How to operate

Summary and additional readings based on Keith Rabois' How to Start a Startup class

PayPal, Square, LinkedIn, and Opendoor all have a few things in common: They’re all billion-dollar businesses, all improve with scale, and they’ve all been built, in part, by Keith Rabois.

This post is my attempt to summarise his How to Operate talk at Stanford’s How to Start a Startup class. It’s worth watching in full, which you can do at the bottom of this page.

People are irrational: build an engine

Forging a company is harder than forging a product for the basic reason that people are irrational.

We all know this, somebody in your life, probably somebody close to you is irrational. And building a company is basically taking all the irrational people you know and concentrating them in the same place for hours a day.

So, what you’re trying to do is build an engine. And the first engine you build won’t be pretty, it might not even work. Which is why people work so hard at the start to bring something together.

Eventually, you’ll construct a high performance machine that no body really has to worry about. Rabois says,

We used to joke about at eBay, that if the Martians took over eBay it would take six months for the world to notice. That’s eventually what you want to get to.

The end goal is to, as Peter Lynch said, “go for a business that any idiot can run, because sooner or alter, any idiot probably is going to run it.”

The role of a leader

Rabois says there is only one book that really outlines how to lead.

High Output Management, by Andy Grove outlines the role of a leader as the person who has to maximise the output of the organisation, meaning “the output of a manger is the output of the organisational units under his or her supervision or influence.”

Following that, the CEO is responsible for everything and a VP is responsible for their part of the organisation, as well as the organisations around them.

What that means is a VP of Engineering is responsible for the performance of the product team and marketing team because they have influence there. The inverse is also true.

So that’s how you measure your people, but you want to focus on output and not input. Don’t measure motion and confuse it with progress.

While this might sound glamorous and people get excited about managing a whole large organisation, it’s about being responsible for the output. More on this later.

Editing: simply and don’t accept complexity

One of the most important things Rabois learned at Square is the concept of editing.

This is the best metaphor I have ever seen in 14 years of running stuff, of how to think about your job. It’s a natural metaphor, so it’s easy to take with you everyday and it’s easy to transmit to each of your employees so they can figure out if they are editing or writing.

Whenever you are doing something, you should ask yourself, am I writing or am I editing? A manager’s primary job is to edit.

The first thing most editors do is take out the red pen, eliminating useless words and phrases, omitting things that aren’t important.

And that’s your job to. The more you simplify, the better your organisation will run. It’s hard to keep track and understand a complicated set of initiatives.

The end goal is to distill the business down to one, two, or three things and teach people to think about it through that framework so they can repeat it without thinking.

Simplify. Don’t accept any excuse for complexity.

You can change the world in 140 characters.

You can build the most important companies in history with a very simple to describe concept.

You can market products in less than 50 characters.

There is no reason you can’t build your company the same way. So force yourself to simplify every initiative, every product, every marketing, everything you do.

Basically take that red and start eliminating stuff.

Ask clarifying questions

You need to ask a lot of questions, and they can be simple questions like “should we try this seven days a week? Or six days?” or fundamental questions like “where’s our competitive advantage?”

But the best thing you can do is find the one, two, three, or four things that matter most and focus on those. That’ll allow you to make decisions rapidly and won’t distract you from your real job of building the company.

Allocating resources

Rabois thinks people who work with you should be coming up with their own initiatives.

This is because your end goal is to use less red ink every day, and a good way to measure that is to see how well you are communicating with your colleagues about what’s important and what’s not.

It’s okay to have a bad day with red ink all over the place but you shouldn’t have a month where you’re editing more than the previous.

Consistent voice

Your job is to ensure consistent voice. When you pick up The Economist, you can tell there is one consistent voice.

Every article feels like it was written by the same person. Ideally, your company should feel the same whether it’s the marketing site, press release or the physical packaging on your product.

It should feel like it was written by one person.

Over time, you don’t want to be doing all the consistent voice editing yourself. You want to train people so they can recognise it themselves. Stripe is world-class at this.

It may be helpful to read Style Guide, by The Economist too.


Following on with the editor analogy. Editors don’t do most of the writing in a publication. That will be true for your company too, and the way you get out of most of the work is by delegating.

When you delegate, you are responsible for everything.

The CEO, founder, there is no excuse.

There is no, there is that department over there, this person over there screwed up.

You are always responsible for every single thing, especially when things go wrong.

So how do you both delegate but not abdicate? 

Rabois is a fan of the management technique called Task Relevant Maturity, and all that really means is “has this person ever done this before?”

