Thoughts on investing in startups

What matters in startups

Most investors agree that these are what startups must have to succeed.

With that combo, most other important things will get worked out.

Investing in people

It is possible to do well investing even if you don't understand a company's technology or the target market. How? Only invest in companies run by teams of people who seem exceptionally competent at achieving most things they set their minds to. If their target market is too small, they will have the talent to pivot. Such teams will provide better than average results in the long run.

Investing in markets

It is possible to do well investing even if you are unfamiliar with a company's technology and a poor judge of management competence. How? Only invest in companies pursuing plausibly very large markets. Because of your ignorance, you will pick a lot of losers. But, because your few winners will capture a lot  of market value, they will cover all the losses.

Of course, the better you get at all the other ways of assessing companies, the more losers you can avoid, thereby improving your portfolio returns.

It takes a team

Most successful startups are founded by a team of 2 or 3 people. Occasionally 4, rarely more, and almost never just 1 person. There are several factors at work.

People related by blood or marriage count as 1 person. Their expected long-term external relationships outside of the company give reasons to act as a unit with regard to company decisions. Many investors will not invest in family businesses.

People who are not committed full time to the company do not count. A startup requires all of its founders' time. They must quit other jobs and any volunteer activities that are not related to the company's business.

The track record fallacy

Investing, especially in the earliest stages of companies, is an activity with orders of magnitude differences between the average and the best results. A single investment can make the difference between an unimpressive and a stellar track record.

Even for the most skillful investor, what will happen to a company in the future is highly uncertain. The probability of a given outcome of an investor's choice for any given investment will be affected tremendously more by random luck than by skill. Much more so than most people assume. Even the probability of the outcome of a portfolio of a few dozen investments will be affected much more by random luck than by skill.

The ratio of the contribution to the probability of an outcome from random luck to that from skill is so great that, to make a statistically significant estimate an investor's skill, would require looking at their performance over thousands of investments. Few startup investors have made investment decisions for thousands of startups.

In venture capital investing, past performance is a poor useless indicator of future results.


With a billion billion monkeys tapping on typewriters for a billion billion years, one will write The Iliad with high probability. How much would you bet that it will write The Odyssey next?

Many destined to fail in the future succeeded by luck in the past. Some who would have succeeded in the future lost by luck in the past. There's no way to tell who is which type.

Good investors stifle psychology

Some investors are more skilled than others. Skilled investors have a significantly higher probability of picking winning investments. So, how can you tell?

[Notice that absence of evidence of skill (such as from a track record) does not mean that there is an absence of skill.]

Investors are people. People make biased decisions (see below). Understanding of biases and effortful self-training to avoid them helps make an investor more skilled. You can test an investor for biased decision making.

<ToDo: Add test questions here>

Even with great introspection and self-training, it is impossible to avoid all bias in decision making. One way to further reduce bias is for two people to work together. They should be thoughtful and open minded but have perspectives as different as possible. Making investment decisions by consensus of the two can help to further reduce the effect of bias on decision making.

Decision biases

Below are summaries of ways investors make bad decisions.

Adverse Selection

Considering only that which is available for lack of a virtue


Being influenced by an immediately preceding feeling about something else


Overlooking the absolute amount and judging relative to a stated reference


Crediting skill for success and luck for failure


Evaluating the probability of things by recalling recent or dramatic examples


Overweighting information that confirms prior beliefs


Splitting the difference or considering the opinions of non-experts or herds


Incorrectly inferring a causal relationship from correlated patterns


A preference for things that are known or comfortable


Reacting differently to the same choice when its benefits or its costs are stated


Overweighting the opinion of somebody with a credential


Judging past events and decisions in view of later-known information


Viewing an absence of evidence is evidence of absence

Law of Small Numbers

Jumping to a conclusion from a statistically insignificant set of examples

Loss Aversion

Holding losers and selling winners early


Overestimating the probability of outcomes that fit a story


Accepting greater risk by inaction to avoid the possibility of blame for failure

Opportunity Cost

Making a decision without the context of alternatives

Sunk Cost

Considering past costs


Studying examples of success while overlooking examples of failure with similar attributes


Belief that recent past random or unrelated patterns predict future events


Overestimating the probability of a desirable outcome

Eliminating psychology

Even with great introspection and self-training and consensus of people with different perspectives (see above), it is impossible to avoid all bias in decision making. However, there is a way to essentially eliminate the biases of human psychology. That is to use an algorithm.

