In April 2016 Jeff Bezos sent Amazon's shareholders his letter for 2015, and one paragraph of it has become the most quoted rule on which decisions deserve a process. Bezos (2016) splits decisions into two kinds: a few are one-way doors, which you cannot walk back through, and most are two-way doors, which you can.
What the letter says, in order
The passage sits in a section about keeping a large company inventive. Bezos describes the first kind like this:
Some decisions are consequential and irreversible or nearly irreversible – one-way doors – and these decisions must be made methodically, carefully, slowly, with great deliberation and consultation.
He calls those Type 1. Most decisions, he writes, are Type 2, and for those the stakes are lower than they feel: "If you've made a suboptimal Type 2 decision, you don't have to live with the consequences for that long." They "can and should be made quickly by high judgment individuals or small groups."
That is the part people quote in leadership meetings. The letter keeps going.
The half the quote drops
The next paragraph is the reason Bezos wrote the first one:
As organizations get larger, there seems to be a tendency to use the heavy-weight Type 1 decision-making process on most decisions, including many Type 2 decisions. The end result of this is slowness, unthoughtful risk aversion, failure to experiment sufficiently...
The sentence ends with less invention, which was the point of the whole section. Read in order, the passage is a warning about speed. The failure he names is a company treating its two-way doors as one-way doors, and he ends by saying Amazon will have to fight that tendency. When the idea is repeated it usually comes out the other way round, as a reason to slow down and be careful, which is the opposite of what he was asking for.
I agree with the warning, and the cost of the opposite error is real too. In the decisions Nutt (1999) studied, managers weighed more than one option in under a fifth of them, and the success rate rose from 56 to 70 percent when they did. A heavy process spent on everything leaves no attention for the few calls that need it. I push back in two places.
A two-way door is only two-way if you notice
"You don't have to live with the consequences for that long" assumes that somebody sees the consequences. In a company of fifty people, a reversible call gets made in a meeting, nobody writes down what it was supposed to produce, and the vendor or the price is still there a year later because nobody checked. The door could have been walked back through, and the people who made the call never learned they were on the wrong side of it.
So I would keep the speed and add two lines to every fast call: a dated prediction, what you expect to see and by when, and a tripwire, the sign that sends you back through the door. Writing both takes a few minutes. Without them the reversibility that justified moving fast exists only on paper. The guide on kill criteria with states and dates covers how to write a tripwire someone can actually check.
Doors change direction over time
The second pushback is that reversibility moves. A price change is easy to undo in its first week and much harder once customers have budgeted around it. A pilot is a two-way door until it becomes a three-year contract. The letter's own word for a one-way door is "consequential" as well as irreversible, and I find it more useful to ask both questions separately, as the guide on which decisions deserve a process does: what does it cost if we are wrong, and how hard is it to undo. Treat a decision as a big bet when either answer is bad, and ask again when a condition you relied on turns out to matter more than you thought.
The two processes, step by step
Both kinds of decision go through the same steps. A big bet does more at each one and adds a few of its own. An everyday call takes five steps:
- Write the decision in one sentence, with what it is for.
- Name the paths, at least the action and doing nothing.
- For each path, write the few things that must be true for it to work and the few things that must not fail.
- Check what you can quickly, a document or a colleague who knows.
- Commit with one dated prediction and one tripwire.
A big bet takes the same five with more care, and adds four:
- A fuller frame, with what triggered the decision, the deadline, the cost of waiting, the lines you will not cross and what is out of scope.
- Real alternatives beyond the obvious pair, which is where Nutt's numbers come from.
- Independent reads from the people asked, given before they see each other's, and a premortem of each path.
- A hearing, where someone argues against the path you lean toward and makes the case for the strongest alternative, and then several dated predictions and a rollout plan.
Midfire asks for both at the commitment. The type decides how many conditions get drafted for each path and whether the hearing runs.
When to move a call from one door to the other
Move an everyday call up to a big bet when one of the conditions it rests on turns out decisive and nobody can check it, or when walking back would now cost real money because contracts are signed and other teams have started building on it. Moving the other way is rarer, and it happens before the decision rather than after it. The best move I know is to split a big bet so that its first step is a two-way door: a pilot with a date and a tripwire, before the contract that would make it permanent.
I read that split as the practical form of Bezos's advice. A company that can turn its one-way doors into a two-way first step decides faster without betting the company on each call, and the letter's case for speed only holds while somebody is still checking which side of the door they are on.