Promoted for the Big Bet

In month eight, Maria read the attendee list. She had not looked at it when the invitation arrived, because nobody reads attendee lists, and she only opened it on the Thursday before for no better reason than that she was avoiding something else. There were fourteen names. Most of them she could place — her manager, two other quality leads, somebody from IT whose name she had seen on ticket responses and never in a room, three people from the Quality Systems programme, which she noted without feeling much, having sat through three of their update presentations and knowing exactly what the roadmap slide looks like. And two-thirds of the way down, a name she has known for a long time.

He joined the year before she did, same function, same grade, same building, and for the first few years they were on more or less the same list of people. He is now two levels above her, in a role with the word excellence in the title, and he will be in the room on Tuesday. Maria can tell you the exact thing that put him there, because she was standing next to it when it happened.

Seven years ago the scrap figure on line four was wrong, and it had been wrong since a system migration, in a boring and specific way: rework was counted twice, once against the station that found it and once against the station that caused it, so line four had been carrying a scrap number about a third higher than reality for two years. Nobody had caught it, because nobody questions a number that has always looked like that. It looks like a property of line four. Maria caught it the way she catches most things, by noticing that a figure had not moved when it should have, and she took it to the man who owned the mapping, and he fixed it in an afternoon. That was a Thursday, and she remembers because she had been planning to leave early.

He caught it about three weeks later, on his own, coming at it from a different direction, and this is where the two of them stop being the same person. He did not take it to the mapping owner. He wrote it up, and he wrote it up as what it plainly also was — evidence that the migration had left errors of this kind all over the plant, that nobody had audited them, and that the plant had been making decisions for two years on numbers it had never checked. He proposed a data-integrity workstream and he got one. It ran for eighteen months with a steering group and a name and a colour on a slide, it found eleven more errors, two of which mattered a great deal, and when he presented it at a regional quality meeting he was promoted about four months later.

Maria has thought about this on and off for seven years and has never once concluded that he was wrong, which is the part worth being careful about. He was not exaggerating, he was not empire-building, and he was not taking her fix, because he did not know about her fix — by the time the workstream started, line four had been correct for a month and there was nothing anywhere recording who had made it correct or when. He had identified a real problem across the whole plant, one that Maria had walked straight past. She had fixed one number; he had noticed a class of failure. He was better than her at something, and the something was not engineering.

What he had was a version of the problem with a shape to it, the kind of shape you can hang a sponsor on, and a budget line, and a start date, and a slide with eleven errors on it. Maria’s version was fixed on a Thursday afternoon by a man who was going to be at his desk anyway, it produced nothing that anyone could point at afterwards, and there is nowhere in that building where it is written down.

So on Tuesday she will sit in a room with the person that story is about, and she has spent most of this week working out what she actually expects from him. He will not be hostile, because he has never been hostile to her about anything. He is good at his job, he is not stupid about tools, and he has spent two years asking in meetings why nothing moves faster. Which is the problem, laid out with unusual clarity in a corridor on a Thursday evening: if she brings the tool into that room, the outcome she should be afraid of is not that he dislikes it, but that he likes it.

She knows what he does with things he likes.

The lazy version of this chapter

There is a way to write this chapter that would be popular and that I am not going to write. In that version the reason nothing gets fixed is that the wrong people are in charge — they are vain, they want monuments, they care about their next job more than about the plant, and the small useful thing dies because it flatters nobody. It is a satisfying story, and most people who have worked in a large company have at least one senior figure in mind who fits it exactly.

I think it is mostly wrong, and worse than wrong, it is useless, because it aims the remedy at the wrong thing. If the problem is bad individuals then the answer is better individuals, and every organisation on earth has already tried that at enormous expense, with leadership development budgets that would fund a small research council.

The claim in this chapter is different and it is harder to live with. You can staff the whole chain of command with capable, decent, well-meaning people — people who would pass any character test you cared to set, who are not vain, who want the plant to do well — and the machinery from Part II still produces the same result, because what gets a person chosen and what makes a person right are two different things, and nothing in the choosing is able to tell them apart. None of that is a claim about anybody’s motives. It is a claim about what a promotion decision can see in the hour it has.

Two questions that look like one question

The finding this chapter rests on is a review of the research rather than a single study, and its power sits entirely in a distinction that sounds like hair-splitting until you hold it for a moment. In 2015 Emily Grijalva and colleagues published a piece in Personnel Psychology pulling together decades of work on narcissism and leadership, and they kept two outcomes separate. The first is emergence: does this person become a leader, get picked, rise, end up in charge. The second is effectiveness: once they are in charge, does it go well.

Narcissism predicts the first and does not predict the second.

The relationship with emergence is positive and it is not small. The relationship with effectiveness is near enough nothing, and once you allow for extraversion, which travels alongside narcissism and does much of the visible work, what is left looks like a curve rather than a line — some is useful, more is not more.

