities.

Miles Lasater

If your intuition and the model disagree, then the challenge is to try to understand why, and hopefully resolve that contradiction.

Spencer Greenberg

I think that's an extremely powerful move, tying your identity to learning, changing, growing in general, not just about histology, but in general.

Miles Lasater

The average startup goes out of business, and a very few of them go on to make so much money that it pays for all the failures.

Miles Lasater

If our technology gets ahead of our wisdom, we're in really big trouble.

Spencer Greenberg
90%
Confidence interval for a metric (example) Learning to state a range with this confidence is a powerful skill.
320 million to 375 million
Example range for US population within a 90% confidence interval Illustrates stating a range for numerical estimates.
50%
Chance of failure for Y Combinator companies This is an estimate from a few years ago, representing a 'best case scenario' for startup success.
10x, 100x, maybe, or more
Potential improvement ratio of most effective charities over average charities Spencer's intuition on the magnitude of difference in effectiveness.