Nine a.m. Three people want you at once.

One says the server is down. One needs you to pick a vendor. The third just wants to know whether the afternoon meeting is still on.

Who do you deal with first?

By evening you feel like a top, and when you look back, nothing seems to have been finished.

This is not because you are bad at your job.

It is because nobody told you that a manager has a production line too.

The book is High Output Management, by Andy Grove, published in nineteen eighty three.

This piece has four parts.

First, how Grove thinks about leading, meetings, decisions, pay and training.

Second, why he claims any job can be split into three steps, and where that split stops working.

Third, how the book’s eight models should be ordered.

Fourth, two things I am still unsure about.

Part one: how he actually manages

Start with the conclusion: your output is not what you do

This is the foundation of the whole book.

Grove gives a formula.

A manager’s output equals the output of the units he directly runs, plus the output of the units he influences.

The first time you read it, it sounds like a truism.

Read it again and it stops being funny.

Because it takes “I was very busy today” and rules it as non-work.

〔图 1：经理人的产出＝直接管辖＋间接影响，以及提高产出的三条路〕

Think about it. A cook can only cook his own pan, however fast he is.

A foreman who starts carrying bricks himself has nobody left to schedule the line.

The server you rescued should have been rescued by whoever was on call.

The call you made should have been made by whoever knows that business best.

Stepping in looks efficient.

The cost is that the line never grows a second person.

So should you let go of everything? No.

The book asks you to change the question: does what I just did affect one person, or a group of people?

That question comes back later. The book has a name for it: leverage.

The breakfast shop comes back again and again.

A breakfast has three parts: boil an egg, toast bread, pour coffee.

The egg takes three minutes, the bread one, the coffee fifteen seconds.

So what does the whole line schedule around?

The slowest one: the egg.

Grove calls it the limiting step.

〔图 2：早餐店的三道工序与限制步骤〕

The bread and coffee are not unimportant. They have to fit into the egg’s gaps, which he calls offsetting.

Applied to you, it stings a little.

What is the real egg in your day?

Not the people who come to you.

It is the one thing only you can do, the thing everything else waits on.

Most people’s problem is not a lack of effort.

It is that they have mistaken firefighting for the egg.

Firefighting gives instant feedback. The egg does not.

So people drift toward firefighting, because every rescue confirms that they matter.

Meetings: a medium, not an activity

The book is blunt about meetings.

A meeting is where a manager does his work. It is not the work itself.

In his words, meetings give managers the occasion for their managerial activity, but a meeting is not an activity, only a medium.

He also puts a price on it.

A manager’s time costs about a hundred dollars an hour.

Ten managers for two hours is a two-thousand-dollar meeting.

So the question is not whether to hold meetings.

It is whether this one is worth two thousand dollars.

He has a hard rule too.

Decision meetings stop moving past six or seven people. Eight is the ceiling.

Think about your weekly meeting with fifteen people in it.

Two hours, twenty people, four thousand dollars.

The output is usually a memo nobody reads.

So should you kill them all? No.

Some meetings can become three lines of text. Some cannot be cut.

The test is not importance. It is whether information actually moves in the room.

Decisions: three stages, six questions

The book splits a decision into three stages.

Free discussion, a clear decision, full execution.

The third one is the counterintuitive part.

It does not require everyone to genuinely agree. It requires that they can live with it and then execute fully.

That sounds like fence-sitting.

What the book is doing is separating consensus from execution.

If you need everyone to nod, the meeting never ends.

If you only need everyone to execute, the meeting gets short.

Nodding is hard. Executing is easier.

In practice he gives six questions.

What is being decided, and by when.

Who is the decision-maker.

Who should be consulted first, who can veto, and who must be told afterwards.

Ask those six and half your meetings disappear.

Why?

Because half the argument is about who has the authority, and that should have been settled in advance.

There is another phrase I like: the peer group plus one.

A meeting of equals tends to stall, everyone probing, nobody willing to speak first.

Bring in one person slightly senior to run it, and the discussion starts moving.

That is not politics.

It is admitting that people freeze in front of their peers.

Pay and training: the two easiest things to get backwards

Grove is unsparing about performance reviews.

