Reducing Friction

One principle, and it’s a subtraction: remove friction before adding more.

The instinct in most organizations facing slowness is to add — more people, more tooling, another initiative, another coordinating role. Adding is visible, fundable, and reportable. Removing is none of those things, which is why it’s rarely anyone’s proposal even when it’s obviously correct.

But removing friction lets you go faster without adding capacity — and the effect compounds, because you aren’t adding energy to the system, you’re stopping the system consuming it.

Practically, that sequences into three moves. The order matters more than the moves do.

First: make ownership unmistakable

Almost everything else is downstream of this. Committees created to compensate for unclear ownership become removable the moment ownership is clear — a genuine subtraction with no replacement needed, which is rare enough in organizational change to be worth going after first.

How it goes wrong. Ownership gets “clarified” by writing it down. Someone produces a responsibility matrix, circulates it, and nothing changes — because a document does not create the condition it describes. Ownership is not a claim about who is accountable. It is a state in which the named person can decide without asking, knows it, and everyone around them knows it too.

The test is not whether a name exists. It is whether that person has ever overruled someone more senior on the thing they supposedly own, and had it stand. If not, the ownership is decorative.

What to do instead. Name the owner out loud, in the room, in front of the people it affects. Then take a decision that used to escalate and let it not escalate. The second half is the entire exercise; the first half is a memo.

Second: move the decision to the information

Not the information to the decision.

Most organizations try it the other way, and it is worth being precise about why that fails. Improving reporting so a distant decider is better informed adds a step to a chain that was already too long. It also cannot work in principle: the summary that reaches the decider is always smaller than what the person on the ground knew, and the difference between them is exactly the judgement you were hoping to buy.

How it goes wrong. Authority is delegated in a memo and retracted in practice. The first time a delegated decision produces a bad outcome it quietly returns upward, and everyone learns that the delegation was conditional on being right. One retraction costs more than ten delegations build.

What to do instead. Delegate the decision together with an explicit tolerance for it going wrong, stated in advance and in public: this is yours, some of these will be wrong, and a wrong one does not send it back to me. Then survive the first bad one without taking it back. There is no other way to do this part.

A test worth keeping. Ask how little context someone needs to make this decision well — not how much you could send them. If the honest answer is a great deal, the decision is still sitting in the wrong place, and no amount of briefing will move it. The same test holds for an agent: the context it has to be handed before it can act is a fair reading of how far the decision still is from the information.

Third, and only then: scale with technology

The question from Technology applies: what capability does this multiply, and is that capability currently good enough to be worth multiplying?

How it goes wrong. This step gets done first, because it is the only one of the three that can be procured. A tool is a decision you can make in a quarter with a budget; the other two require confronting people about authority. So the tool arrives, the friction stays, and the organization concludes — reasonably, from the evidence in front of it — that the tool didn’t work.

When the boundary will not move

Some boundaries are fixed. A regulator requires them, a joint venture created them, an acquisition never finished integrating, or someone’s position depends on the boundary staying exactly where it is. Pretending otherwise costs a year.

When a boundary can’t be removed, the goal changes from removing the handover to making it cheap:

• Make the interface explicit. If work must cross, publish how — documented to the standard you’d use for an outside consumer.
• Move the wait off the critical path. A handover that happens in parallel costs elapsed time once, rather than every time.
• Reduce the frequency rather than the cost. Ten cheap crossings can be worse than two expensive ones.
• Name it as a known cost. Put a number on it and revisit it annually. Boundaries nobody questions outlive their reason by a decade.

What improvement feels like from the inside

Nothing dramatic, and people should be warned about that in advance.

An organization that has removed real friction does not feel fast. It feels ordinary. A meeting stops being scheduled and nobody remarks on it. A decision gets made in the room where the problem came up. Someone notices that a thing they used to dread has become unremarkable.

The absence of a wait doesn’t announce itself, which makes improvement nearly impossible to celebrate and very easy to under-claim. That is precisely why the numbers matter. Without a before and an after, nobody will believe anything happened — including the people who did it.

Say it back before anyone leaves the room

Everyone Left Agreeing is the cheapest friction in this paper to remove and the most reliably skipped, because removing it costs a small amount of social nerve at exactly the moment nobody has any spare.

Say it back. Not a summary of the discussion — a statement of what you are going to go and do, in your own words, out loud, while everyone who could correct you is still present. So I’m going to do X, and I’m assuming Y is out of scope. Half the time everyone nods and you have lost ninety seconds. The other half, somebody says wait, Y is definitely in scope, and you have recovered three weeks.

Write down what was assumed, not only what was decided. Decisions get minuted almost everywhere. Assumptions almost never do, and assumptions are the part that diverges. A decision record with no assumptions in it tells a future reader what was chosen and nothing about the world in which choosing it made sense.

Treat “we’re aligned” as a hypothesis. It is an accurate description of how a conversation felt. It is not evidence that eight mental models match, and it is most likely to be said in exactly the meetings where they don’t, because a discussion with no friction in it is a discussion where nobody surfaced their picture.

The reason this works better from the top of the room is worth stating plainly: whoever has the most standing pays the smallest social cost for going first. If you are the senior person present and you say it back before anyone else does, you have made it free for everyone else. If you never do, you have made it expensive, whatever you have said about psychological safety.

All of which recovers capacity, and recovers it permanently. It says nothing whatsoever about what the capacity should then be spent on, and an organization that gets this half right and stops here has built a very efficient way of arriving somewhere it never chose.

This cultivates: Everyone Left Agreeing, in Part II.