Nobody told me product decisions are reversible or not

I used to agonise over every product decision like each one was permanent. Database choice, colour of the button, whether to add auth, which framework — all weighed with equal dread. Then I borrowed an idea from Bezos via a blog post: decisions are either one-way doors or two-way doors, and the worst thing you can do is treat a two-way door like a one-way.

Sorting my actual decisions

Two-way doors (decide fast, fix later): button copy, page layout, feature flags, most library choices in a well-factored codebase. Wrong choice costs an afternoon.

One-way doors (decide slowly, with data): domain and name, data model fundamentals, the pricing model, who the product is for. These get research, prototypes, and time.

What the lens changed

I stopped researching button libraries for a week and started asking whether I could undo each choice in under a day. If yes — pick anything reasonable, ship, learn. If no — earn the decision.

The paralysis I mistook for rigor was mostly misfiled caution. Most decisions are cheap because they're reversible. The few that aren't deserve the fear I was spreading everywhere else.