He says it is hard because our culture teaches us to avoid conflict, and a review requires saying things to someone’s face.

He also makes an observation that made me pause.

Managers write long improvement notes for poor performers, and a few lines of praise for the “stars of tomorrow”.

The priority is exactly backwards, because most of the company’s results come from those people.

You can write pages for a poor performer and he probably will not become someone else.

One sentence of recognition can decide whether the person on the way up stays.

He separates money just as cleanly.

Reviews are reviews. Raises are raises.

Why keep them apart?

Because once they are mixed, the person only hears the money.

And if they only hear the money, they never hear what to change.

He also does the arithmetic.

For a senior manager, performance bonus can be half of total pay. For a middle manager, roughly ten to twenty-five percent.

The higher the share, the more the company is using money to move behaviour.

It should not be treated as part of salary, and certainly not as a score.

Treat it as a score and the team starts performing for the score.

Training, meanwhile, he treats as one of the highest-leverage actions available.

He spent sixteen hours preparing four classes.

Ten people in his department, about twenty thousand working hours over the coming year.

One percent better performance is two hundred hours.

Do that arithmetic once and it is obvious: sixteen hours to buy two hundred.

Which manager would refuse that trade?

Most do, in practice.

Because the sixteen hours are paid now and the two hundred arrive a year later.

Part two: how far to split

Three steps: making, assembling, testing

The book says any job can be broken into three basic actions.

Making changes the thing itself. Assembling puts parts together into something new.

Testing is checking whether it is right.

Boiling an egg is making. Putting egg, bread and coffee on a plate is assembling.

Looking at the coffee to see if it has brewed is testing.

Selling a new product is assembling. Writing a compiler is assembling, with different parts.

The classification feels almost too plain to be useful.

But it has one practical benefit.

Once the work is split, you immediately know where the checks go.

A bad egg should be caught before it goes into the pan, not at the customer’s table.

Checks come in two kinds.

One is customs: inspect everything, send rejects back to the previous step.

The other is a monitor: sample, compute a defect rate, let the line keep running.

His rule of thumb is to start with the cheap monitor.

Unless history shows that one defective unit slipping through causes a disaster.

It looks like first principles, but it is not

Be careful here.

Many people will equate “split into three steps” with first-principles thinking.

First principles says: do not reason from what everyone else does.

Go back to the basic facts and work it out from there.

The two do share something.

Both reject copying by analogy, and both require breaking a whole into units you can handle separately.

The difference is the purpose.

First principles asks what the most basic fact is, and splits until nothing is left, in order to rebuild.

Grove asks which step is slowest and which is most likely to fail.

He splits until the line can be scheduled and checked, then stops.

One seeks truth. The other seeks control.

So his split has a boundary.

The book never claims management reduces to physics.

It only claims that if you split the work, you can find the limiting step and the place to put a check.

That is the most honest thing about this book.

It does not pretend to be a science.

It is a craft for making output predictable.

So when does the split fail?

When the steps themselves cannot be named.

Write a strategy from scratch and try to split it into making, assembling and testing. The three steps you get are fake.

Force the split and all you get is a professional-looking flowchart.

After the split, watch out: metrics lie

Once the work is split, you want a number for each step. The book encourages that.

But metrics have a flaw.

Once a number becomes a target, it stops being a good measure.

The book’s answer is to pair them: if you cut inventory, you must watch stockouts at the same time.

Why?

Because inventory alone will fall while customers arrive and find empty shelves.

Watch a single number and people will push on that number while everything else collapses.

This is worth more today than in 1983.

We have far more metrics now, and they are far easier to game.

Measure lines of code and you get padded code.

Measure response time and you get people splitting tickets into fragments.

So splitting is not the end.

After you split, ask one more question: will this metric teach people to do the wrong thing?

Part three: ordering the eight models

Not a list of eight, a chain

Each model in the book holds up on its own.

The output formula, the limiting step, leverage, metrics, modes of control, maturity, one-on-ones, decisions, training.

Put them together and it gets confusing.

My reading is that they are not eight parallel items.

They are eight positions on one chain.

The chain starts with the output formula: define what “done” means.

Second is seeing: can you tell whether the line is moving without relying on your gut?

