Decision Throughput: Why the Founder's Real Job Is Removing Decisions, Not Making Them

Every decision in your company waits in one line, and the line ends at you. You can be brilliant and still be the slow part of your own business. The thing to manage isn't the quality of your decisions — it's how many can move through a single brain in a week, and that brain doesn't scale.

The line ends at you

Picture the last normal week at your company. Now picture every choice that needed a yes — a price exception, a hire, a refund over some threshold, whether to ship the thing or hold it a week. Trace each one. Most of them, if you're honest, ended at your desk and waited there for your attention to land.

That waiting is invisible, which is why it's dangerous. Nobody puts "stalled because she hadn't replied" on a dashboard. The thread just goes quiet. The decision sits in a Slack message you'll get to. The customer gets an answer Thursday instead of Monday, and nobody connects the slowness back to you, because you were busy the whole time. Being busy and being the bottleneck feel identical from the inside.

I ran a company this way for longer than I'd like to admit. I mistook the queue outside my door for proof that I mattered. It wasn't. It was proof the system couldn't run without me, which is a different and much worse thing. A business where everything routes through one person has a hard ceiling, and the ceiling is that person's working hours minus the time they spend sleeping and pretending to.

The metric is throughput, not speed

Here's the reframe that changed how I work. Stop trying to make decisions faster. There's a floor on how fast one person can think well, and you'll hit it, and then you'll start making worse calls under time pressure. Speeding up the bottleneck doesn't remove the bottleneck.

What you actually want to raise is decision throughput: the number of good decisions your business can move through in a week, across everyone, not just you. These are different levers. One asks how quick your brain is. The other asks how many brains are allowed to decide.

Think of it like a kitchen during a rush. You can train the head chef to plate faster, and it helps for about a night. Or you can let the line cooks finish their own plates without the chef tasting every one. The second kitchen serves three times the covers, and the food doesn't get worse, because the chef set the standard up front instead of inspecting at the end. Throughput went up because the number of people allowed to say "this is done, send it" went up.

Your company is the kitchen. The question isn't whether you're a good chef. You probably are. The question is how many decisions per week you've made it structurally impossible for anyone but you to finish.

The one-way vs. two-way door test

Jeff Bezos wrote about this in his shareholder letters, and it's the cleanest tool I've found for deciding what to give away first. He split decisions into two types. One-way doors you can't walk back through — they're hard or impossible to reverse, so you study them, you slow down, you bring in judgment. Two-way doors you can walk back through. You try it, and if it's wrong, you reverse it and try again.

Most decisions, he argued, are two-way doors. Reversible. And reversible decisions should be made quickly, by whoever is closest to the work — not escalated up and treated with the gravity of something permanent.

The founder's mistake is treating every call like a one-way door. The wording on a landing page is not a one-way door. Which vendor to trial this quarter is not a one-way door. A refund for an angry customer is the most reversible thing on earth. But we hold these like they're irreversible because they carry our name, and the held-up cost is real even when the decision is trivial.

So run the test out loud. If we get this wrong, how hard is it to undo? If the honest answer is "we change it back in an afternoon," it should never have reached you. Give those away first, in a batch, this week.

How to actually take it off your desk

Giving a decision away badly is worse than keeping it. You say "you decide" with no context, the person guesses, it goes sideways, and you conclude you can't delegate. You can. You just handed over the verb without the boundaries.

What works is handing over the decision plus three things: the goal it serves, the constraints it can't cross, and the dollar or risk size above which it comes back to you. "You own refunds. Goal is the customer leaves willing to recommend us. Constraint is don't set a precedent we can't repeat. Anything over X, or anything that smells like fraud, flag me." Now they can decide a hundred refunds without you, and the few that genuinely need you still arrive.

Write the boundary down once instead of re-explaining it through ten one-off approvals. A short written rule — what's in bounds, what triggers escalation — is the artifact that actually removes the decision, because it lets the next person decide without finding you first.

Then do the hard part: let the first few be imperfect. The cost of someone deciding 85 percent as well as you would have, today, is almost always lower than the cost of it sitting in your queue until Friday at 100 percent. Throughput beats polish on reversible calls. Reserve your polish for the doors that don't swing back.

"But my judgment is the product"

Here's the honest objection, and I held it for years. Some of these calls really do need me. The taste, the standard, the read on a situation that took a decade to build — that's not a bottleneck, that's the value. Hand it to someone with less context and quality drops where it matters most.

Sometimes true. The door test exists precisely so you keep the one-way doors. Pricing architecture, a key hire, the decision to enter or leave a market, anything that bets the company — keep those. Nobody is telling you to give away the irreversible.

But watch the trick your mind plays. "My judgment is the product" quietly expands to cover everything you've simply never written down. The reason only you can approve the copy is often that the standard lives in your head and you never made it legible. The fix isn't to keep approving forever. It's to extract the standard once — the three things you check, the two things that are dealbreakers — so the judgment can run without you in the chair. Judgment you can teach is a system. Judgment you can only perform is a cage, and you built it, and you're in it.

What you're really building

There's a quiet test for whether you've done this right, and it has nothing to do with how you feel at your desk. Take a real two-week break. Don't check in. When you come back, was there a pile of stalled decisions waiting, or did the company keep moving and meet you with a short list of the genuine one-way doors that were correctly saved for you?

That second outcome is the goal. Not a company that runs without judgment — a company where judgment is distributed, written down, and bounded, so the founder is the rarely-needed escalation path instead of the default queue.

This is also, bluntly, what makes a company worth more than its founder. A business that can only decide as fast as one person is a job with extra steps and more risk. A business that has raised its decision throughput — where dozens of good calls happen daily that you never see — is an asset. It can grow past your hours. It can survive your vacation, your sickness, your attention going somewhere else for a quarter.

That's the real work. Not being the smartest decision in the room. Building a room that decides well when you've left it. If you want a structured way to map which decisions to move off your desk first, that's what we work on at marsa.ai/business.