How Decisions Get Made: Approval Chain or Single Signature?

At an owner-run company, one afternoon, the shape of the data model was settled in a twenty-minute conversation nobody sat down for. “Let’s do it this way for now, we’ll fix it later.” We spent the next two years inside that “later.” Every new feature we wrote had to route around those first twenty minutes, and every detour made the next one a little more expensive.

Around the same years, at a corporate, we waited three months for a decision on which library the team would use. At the end of the three months there was no decision — because the scope of the project had changed in the meantime and the library was no longer needed. Nobody counted that as a failure. “The decision resolved itself,” they said.

Two stories, two different faults. But both point at the same root: companies don’t fail at going fast or going slow; they fail at choosing which decision gets which speed. In the vocabulary I set up in the first post of this series, the first axis was “where are decisions made?” This post stands on that axis.

The reversibility frame: which door are you walking through?

The most productive way to argue about decision speed is to sort decisions by how reversible they are.

A two-way door is a decision that’s cheap to walk back. A one-way door is a decision where walking back is either impossible or more expensive than doing it again from scratch.

Move fast through a two-way door: if you’re wrong, you turn around, and the cost is a few days of work. Move slowly through a one-way door: if you’re wrong, you’ll live inside that decision for years.

That much is easy. Here’s the hard part: company types get this systematically backwards.

	One-way door (should be slow)	Two-way door (should be fast)
Examples	The core of the data model, a clause promised to a customer, a hire, the pricing model, who owns a system	A library choice, folder structure, the layout of a screen, an internal tool, the format of a meeting
Where the owner-run company gets it wrong	Decides in twenty minutes, says “we’ll fix it later”	Usually gets this right; speed genuinely is valuable here
Where the corporate gets it wrong	Usually gets this right; slowness genuinely is valuable here	Forms a committee, sits on it for three months, the decision goes stale

The owner-run company sprints through the one-way door, because speed counts as a virtue there and there’s no mechanism to notice the door only opens one way. The corporate builds an approval chain in front of the two-way door, because there, every decision is assumed to carry the same weight.

The first thing to look at when you join a company isn’t decision speed. It’s whether the speed changes with the door. If it doesn’t — if every decision moves at the same pace — that company has never thought about reversibility at all. There’s a bill for that at both extremes.

The real benefit of an approval chain isn’t efficiency

The most common defence of the approval chain goes: “It prevents mistakes.” That isn’t true. An approval chain doesn’t prevent a mistake; it slows the mistake down and spreads it around. If you’re determined enough, you can produce genuinely terrible decisions that have passed through an approval chain — and most corporates have.

The real benefit is something else: it separates the decision from the person.

When a decision passes three signatures, what that means is three people have understood it. They have to; they’re going to sign it. That’s what keeps the decision from walking out the door when the person who made it leaves tomorrow. The reason the answer to “why did we pick this database?” is written down somewhere in a corporate isn’t that someone is unusually diligent — it’s that the structure of the decision required at least two other people to know about it.

Not many people put it this way, because it sounds defensive. Said out loud in a meeting, “we have process so that nobody is indispensable” makes people feel disposable. But that isn’t what it means. What it means is: the institution’s memory should outlive the memory of the person who decided.

This benefit has a limit too, and honestly, the limit arrives early. The moment an approval chain starts getting signed by people who don’t understand the decision, it does nothing — it’s just a delay spread across calendars. If the signature has become a formality rather than a responsibility, that chain isn’t separating the decision from the person; it’s producing a decision nobody owns. How you spot that is a subject for a later post in this series.

The real cost of a single signature isn’t speed

It’s easy to romanticise the other side: the decision comes out fast, it has an obvious owner, nobody can hide behind anybody. All true. But the real cost of a single signature isn’t the speed itself — it’s what the speed gets attached to.

In a single-signature company, a decision gets attached to the owner’s mood that day, to the last person he spoke to, and to the cash position that morning. None of this is bad faith; it’s the ordinary condition of being human. But as inputs to a system, they are noise. If you ask the same question on two different days and get two different answers, the thing deciding is no longer your argument — it’s your calendar.

And the real damage isn’t even here. The real damage accumulates in the team.

