The Big Factory

There is a mountain in the Alps that Maria’s company put on a brochure once, a clean triangular peak, the kind a child draws. The brochure was making a point about permanence. We have been here for over a century, it said, more or less. Solid as the mountain.

The comparison is better than the brochure knew, though not in the way it intended.

Because the mountain is not permanent, and it is not solid, and it is not particularly a monument to the dramatic event that made it. The peak everyone photographs was thrust up in a geological instant, a violent spasm of tectonic collision, the kind of event that gets into the textbooks. But that thrust is not what gives the mountain its shape today. What shapes it now, continuously, is the opposite of an event. It is erosion: rain, ice, wind and gravity, removing a few grains at a time, every second, for millions of years. The dramatic uplift built the raw mass. The invisible grinding carved everything you would recognise.

Nobody watches the erosion. You cannot watch it, because it moves too slowly to see. If you stood on that peak for your entire life and stared, you would perceive nothing changing at all. And yet the erosion is winning, and it always wins. Every mountain range has a life cycle that ends the same way, with the uplift brief and finite, the grinding patient and relentless, and given enough time the grinding takes the whole thing back down to a plain.

A large manufacturing company is a mountain in exactly this sense. Its shape is made by two forces on wildly different timescales, and it systematically mistakes which one matters.

The uplift

Let me start with the force everyone can see, because it is the one the whole organisation is built to serve.

A car plant is a staggering object. Not just large, but committed. When a manufacturer decides to build or retool a factory it is spending a sum with a lot of zeroes, on equipment that will be bolted to a specific floor in a specific country for a specific purpose, and that cannot meaningfully be undone. You do not half-build a body shop. You do not retool a paint line as an experiment and roll it back next quarter if it disappoints. The press that stamps door panels weighs as much as a house and was designed around the exact geometry of a car that will be in production for years.

These are the uplift events. Enormous, rare, dramatic, and irreversible, which is the word that will run through the rest of the chapter.

The economists have a precise language for why irreversibility changes everything, and it is worth having, because it explains behaviour that otherwise looks like simple cowardice. In the late 1980s and early 1990s, Avinash Dixit and Robert Pindyck worked out the formal consequences of investing under uncertainty when you cannot take the investment back. Their key insight is deceptively simple: when an investment is irreversible, committing to it kills your option to wait, and that option has value. The ability to hold off, to see how things develop, to not be locked in — that flexibility is worth something real, and an irreversible commitment spends it. So a rational manager facing a one-way door does not just weigh the cost against the return. They add a third term, the value of the option they are about to destroy by walking through it, which means they demand a much higher return before committing and they wait a good deal longer than a naive calculation says they should.

You can measure this waiting. Firms use hurdle rates, the minimum return a project must clear to get funded, and the striking, well-replicated finding is that these rates sit far above what the financial theory says they should. Companies routinely demand returns three or four times their actual cost of capital before they will commit, and as we saw in Chapter 2, in one of the sturdiest findings in the whole book, those rates barely move even when interest rates collapse. The bar is high, it stays high, and it exists precisely because the doors only swing one way.

There is a second reason the big commitment is the only kind that feels natural, and it is physical rather than financial. Investment in a factory is lumpy. It does not flow in smoothly, it arrives in enormous discrete lumps. The plant-level data on this is unambiguous. A manufacturing plant will go years changing very little and then, in a single concentrated burst, replace a huge fraction of its capital at once. Doms and Dunne, studying thousands of US plants, found that around half of a plant’s investment over a fifteen-year period was compressed into just two or three consecutive years, with the rest of the time spent in relative stillness. The factory does not evolve continuously. It sits, and then it heaves.

Put irreversibility and lumpiness together and you get an organisation whose entire decision-making apparatus is shaped around the big, rare, one-way bet. That is what its senior people are trained to evaluate. That is what the committees, the hurdle rates, the capital-approval processes and the whole expensive machinery of governance exist to handle. And it is right to exist. I want to be very clear about this, because the easy version of this book would cast the big-bet machinery as the villain and it isn’t. When the decision genuinely is a house-sized press bolted to a floor for a decade, you should deliberate for months, demand a high return, and guard the door. That caution is correct.

The problem is not the caution. The problem is where the caution gets applied.

The one-way door and the two-way door

Here is a distinction that will do a lot of work for the rest of the book, and it comes, as it happens, from the world Maria’s company is now being told to imitate. Jeff Bezos set it out in Amazon’s 2015 letter to shareholders, and it has been quoted so often since that a good many of the people using it no longer know whose it was.