The more mature they are in a task, the more rope you give them. If someone is new to something, you’re going to instruct and constantly monitor. And if they’ve done it before you’re going to delegate heavily.

Your management style is dictated by your employee.

It's actually a good thing if you do reference checks on somebody and half the people you call say they are a micromanager and the other half say they actually give me a lot of responsibility.

That’s a feature not a bug.

When to delegate

Rabois learned how to delegate from Peter Thiel.

Basically, you want to draw up a two by two matrix that has your own level of conviction about a decision on one axis from extremely low to extremely high and the other axis with the consequence of the decision from low impact to catastrophic.

If the decision is low consequence and you have low confidence in your own opinion, you should delegate and delegate completely, let people make mistakes and learn.

And if the consequences are dramatic and you have extremely high conviction you are right, you shouldn’t delegate. But you should explain your thinking about why you are making the decision.

Okay, you’re the boss

When Rabois was at LinkedIn, he had a talented colleague who would occasionally get annoyed if he didn’t agree with his opinion on something.

So Rabois would spend his time trying to persuade him why he was making the decision a certain way.

And the colleague’s trump card would be “Okay, you’re the boss.”

And that to me was like I was burning a lot of social capital.

Every time he said that I knew I was creating a really thin line and ultimately that was going to backfire if I did that too often.

You want to track the times that you are doing that.

Let people make mistakes

An example of this is at Square, one of my favourite people in the world and my second hire, first marketing hire, had this program he wanted to run called Inner Square which allowed Square merchants to give out, imagine a food truck outside put out ten Squares on the counter and people could just grab them.

And Kyle had this great idea that this would be an awesome marketing program. Squares would spread Squares to other people and to some extent it was on brand. So it didn't have catastrophic consequences.

Each of these ten Squares didn't cost that much money, so financially we could afford to do it.

But at that time, my ten years of experience said it was not going to work on a meaningful enough scale for our metrics and I preferred not to do it.

Kyle was so excited about this that I decided to just let him do it. He learned that when you measure this thing, it's not massive. It doesn't create massive value for the company.

It did require a fair amount of operational complexity to ship all these Squares to people and figure out how to get them, etc, etc. But it allowed him to be excited about his job and to learn how to filter future ideas.

So it was totally worth letting him make the “Mistake."

Editing the team

It’s impossible to have a perfect team and you definitely won’t get that from the start.

Rabois has a good analogy for hiring that he calls barrels and ammunition.

When most people hire a lot of people, they expect that more people = more horsepower = higher velocity of shipping.

But it turns out that it doesn’t generally work that way. When you hire more engineers, you won’t automatically get more done. The same is true for designers, business people, and customer success staff.

And that’s because “most great people are actually ammunition. But what you need in your company are barrels. And you can only shoot through the unique barrels you have.”

When you stock your barrels with ammunition, you increase the velocity of the company. Barrels are very difficult to find. When you have them give them lots of equity, promote them, take them out to dinner, because they are irreplaceable and culturally specific.

A barrel at one company may not be a barrel at another company.

Finding barrels

A barrel is a person who can take an idea from conception through to shipping and bring people along the journey with them. And that’s a culturally specific skillset.

How can you tell who is a barrel and who is not? One is you start with a very small set of responsibilities, it can be very trivial.

It can be something like, I want to reward the engineers in my office at nine o'clock every night with a nice cold, fresh smoothie. This is actually a real example.

I was frustrated, our engineers were working really hard, and maybe 20%, 30% would stay late in the evening and we had already served them dinner but I wanted to give them something cool to reward them. You can think about alcohol but that's a little complicated.

So smoothies were probably a little bit better than pizza, which drains you of energy.

But nobody could get smoothies to show up in my office at nine o'clock sharp, that were cold, that tasted good, and that were delivered in the right place that the engineers would find them.

You would think this is simple but in fact it took two months to get this done. So we had an intern start, and I think on his second day I was explaining this problem, and he said, well I will do it.

And I was looking at him like there was no way.

I have seen my office manager fail, my assistant fail, who were actually pretty good. This just isn't going to happen.

And low and behold they show up. On time, cold, delivered at the right place, and my first instinct was great. Nothing about the smoothies, but now I can actually give him something more important that is more complicated to do.

If you’ve hired someone and people go up to their desk, particularly if they don’t report to them, it’s a sign they believe the person can help them.

If you see this consistently, those are your barrels.

Expand the scope of responsibility

The story above is exactly what you want to do with all your employees, every single day, until they can’t handle anymore responsibility.

And what you want to do is keep expanding it until you see where it breaks and that's the role they should stay in. That level of sophistication.

