Introduction

This post owes a great deal to prior work and thought by Spencer Greenberg, Eric Gastfriend, and Peter Hartree.

This post is a summary of the object-level thought on what makes a startup high impact which we developed while working on the Good Technology Project.

A lot of this material is more-or-less obvious applications of EA thought to startup theory. Nonetheless, it managed to be surprising and useful to people, so perhaps it is less obvious than it seems. I’ve condensed the presentation given the intended audience of this post - there is a lot more to say on many of these points. This material might have eventually developed into a “guide” to effective entrepreneurship.

In addition, some of the material relates to how to manage a startup in later stages. We never really got a chance to try that out, so it is especially speculative.

What makes a startup high impact?

We’re interested in startups because we think that they might be a mechanism by which we can have a large positive impact on the world. But what are the qualities that we should look for in a startup?

Impact model

Before we get started on assessing how good a company is, we should try and get clear on how that company benefits the world, to make an impact model for the company.

The first big consideration is who the company helps. Usually there will be one group in particular that you are expecting to benefit. A good way of figuring out who these are is to consider the various groups of “affectees” for your company.

Customers

Customers are the most obvious people who benefit (or suffer) from the existence of a company. They pay a cost in money and time, and they gain your product in return.

For example, Mesh Power’s1 primary beneficiary group is its customers (insofar as you think that promoting clean energy over burning kerosene might have environmental benefits, Mesh Power may also have some externality benefits, see below).

While we can usually assume that people will buy things that actually improve their lives, this isn’t universally true. Cigarettes and addictive drugs or games are examples of this.2

Third parties

The operation of your company will also affect people who are not part of the transaction, or even involved at all. These effects are called externalities. Often these are positive, in the case of innovation and economic growth, but they can also be negative, such as pollution, developing dangerous new technologies, or causing technological unemployment.

For example, despite being a car company, arguably Tesla’s primary beneficiary group are third parties, because accelerating the progress of electric cars and storage will help to ameliorate climate change.

An important class of externality is benefits produced by your customers, which will often happen if you’re selling to businesses or institutions. For example, disease outbreak monitoring systems may be sold to governments, but the beneficiaries are the people who don’t get ill because of the government’s’ improved preventative action.

Employees

A third category of beneficiaries is your employees. They will gain pay and satisfaction from working for you, but will also spend their time. In bad cases they could experience physical or psychological harm because of the job.

For example, one of M-PESA’s beneficiary groups are among its employees, since it needs lots of places for customers to buy and sell mobile money, and this provides additional income for a lot of relatively poor shop owners.

Impact mechanism

The next thing to do is to work out how you think your startup will actually affect your target group of beneficiaries. This is likely to be very uncertain, especially if you expect to create an impact through externalities. However, it’s better to explicitly write down what you’re uncertain about nonetheless.

For example, here’s one mechanism by developing a better test for drug-resistant TB might improve wellbeing:

  • Decrease cost of TB test
  • Increase availability of test in low-resource areas
  • Accurately distinguish more cases of drug-resistant TB from normal TB
  • Give more drug-resistant TB sufferers the correct drugs
  • Cure more people of drug-resistant TB than otherwise
  • Fewer people go through the lengthy suffering of drug-resistant TB
  • Increase wellbeing

There may well be several such mechanisms, of course!

Once you have an explicit impact mechanism, that gives you a two useful things: a set of hypotheses about how your impact occurs, which you can test; and a set of stages in the mechanism which you can measure.

Most of these won’t be things you can test or measure now, but it’s worth thinking from time to time whether you might be able to measure more of them. For example, in early development you might focus on measuring the cost of the test, but as you roll out you might also be able to measure improvements in availability.

Assessing the impact model

We can apply our usual INT heuristics in this case, although we can pick out some particular considerations for the domain. These can work both for picking out a broad problem area, and for directly picking out factors relevant to a particular impact model.

Scale

As ever, we care about both how many people we help and how much we help them.

We should think about maximum scale here: if you could eventually sell your product to everyone on Earth, that’s better than if you’re limited to just one national market. If we think about our possible beneficiary groups, third parties tend to be the biggest group, followed by your customers, and then your employees.

Similarly, a life-saving product is much better for each person than something that merely saves them some money.

Tractability

There are a couple of big things that affect tractability.

The first is obvious: the problem may be hard. Or the problem may be easy, but making it profitable may be hard. And we’re primarily thinking about businesses here, so if you can’t make it profitable, you can’t do it.

Secondly, you might not want to do it. Running a business is hard work, and you face pressure not only to drop out, but to cave in on issues where your investors or advisors may not be aligned with what you want. If your beneficiary group is your customers, then your profit goals and your impact goals are relatively aligned, so this may be easier.

In other cases this is less likely. For example, Uber (may be) benefiting its 1.5 million drivers. But they are not incentivised to employ these people, because doing so costs them, so as soon as they can automate them away, they will.

Finally, you might not be able to figure out what to do. Even if you can identify the problem, you may not be able to figure out a plausible mechanism to actually have an impact on it, or your mechanism might fail to work.

Tractability issues result in two big failure modes:

  • The business fails entirely
  • The business succeeds, but it has a low or negative impact

Counterfactuals/Neglectedness

Assuming that you start a business that solves a real problem, we can assume that someone would have solved it eventually. That means that the effect you have is the difference between those two, which will look like getting X extra years of the solution. We can call this your time advantage.

Generally, the bigger the time advantage the better. If the problem is big enough, then even a short time advantage may not be a problem - getting a malaria vaccine a year earlier would be huge!

But generally bigger is better. There are a few ways you might have a big time advantage:

Firstly, the technology you use has existed for a while but hasn’t been applied to the problem that you are applying it to. That suggests that it would continue to be unsolved in that way for a long time if you don’t do it.

Counterintuitively, this suggests that you should stay away from new technologies: it is very likely that someone will try “machine learning for X” relatively soon, so it is unlikely to be neglected.

Another common case is that the problem requires an unusual combination of skills, knowledge, or inclinations. For example, you might know about both financial services and the developing world, while also being altruistic. Combinations of traits are correspondingly rarer - if you have at least one moderately rare skill, then it is likely that you also have one very rare combination of skills. It may be a long time before this combination comes along again, and so if there are problems that require it, they may go unsolved until then.3

This suggests that you should look especially hard for problems that only you (or you and your friend with the other unusual skills) can solve, because that is likely to give you a big time advantage.

Finally, the incentives to solve the problem may be lacking (e.g. the customers are poor). This is a tough case, because those incentives will also be lacking for you. So you need a good story about how you are going to keep your impact on track. Many benefits to third parties have this form. Often if the externality is innovation then a strong founder can ensure that most of the benefit is produced before they are phased out. For example, Tesla has chosen to give away their patents for free, which might not have happened with a less altruistic CEO.

  1. Sadly, it looks like they’ve gone bust since I last checked, but they’re still a good example of the principle. 

  2. Spencer Greenberg’s podcast discusses some of the ways startups can unexpectedly cause harm. 

  3. Peter Thiel talks about “secrets” which are unusual beliefs that you have which make you think that a problem is soluble, even though the general belief may be that it is not. These are another thing that can make you unusual.