The idea is simple enough. Some decisions are one-way doors, hard or impossible to reverse, so you had better be sure before you walk through. Others are two-way doors, and if it doesn’t work you walk back through and the cost of having tried is small. The whole art, the framing says, is telling them apart, because the two kinds of door deserve completely different treatment. One-way doors want deliberation. Two-way doors want speed, because the cost of hesitating exceeds the cost of a reversible mistake.

Now hold Maria’s tool up against that distinction. Six hours to build, eighteen euros a month to run, on her own account. If it had failed she would have closed the browser tab and lost an afternoon. It is a two-way door, and close to the purest two-way door imaginable: nearly free to try, nearly free to reverse, nearly free to walk back out of.

And here is the failure, stated as plainly as I can. The organisation only owns one kind of door-handling, and it is the wrong kind for Maria.

The manufacturer’s every instinct, every process, every trained reflex is built for one-way doors, because that is what factories are made of. So when a two-way-door idea like Maria’s approaches the machinery, it gets one-way-door treatment. It gets the deliberation, the demand for a business case, the request to prove the return, the alignment review and the wait, which is all apparatus that is correct for a house-sized press and catastrophic for a six-hour tool. You cannot ask a two-way door to clear a one-way-door hurdle rate. The tool that would have paid for itself in a fortnight is asked to justify itself as though it were a capital project, and it can’t, and it dies. Not because the caution is wrong, but because the caution was built for a different kind of decision entirely and nobody ever separated the two.

This is the tilt Chapter 4 kept pointing at. The organisation protects the big and the owned, not out of malice or even preference, but because its entire decision-making muscle was built around irreversible bets and it applies that muscle to everything, including the reversible things that should never have been near it.

The resource nobody budgets

Capital is the tilt everyone can at least see. There is a second one, subtler and in some ways more decisive, and it explains something Chapter 1 left hanging.

Remember Maria’s rival, the Quality Systems programme with eighteen months, two consultancies, a budget quoted in disbelief, and a roadmap slide that was, everyone agreed, very good. Nothing shipped to her plant. Maria’s tool worked. And yet in any contest between them the programme would win effortlessly, without the two ever being formally compared. Why?

The answer is a resource that never appears on a budget line but is scarcer than money, which is attention.

Chapter 1 set out the mechanics: Ocasio’s finding that a firm’s behaviour is governed by how it allocates the attention of its decision-makers, and Bouquet and Birkinshaw’s two routes to being noticed — weight, which is structural, and voice, which is effortful and can be spent. What matters here is the part that becomes specific to a factory.

Weight in a manufacturer is not a metaphor. It is the same lumpiness this chapter has just described, expressed in a room rather than in a capital plan. The Quality Systems programme carries budget, consultancies, an executive sponsor and a place on the transformation roadmap, and every one of those is a claim on the capital process that the previous section showed is built to serve exactly that shape. The mass that gets it through the gate is the mass that holds the room. It is one property doing both jobs.

Which is why the two were never in a fair fight, and never in a fight at all. A fight would require both to be in the room, weighed against each other. This is the sharper and crueller version of what Chapter 4 called territory: the increment does not lose the argument, because there is no argument. It is not defeated, it is unseen, and you cannot appeal a decision that was never made.

That is why the roadmap slide can be excellent for three years while nothing ships. The slide is not there to produce an outcome. It is there to hold the weight, to keep the programme’s mass parked in the centre of the room’s attention, and it succeeds at that completely, which is the only thing it was ever measured on.

The clock

There is a third tilt, smaller than the other two but worth naming because it compounds them.

Manufacturers run on cycles. Budget cycles, model-year cycles, planning cycles, the great slow rhythm of a product that takes years to design and years to build and years to sell. This is not a flaw, it is a necessity given how long the physical bets take to pay back. You cannot run a car programme on a quarterly whim.

But innovation does not keep that time. A good idea does not arrive in Q3 aligned to the planning calendar. It arrives when it arrives, half-formed, needing a few weeks of quiet iteration to become anything, which is the delayed-assessment window from Chapter 1. And the cyclical organisation has nowhere to put a thing like that. It has a slot for next year’s capital plan and a slot for this cycle’s approved projects and no slot at all for a small useful thing that showed up in the wrong month and needs six weeks to prove itself. So the idea waits for a cycle that has no room for it, and by the time a cycle comes round that might, the moment has passed, and often so has the person’s willingness to raise it again.