Sadly to admit, computers are better decision makers than humans. In almost all fields where algorithmic decision making has been applied, it has given better results. Below is one algorithmic approach to making the decision about whether to invest in a given startup company.

Algorithmic decision making

This is how I compute a fair current valuation and decide whether to invest in a startup. All estimates are highly uncertain, but an uncertain estimate is better than no estimate at all.

1. Estimate most likely year of exit

It is impossible to know when a liquidation event will occur for a startup. However, calculating a valuation requires  using some year.  [Ideally, the whole valuation calculation could be done based on a probability distribution, but picking a single year is more practical.] It is unlikely that a good startup investment will have an exit in 1 year. 10 years is unlikely, too. Guess which number of years from now has the highest probability of the company having an exit.

2. Estimate the market opportunity

First, ignore the big TAM, SAM, and SOM circles of a top-down market estimate. Think about the following.

a. Who is the customer who pays money to the company in exchange for its product or service?
b. If everything goes as planned, how many customers will there be?
c. How much will each one pay on average per year?

Multiply the number of customers by the average amount they will pay per year in the year of highest probability of exit. That's the expected annual revenue. Subtract the annual cost of producing and providing the product or service from the revenue to determine the expected annual profit.

3. Scale by a valuation multiple

Company valuations assume that sales of current products will continue for some number of years. This is the valuation multiple. It is typically in the range of 2 to 10. It depends on the industry and how this company compares to other similar companies. If you don't know, use 3x. Multiply the expected annual profit by the valuation multiple to get the best-case future valuation.

4. Discount for probability of failure

There are many ways that startup fail or give disappointing returns to investors. Here are 12 common ones.

Do thorough due diligence research, filling out a checklist with narrative descriptions. Read it carefully and think about it. Then, guess a probability in the range from 0 to 1 for each of 12 possible kinds of failure NOT happening. That is 1 minus the probability of the failure happening.

Multiply the best case future valuation by each of the probabilities to get the risk discounted future valuation.

5. Discount for the cost of capital

If the investor doesn't invest in the startup, they will invest in their best alternative, which will have a return on investment.

Divide the risk discounted future valuation by (1+R)^Y where R is the rate of return on the best alternative investment and Y is the number of years until the exit. That is the present value of the company.

6. Compare opportunities

Consult with possible co-investors and possible future follow-on investors for their view of the deal and adapt your view of how good the deal is based on your trust of others' ability to assess fair valuations.

Subtract the best valuation that you can negotiate with the current owners of the company from your estimated present value to see how good of a deal is being offered. If the current owners wouldn't accept your valuation, pass on the deal.

Next, look at all the deals available that you could do and sort them by how good they seem. Think about your investing budget (how far behind you are on a target rate of deploying capital). Figure out how many of the best available deals you could do within your budget. When the deadline comes to commit to the deal at hand, if it is one of the top deals within your budget portion, do it. Otherwise, pass.


Macrosoft is a software-as-a-service company raising a round of funding. They make business managements software specialized for yoga studios.

1. It is likely to grow and be acquired by some large company after building out a range of increasingly useful offerings and building a customer base. It is not likely to get acquired in 3 years. Maybe 4. Based on other SaaS companies, 5 years is a bit more likely than 4. It will probably get there before year 6. So, the year of highest probability exit is 5.

2. Macrosoft expects to charge $250 per month ($3000 per year) from gyms. There are about 15k gyms who would choose Macrosoft over other offerings. There will be a lower level product for yoga studios at $150 per month ($1800 per year) with expected adoption by 25k studios. Together, that is annual revenue of about $90M. Developing and providing the software will cost about $20M. Profit is therefore $70M.

3. B2B SaaS companies might have a valuation multiple of about 5x, giving a best-case future valuation of $350M.

4. These are the 12 risk discounts.

The product of all the failure risks probabilities is .25. The future value of the company is about $88M.

5. The investor expects a 25% ROI on alternative investments. At 5 years, that is a cost of capital discount of .33. The result is a $29M present value of the company.

6. If the company is raising money at a $15M valuation, you have a deal that is good by $14M. You are thinking of investing about $500k per deal, you are looking at about 100 deals closing in the next 3 months, and you plan to invest $3M over those 3 months. $1.5M / $500k is 3 deals to be chosen out of the 100. On the day to decide whether to invest in Macrosoft, if a $14M good deal is in the top 3, you do the deal. Otherwise, you pass.