In our experience, the worst decisions at a startup are rarely the wrong ones. They're the unclear ones.
A wrong decision can be fixed. You learn something, you change course, you move on. An unclear decision is harder. Nobody is sure who made it, so nobody is sure who can change it. People keep debating after the meeting ends. The outcome belongs to everyone, which means it belongs to no one.
We've run into this at Agora, and I expect we'll keep running into it as we grow. So we've adopted a simple rule. Every material decision gets put into one of three categories before the discussion starts. The people involved are responsible for naming the category up front.
Democratic. The team decides together. We share the data, argue the options, and take a majority vote. This works when broad input really does improve the answer, and when the decision doesn't need one owner to move quickly.
Republic. A manager decides on behalf of the company. They gather context, ask the right people, make a clear call, explain it, and own the result. Most of our day-to-day work runs this way, whether it's operations, product, hiring, or a team's own priorities.
Dictatorship. I decide, as CEO, on behalf of the company. This is for a small set of decisions that are highly strategic, time-sensitive, company-defining, or that cut across every team and need one final call. I'll usually ask for input first, often a lot of it. But the decision is mine.
That last category makes people uncomfortable, and I understand why. The word is meant to be blunt. Here's the reasoning.
Some decisions get worse when you try to make everyone happy. Product development is the clearest example. Say two people have strong, coherent visions for a feature. One wants it simple and opinionated. The other wants it flexible and complete. Either could work. The compromise, a bit of both, usually doesn't. You end up with something too complicated to feel simple and too limited to feel complete. Splitting the difference doesn't give you the average of two good outcomes. It often gives you the worst one.
Good products tend to come from a single point of view held with conviction. Someone has to pick a direction and commit to it fully.
That commitment cuts both ways. If you make the call, you live and die by it. You don't get to point at the process when it goes wrong, and you don't get to call it a team decision after the fact. That applies to managers in a republic decision, and it applies to me most of all. Authority without accountability is just control. We're not interested in that.
The framework has limits. Sometimes we'll put a decision in the wrong category. Sometimes a democratic vote will be the wrong tool, or a decision I keep for myself should have gone to the person closest to the problem. When that happens, the fix is to say so and change the category, rather than quietly working around it.
What the framework mostly prevents is performative consensus: meetings where everyone appears to agree, nobody is actually deciding, and the real decision gets made later by whoever cares most. Naming the category up front ends that. Everyone knows whether they're being asked to vote, to advise, or to execute. They also know who to hold responsible.
The goal isn't to make every decision democratic. The goal is to make every decision clear, fast, and accountable.
We're building something that tries to find the best product in the world for each person who uses it. That takes a lot of decisions made well, and made quickly. This is how we try to do it.
