Writing · Speed
Most decisions are two-way doors
Jeff Bezos split decisions into those you can undo and those you cannot. Most companies treat both the same way, and that is why they are slow.

In his letter to shareholders for 2015, Jeff Bezos divided decisions into two kinds. Some are consequential and cannot be undone. He called them Type 1 decisions, or one-way doors, and they deserve to be made slowly, with care and consultation. Most decisions are different. They can be changed, and if you walk through and do not like what you find, you can walk back. Those are Type 2 decisions, or two-way doors, and Bezos wanted them made quickly, by small groups or single people with good judgement.
The trouble, he wrote, is that as companies grow they start to treat most decisions like the first kind. Everything slows down, and they stop experimenting. In the same letter he noted that Amazon had become the fastest company ever to reach $100bn in annual sales. A year later he added a rule of thumb: most decisions should be made with about 70% of the information you wish you had. Wait for 90% and you are probably being slow.
Small companies copy big ones
In a footnote Bezos named the opposite mistake: companies that make irreversible decisions as casually as small ones tend to die before they get large. Young companies manage to make both mistakes. A team of six can run a decision about a landing page like a merger. The founders want consensus and the investors want a slide. A month passes. Meanwhile a competitor has tried three versions and kept the best one.
Most of what a young company decides is a two-way door: the price, the first market, the wording of the offer, the accounting software, even the name of a product. If one of them is wrong, it takes a few weeks to fix. Spending months to get it right first is the more expensive mistake.
The few doors that lock
A handful of choices cannot be walked back, and they are worth all the time they take. They include who the founders are, how the shares are split, which investors come in on what terms and where the company is registered. Hardware and energy add a few more, such as a factory or a certification that takes years.
That is where we take our time. On the board of a young company the most useful question is often the simplest: if this turns out wrong, can we undo it, and what would it cost? When the answer is yes and not much, the meeting can end.
Disagree and commit
For the moment after the argument Bezos used an older phrase, heard at Sun Microsystems and Intel before him: disagree and commit. He used it on himself. When a team at Amazon Studios wanted to make an original he doubted, he told them so, then wrote back that he disagreed and committed and hoped it would become the most watched thing they had ever made.
Few teams manage it. It needs people who can lose an argument without losing interest. It is one reason we look so closely at how founders argue before we back them.
Sources
More writing
PragmatismA prototype ends the argumentJames Dyson built 5,127 prototypes before his vacuum cleaner worked. Jeff Hawkins carried a block of wood in his pocket. A model settles what a deck cannot.
EnergyAI runs on electricityAI will help double the power data centres use by 2030. Microsoft, Google and Amazon have turned to nuclear power. AI now waits on the plug.
RoboticsRobots learn from peopleLanguage models learned from trillions of words. Robots have no such library yet. In robotics the scarce thing is now data on how people work.