People on a team make a decision, and the decision gets overruled. They make another, and that gets overruled. After the third one, they stop deciding. They don’t do it as a conscious protest but as a quiet act of energy conservation: it’ll get changed anyway. Over time this becomes a habit, and then a character trait — learned helplessness. Nothing gets done without being asked, every detail gets escalated, and the word “initiative” turns into an in-joke on the team.

The owner sees this. But he reads it as something else: “Nobody takes initiative.” And most of the time his fix is to make even more decisions himself. This is the most insidious damage an owner-run company does, because it’s circular and it feeds itself. I’ve watched it play out in three separate companies, in exactly that order.

Question: Why doesn’t my team take initiative? Answer: Count how many of the team’s decisions you’ve overruled in the last six months. If the number isn’t zero, that’s your answer. People don’t make decisions they know will be reversed; not deciding is cheaper.

”Who can reverse this decision?”

I closed the first post with this question, because it’s the fastest way to read a company.

If the answer is a role (“the architecture board,” “the tech lead”), you’re inside a corporate structure. The decision belongs to a seat; the person in the seat can change without changing the decision’s standing.

If the answer is a name, you’re at an owner-run company. That isn’t bad — in its healthy form, the decision has an obvious owner, which makes it a structure you can hold to account. In its sick form, that one person’s mood is the company’s only real roadmap.

If the answer is “nobody,” stop. That isn’t a third type of company; it’s a symptom. A decision that can’t be reversed is usually a decision nobody owns. And a decision nobody owns is a decision nobody reviews — because reviewing it would mean owning it.

Three desks

The same decision looks like a completely different thing depending on which desk you’re sitting at. And every desk has a legitimate constraint.

From the senior engineer’s desk. Your job isn’t to make the decision — it’s to frame it. The sentence goes: “There are two ways to do this. The first is done this week but it locks the data model into this shape. The second takes three weeks and doesn’t lock anything. I’d recommend the second, because this is a one-way door. Your call.” If you can say that sentence, you’ve influenced the decision without taking ownership of it. That’s a large part of what seniority means beyond the title; I’ve written about that separately. Your constraint: you don’t see all the inputs. Some of the “wrong” decisions made without knowing how much cash the company burned that month, or which customer is about to walk, aren’t actually wrong.

From the manager’s desk. Your job is choosing which decisions to push upward. It’s a balancing act and both ends are bad. Push everything up and your own signature loses its value; the people above you stop seeing you as a decision point and start seeing you as a mailbox. Push nothing up and, on the day a decision genuinely needs to go up, they’ll route around you — because they’ve learned your desk doesn’t produce decisions. Your constraint: there’s no manual for finding that balance, and you only discover you’ve landed on the wrong side of it months later.

From the owner’s desk. Your job is choosing which decisions to let go of. It’s the hardest job on this list, because letting go feels like losing, and the company’s risk really is coming out of your pocket. But a delegated decision comes with exactly one rule, and it isn’t negotiable: overrule a decision you handed over, once, and the handover is finished. The team won’t decide again; they might go through the motions out of politeness, but inwardly they’ve switched to waiting. So when you’re picking which decision to let go of, the real question isn’t “will they get this right?” — it’s “can I live with it if they get it wrong?” Never let go of a decision you can’t live with getting wrong; pretending to let go does more damage than never letting go at all. Your constraint: you can’t know everything, but you feel obliged to decide everything — and that feeling gets stronger as the company grows, when it needs to be doing the opposite.

All three desks are right. All three are incomplete. The senior engineer can’t see the cost of the decision, the manager can’t see which door it is, and the owner can’t see the mark it leaves on the team.

Closing

Decision speed isn’t a virtue. It’s a setting. Being fast at the wrong door costs exactly as much as being slow at the right one.

When you join a company, don’t let your first question be “do decisions get made fast here?” Let it be this: which decisions get made fast? If the answer is “all of them” or “none of them,” there’s a problem with decisions in that building — and that problem will do more to determine how much of your work is actually yours than your salary ever will. If that math doesn’t work out for you, the conversation to have is a different one; I wrote that up separately under stay or leave.