The book’s method is to open windows in the black box: leading indicators, linearity indicators, trend indicators, the staggered report.

The staggered report is the most useful of them.

Put last month’s forecast next to this month’s actual, every month, and you can see whether your judgement is sharpening or drifting.

Third is allocation: where your time goes.

That is leverage, how much output one managerial action moves.

He also warns that leverage can be negative.

Your pessimism spreads through the team, and the whole room slows down.

Delaying a decision is making a wrong one.

Taking the work back from a report is denying them the chance to grow.

Fourth is motivation: why people do what you ask.

The book gives three modes of control.

Free market, where self-interest governs. Contractual obligation, where rules and promises govern. Culture, where shared values govern.

Which one to use depends on two variables: how complex the work is, and how ambiguous.

The harder something is to quantify, the less you can rely on self-interest alone.

Fifth is leading: the same report needs different handling at different levels of maturity.

Maturity here means familiarity with this particular task, not the person’s overall calibre.

Low maturity, give clear instructions. Medium, talk it through. High, step back.

Sixth is feedback: a one-on-one should be run by the report, not by you.

Be honest in reviews, and keep reviews away from raises.

Seventh is decisions and meetings: the place where you and a group exchange information and set direction.

Eighth is capability supply, which is training.

〔图 3：八个模型串成一条产出链〕

If you can only start with three

Ranked by importance, I would order them like this.

First, the output formula. It decides what question you ask yourself every morning.

Second, leverage. It decides who gets your time.

Third, modes of control. It decides why you keep burning hours on “I told them and they didn’t listen”.

The other five are tools for those three.

What happens if you get the order wrong?

You become that person.

Meetings run well, one-on-ones are regular, review forms are beautifully filled in.

And the output of the line has not moved in three years.

Because the first three were never done, and everything after is decoration.

Part four: two things I am unsure about

Should managers teach, or is that half an expert teaching half an expert?

This is the sharpest question in the brief.

Grove is clear.

Training is a manager’s job, not HR’s.

He points to his own sixteen hours of preparation.

But the division of labour is far finer now than in 1983.

A manager teaching an algorithm engineer easily becomes “this is how we did it in my day”.

My view is that it depends on what is being taught.

Judgement can be taught: when to use this approach, what it costs, where to look first when it breaks.

Those are things a manager buys with tuition every week.

Technique cannot be taught by a manager. The specialist knows the implementation better.

Teach it anyway and you pass on outdated habits.

So the question should change.

It is not whether managers should teach.

It is where the boundary is.

The boundary is: experience and judgement are yours, frontier and detail belong to the specialist.

The book leaves room for this.

What he stresses is teaching people how to do the work, not doing it for them.

A 1983 book: what still holds

Start with what has expired.

The setting is a semiconductor plant with production lines, workshops, everyone on one floor.

Remote work, time zones, knowledge work: none of it is tested.

The book says a manager who leads people should have six to eight reports.

In today’s matrix organisations, that number needs a discount.

The money is all in 1983 dollars. A hundred dollars an hour is several times that now.

Now what has not expired.

First, treating management as a production system.

That view is still rare. Most people still discuss a quantifiable thing using words like leadership, vision, execution.

Second, leverage.

The more information and the more fragmented the collaboration, the more this metric matters.

Third, the three modes of control.

The most common mistake in remote teams is assuming a policy is enough.

They forget that contracts need trust behind them, and culture needs shared interest behind it.

Fourth, keeping reviews and pay apart.

This is rarer today, because many companies have tied them tighter.

〔图 4：1983 与今天，什么还成立、什么过期了〕

One last judgement of mine.

What is genuinely ahead of its time in this book is not a prediction.

It is that management was turned from a personality trait into a craft that can be taught, checked and costed.

Forty years on, the technology has turned over several times. That part has not changed.

Afterword

Back to nine a.m.

Three people want you at once. Who first?

Grove’s answer is not “whichever is most urgent”.

It is whichever is closest to your limiting step.

The other two either get taught to someone else, or wait twenty minutes.

You are still leading the same three people.

What changed is that they may no longer sit next to you.

Management is not finishing the work.

It is letting the output pass through you.