But some people will surprise you. There will be some people that you do not expect.

With different backgrounds, without a lot of experience that can just handle enormously complicated tasks. So keep testing that and pushing the envelope.

Only talk to Peter Thiel about one thing

Peter Thiel would insist everyone at PayPal could only do exactly one thing. He would say “I will not talk to you about anything else besides this one thing I assigned you.”

The insight here is that people will only try solve problems they know how to solve.

No one wants to bang their head against the wall day after day. And this cascades down the organisation so no one is really solving the most important thing.

Try giving everybody one thing to prioritise.

Don’t make all decisions yourself

Give people the tools to make decisions at the same level that you would make them yourself.

Build a dashboard that simplifies the company’s value proposition into a set of metrics that the whole company can rally behind.

Then measure what fraction of your employees are using that dashboard everyday.

If you’ve done your job well, it should be close to 100%. This dashboard needs to be as intuitive to your internal users as your product is to external users.

Remember, you want to measure outputs not inputs.

Pairing indicators

When you measure one thing and only one thing, the company tends to optimise towards that, often at the expense of something important.

The example Rabois uses is fraud rates. It’s really easy for the risk team to lower the fraud rate by treating every user as a suspect.

Then you have the lowest fraud rate in the world, you also have the lowest level of customer satisfaction score.

In this example, you want to pair the fraud rate with your false positive rate. This forces the team to innovate.

Or with your hiring team, you can pair the number of interviews with the quality of people being interviewed.

Look for anomalies

You don't actually want to look for the expected behaviour.

So a famous example was at PayPal. None of the top ten markets that the company was planning on going after included eBay.

One day, someone noticed that 54 of the sellers actually handwritten into their eBay listings, please pay me with PayPal and brought this to the attention of the executive team at the time.

The first reaction from the executive team was, what the hell is going on? Let's get them out of the system, that is not the focus.

Fortunately, David Sacks came back the next day and said, I think we found our market. Let's actually build tools for these power sellers instead of forcing them to write into their listing, pay me with PayPal. Why don't we just have an HTML button that they can just insert? And that actually worked.

Then he thought, why should we have them insert it each time? Why don't we just automatically insert it for them? They can just insert it once, then every listing they have forever will have it automagically appear there.

So that became the success for PayPal.

Be transparent

Metrics are the first step. Everyone should have access to what’s going on.

The other thing Rabois likes to do is take everything through board decks, and actually review every single slide with every employee after each board meeting. If you can remember the feedback you got from your board, you should pass that on too.

Every meeting should have notes that are sent around to the entire company. This means everyone can keep track of what is interesting and not feel excluded.

When Rabois was at Square, every company room had glass walls so people could see exactly who was in the meeting and who was meeting who.

At Stripe, everyone has access to everyone’s emails.

The Score Takes Care of Itself

The other book Rabois mentions in the lecture is The Score Takes Care of Itself, by Bill Walsh.

The basic point of the book is if you can get all the same things right, you don’t have to worry about the big things.

The big things are a byproduct of what you do everyday to get the small things excellent. As an example, Walsh wrote a three page memo on how to answer the phone.

And that may sound absurd but what his point was organisation as a whole does everything exactly the right way.

Every detail matters.

That is the sort of detail obsession that building this sort of company requires.

Examples that may be a bit more practical for you instead of circuit boards may be, what sort of foods do you serve people?

It actually matters more than you might guess. When people don't like the food you serve them, what do they do?

They go gossip, they go complain to their friends, they walk over to someones desk. Then all of a sudden that lunch that they are complaining about is what they are spending most of their time gossiping and complaining instead of brainstorming.

You don't have this serendipities idea matching another serendipities idea that creates a spark instead they are all wallowing and whittling around.

The best thing you can do is give people the food they want or the food that’s good for them that makes them more productive.

Ending thoughts

My favourite quote of the talk:

So it may seem like this glorious job you thought you had is more like running around being a TaskRabbit for people.

But it is to take things off their plate that is a distraction so they can be high performance machines.

And if you take enough things away from people to distract them and give them the tools to be successful, all of a sudden your organisation produces a lot more.

Stanford CS183B: How to Operate (Keith Rabois)

Additional reading

These are books that are roughly related to Rabois’ talk. Hopefully you find something good to read.

More lessons from Keith Rabois


This is a super interesting topic, I recommend you get started with the following books:

Company building

I recommend reading the following books to get a deeper understanding of what Rabois is talking to:

Management and work

These are books that have changed the way I think about management and work.