On making decisions you can’t reverse

This essay is one of a small group I keep coming back to. I revise it lightly each year as my thinking sharpens. Last revision: April 2025.

I.

There’s a category of mistake I make often, and the people I work with make often, and it took me about a decade to name it: we treat reversible decisions and irreversible decisions with the same level of care. We deliberate on which bug to fix first the way we deliberate on whether to take a job. We dash off architectural choices with the casual energy we should reserve for picking what to have for lunch.

Most decisions are cheap to reverse. The cost of trying the wrong thing is roughly the cost of trying the right thing later, plus a small overhead. Calibrating your deliberation to that overhead is rational. If reversal costs an hour, deliberating for two is wasteful.

II.

But some decisions are irreversible — or near-enough — and the asymmetry between “cheap to reverse” and “impossible to reverse” is enormous. A wrong fundraise is hard to walk back. A wrong hire at staff+ has a multi-year shadow. A wrong commitment to an open-source ecosystem can outlive the person who made it.

Bezos famously calls these type 1 decisions, and I think that framing has held up. The point isn’t that type 1 decisions are rare; they’re actually pretty common. The point is that they’re rare relative to total decisions made, and the spread of consequences is so wide that ignoring the difference is dangerous.

III.

What I try to do, with mixed success, is name the type up front. Before the meeting starts, ideally before it’s scheduled. “This is a type 1 decision” or “this is type 2”. The naming changes the conversation. Type 2 decisions get a quick discussion and a try-it-and-see follow-up. Type 1 decisions get a slower process, more dissent, more research, more time deliberately spent imagining how you’d feel having gotten it wrong.

IV.

The trick — and I think this is the bit most people get wrong — is that the type isn’t fixed. You can sometimes convert a type 1 to a type 2 by adding optionality. The hire that includes a 6-month trial. The architectural choice with a documented exit ramp. The ecosystem commitment with an explicit one-year review. Adding optionality is itself a type-2 decision — it’s cheap, and it makes future you’s job easier.

I think one of the most under-rated skills in product work is converting type-1 decisions into type-2 ones, deliberately, before you have to make them.

V.

The opposite mistake is also common. You can over-engineer a type-2 decision until it becomes type 1 — you make it so expensive to undo that it might as well be permanent. Big monorepo migrations have this property. So do brand redesigns that sprawl over a year. By the time the question “was that the right call?” can be answered, the cost of reversing it is too high to bear, and you’re committed by inertia.

So the practical asks are these. Name the type. Convert when you can. Be wary of letting type-2 work calcify into type 1. The work doesn’t get easier. But the conversations do.

If this resonated, you might also enjoy Opinionated, not prescriptive — a related piece on why systems with strong points of view are easier to work with than systems with rigid rules.