Decisions, disagreement, and speed

Decision speed is the thing I've watched break first, in every organisation that grew past its ability to move.

Not quality. Not communication. Not the hiring bar.

Decisions.

The same call that took an afternoon at fifteen people takes eleven days at sixty, with no individual step along the way being unreasonable, and nobody experiences it as a problem because everybody is busy the entire time.

What happens is that decisions acquire participants.

Someone should probably be consulted. This touches another team. Let's cover it at the sync on Thursday.

Each addition is defensible. The compounding is brutal.

Classify before you discuss

First question in any decision conversation, asked out loud, before the substance.

How expensive is it to be wrong, and how fast could we undo it.

Reversible, cheap to undo. One person decides, now, and tells people. No meeting.

That's the overwhelming majority of them, and applying any process at all is a net loss even when it produces a slightly better answer, because the deliberation cost gets paid every single time and the improvement is marginal.

Irreversible, or expensive to undo. Slow down deliberately.

Write the document, get the alternatives on paper, invite the disagreement, sleep on it. Your primary datastore, your tenancy model, a senior hire, a public API. Anything you can't unwind inside a quarter.

Almost every decision pathology comes from applying the wrong process to the wrong category.

Slow process on a reversible decision is the common one, and it's the reason organisations get slower as they get bigger.

The rarer, more expensive one is a one-way door made casually, because somebody confident happened to be in the room.

And the useful property is that it converts an argument about the decision into a much shorter argument about which type it is. People agree about that far more readily.

Name the decider

Most stuck decisions aren't stuck on the substance.

They're stuck because nobody knows whose call it is, and everybody is waiting for somebody else to conclude.

So say it out loud. "This is Priya's call, she'll decide by Thursday, tell her what you think before then."

Fifteen seconds. It unblocks most of them.

Two things make it work.

The decider is one person, never a group. And they're accountable for the outcome rather than for running a fair process, which is what stops the role decaying into chairing a consensus.

The corollary matters more than it sounds. The decider is usually not the most senior person available.

It should be whoever holds the most context, and that's frequently somebody three levels down.

Every decision you pull upward teaches the org to route decisions upward, and you'll be complaining about exactly that within a year while continuing to cause it.

What you keep is the right to be told, and occasionally the right to overrule.

Use the second one rarely, and explain it every time.

An overruled decision without an explanation costs you two things. This decision, and the next five that person would have made without asking you first.

Disagree and commit, done honestly

The phrase gets used constantly, usually as a way to end a conversation, which is the dishonest version.

Done properly it has three parts. All three are load-bearing.

The disagreement has to be genuinely heard, on the record, in enough detail that the person can tell it landed. Not "noted."

Then the decision is made, by the named decider, and the reasoning is stated, including which arguments moved it and which didn't.

Then the person who disagreed executes it properly. Not sabotage, not "well, I said this wouldn't work," not a slow version.

That obligation is real. An organisation where dissenters implement half-heartedly is one where no decision ever gets a fair test.

What you owe them in exchange, and this is the part usually missing: an explicit note of what they predicted, so that if they turn out to be right, it's on the record and they get the credit. That's what makes commitment safe rather than humiliating.

The version I'd refuse is leadership saying "disagree and commit" before any disagreeing has actually happened.

That's just "commit." Everyone can tell.

Revisiting without relitigating

Decisions do need to be revisited, and the difference between revisiting and relitigating is whether something changed.

New information, changed conditions, or the trigger you wrote down at the time. Revisit.

That isn't weakness, it's the system working, and a leader who never reverses anything is a leader nobody brings new information to.

Same information, different mood, somebody who lost the argument and is trying again with a bigger audience. No.

And name it as such, plainly, once. "We decided this in March, nothing has changed, I'm not reopening it" is a complete answer.

What makes this cheap is writing the revisit conditions down when you make the decision.

"We'd revisit this if we exceed X, or if the vendor's pricing changes."

Then the conversation is about whether the condition has been met, which is checkable, rather than about whether the original decision was wise, which isn't.

The tell that your decision speed is degrading

Count the calendar days between a decision first being raised and it being made. Last five big ones.

If the median is over a week and none of them were one-way doors, you have a decision speed problem.

Almost certainly one of three causes. Nobody knows who decides. People are waiting for a meeting that happens weekly. Or the decider is waiting on a certainty that was never going to be available.

The third is hardest to see in yourself.

Some proportion of your decisions should turn out to be wrong. If none of them do, you're waiting too long, and the cost of that is invisible, because it gets paid in things that didn't happen.