Capital tilts toward the big, attention tilts toward the heavy, and the clock tilts toward the scheduled. Three separate forces, three separate mechanisms, all pointing away from the small, the light, the unscheduled and the reversible, which is to say away from exactly the kind of idea most likely to be trapped in the fleet.

The convergence, and a caution

I want to step back and name what has just happened, because it is the spine of the book assembling itself, and then I want to undercut it slightly, because that is the deal we made.

Across the last two chapters three scarce resources have each turned out to be biased in the same direction. Capital, because irreversibility and lumpiness make the big rare bet the only thing the machinery is built to handle. Attention, because weight pulls it toward mass and it is zero-sum. And careers and status, from Chapter 4 and returning in force in Part III, because the people who get promoted to allocate the first two are, as we will see, disproportionately the ones drawn to the big visible bet.

Three different currencies, and three different bodies of research that as far as I can find have never been introduced to one another. The investment-irreversibility economists do not cite the attention-allocation strategists, who do not cite the leadership-selection psychologists. And all three, independently, bend the same way. That convergence — the claim that these separate forces compose into a single systematic tilt against incremental innovation — is the argument of this book and it is mine. The individual findings are documented and I have cited them. The synthesis is the thing I am adding, and you should hold it as a strong argument rather than a proven fact, because no study proves it. It is an assembly.

And now the caution, which is real and which I would be dishonest to skip.

The whole capital story rests on the premise that manufacturing investment is enormous, lumpy and irreversible. That has been overwhelmingly true for a century. But it may be becoming less true at the edges, right now. Flexible automation, additive manufacturing, modular and software-defined production lines are slowly making some manufacturing capital less monolithic and less irreversible than it was. A line you can reconfigure in software is not quite the one-way door a welded body shop is. If that trend runs far enough, the capital tilt I have just described could soften.

I do not think it undoes the argument, for two reasons. First, the softening is partial and uneven, because the house-sized press is not going anywhere soon and most of manufacturing is still built around genuinely irreversible bets. Second, and more importantly, capital is only one of the three tilts. Even if flexible manufacturing dissolved the capital story entirely, attention would still be zero-sum and would still flow to weight, and careers would still reward the big bet. The mountain would still erode in a direction the organisation cannot see. But I flag it honestly, because the strongest single leg of the capital argument is the one with a genuine question mark hanging over its future, and a reader who works in the most advanced and most flexible corners of manufacturing may already feel the floor shifting under it.

Which, if you want the optimistic reading, is rather the point of the whole book. The forces that bury the increment are not laws of nature. They are features of how organisations happen to be built, and things that are built can be built differently.

Back to erosion

Return to the mountain, because now the metaphor can carry its full weight.

The manufacturer is built, entirely and rationally, to serve the uplift, meaning the rare, enormous, irreversible bet. That is where its capital goes, where its attention pools, where its calendar bends, where its heroes are made. And the uplift matters, because it built the mass. But the uplift is not what determines the company’s shape over time, any more than the tectonic thrust determines the mountain’s. What determines the shape in the end is the erosion: the continuous, invisible, grain-by-grain accumulation of small changes. The incremental improvements. The learning. The thousand tiny adaptations that compound, unwatched, into the difference between a company that is quietly getting better every day and one that is quietly getting worse.

And the organisation cannot see its own erosion. It has no instrument for it. Every instrument it owns — every committee, hurdle rate, roadmap and budget cycle — is calibrated for uplift, for the discrete visible event. The erosion happens below the resolution of everything the company uses to perceive itself, which means the single most decisive force acting on a manufacturer over time is the one force its entire apparatus is structurally blind to.

That blindness has a name we have been circling for four chapters. It is the fleet. The erosion is the submerged innovation, the small compounding improvements happening constantly, below the waterline, in the accounts and afternoons of people like Maria, entirely unseen by an organisation whose every sense is tuned to watch for mountains being thrust into the sky.

We have now seen the machinery, and why it is built the way it is, and why every part of it is individually reasonable. What we have not asked is the question that should be nagging by now, the one a sceptic would have raised chapters ago. If all of this is so well understood, if the erosion is so valuable, if the fixes seem so obvious, then why in sixty years has nobody fixed it?

That question turns out to have a darker answer than the machinery does. It is the whole of Part III.

Next: Chapter 6 — the increment that would pay for itself, and the arithmetic that means nobody will ever find out.