A word about that term, because it does damage every time it appears. What these studies measure is a normal-range personality trait rather than the clinical disorder: confidence in one’s own judgement, comfort with authority, a willingness to be the centre of a room. Plenty of people who score high on it are people you would call impressive. Very few are people you would call narcissists in conversation, and reading this literature as a list of bad people is the fastest way to learn nothing from it.

The result that should worry a selection panel came a few years earlier. Brunell and colleagues in 2008 ran the emergence question in group settings and found the high scorers rising to the top even where the people watching had been trained to assess leadership.

Training the observer did not remove the effect, which tells you the signal is not being missed through carelessness.

Cameron Anderson and colleagues supplied the reason in 2012, in work on what they called the status-enhancement account of overconfidence: overconfident people gained standing in their groups, and they gained it because the overconfidence was read as ability, and it was read as ability because nobody spotted it as overconfidence. The group was not tolerating a flaw it could see. It could not see it.

Tomas Chamorro-Premuzic assembled all of this for a general audience in 2019, and I will flag that book as what it is, a trade book drawing on other people’s data rather than a study, because his central line is still the cleanest statement of the mechanism I know: confidence gets used as a stand-in for competence, and the two have very little to do with each other.

Now set that against the room a promotion actually happens in — an hour of conversation, three or four busy people, a written record of achievements the candidate wrote themselves, and perhaps a reputation arriving second or third hand. Confidence is fully visible in an hour. Competence is not visible in an hour, and in a great many jobs it is not cleanly visible in three years, because the comparison does not exist and nobody ever gets to watch what the plant would have done under the other candidate.

So the panel does what every other mechanism in this book does, and takes the signal it can read in place of the one it cannot, because the one it can read is the only thing in the room.

That should be familiar by now. In Chapter 6 the benefit that could be forecast pushed out the benefit that could only be discovered. In Chapter 7 the cost of protected time was a budget line and the benefit was not, so the programme died in the first bad quarter. This chapter is the same trade in a third currency. It is one mistake showing up in three places, and the currency is never really the point — whatever can be measured easily will stand in for whatever cannot, and there is always something that can be measured easily.

What the selected people then do

Half of this chapter’s job is working out who gets picked. The other half is showing that the picking has consequences for the particular thing this book cares about, and here the evidence is older and much better funded, because it is finance research and finance research follows the money.

Michael Jensen’s 1986 paper on free cash flow is where the idea starts that managers over-invest and over-expand for reasons unconnected to the return, and it names the reason plainly: status, power, freedom of action and pay all rise with the size of the thing you run, so growth pays the person deciding whether or not it pays anybody else.

Ulrike Malmendier and Geoffrey Tate then measured it. Their work in 2005 and 2008 picked out overconfident chief executives by their habit of holding company stock options long past the point where a sensible person would have cashed them in, which is an elegant way to catch someone who believes their own forecasts, and found that those executives buy other companies far more often — roughly sixty-five per cent more likely to make an acquisition, more likely to destroy value when they did, and most of all where the firm was sitting on plenty of its own cash and therefore never had to convince an outside lender of anything.

Richard Roll had proposed the shape of this in 1986 as the hubris hypothesis, and Mathew Hayward and Donald Hambrick tested part of it in 1997 across a hundred and six large acquisitions, finding that signs of hubris in the chief executive predicted how much of a premium the company paid. The more certain the person, the more they paid.

Arijit Chatterjee and Hambrick ran the personality version directly in 2007, in a study that gets quoted more for its cleverness than for its result. They could not sit chief executives down with a questionnaire, so they built their measure out of the clues those executives leave behind: how prominent the chief executive’s photograph is in the annual report, how often they are named in press releases, how much they say “I” in interviews, and what they are paid next to the second-highest-paid person in the firm. Higher scores went with bolder strategy, grander gestures, and more and larger acquisitions.

The result I want is the last one, and it is the one that keeps this chapter honest. Performance under those executives swung much harder in both directions, bigger wins and bigger losses, and on average came out no better and no worse than under everybody else.

That removes the villain from the story completely. These are not people making bad decisions; they are people making risky ones, and risk is not a moral category. A firm that never takes a big swing is not obviously in better health than one that does.

But risk is not neutral for the person taking it, and this is where the individual and the company come apart. The good outcome of a big visible bet has a name on it, and the name is theirs. The bad outcome spreads out and lands nowhere in particular, because by the time it shows up the market had moved, the integration was harder than anyone expected, the supplier failed, and in any case they are running something else now. Same odds, very different personal payoff. Nobody has to be cynical for that to work, and nobody has to notice it either.

There is one more piece and it arrives with a warning attached, as promised in Chapter 2. Barry Staw’s 1976 study, the one everybody knows by its title about being knee-deep in the Big Muddy, found that people who were personally responsible for a failing course of action put more resources into it than people who had inherited the same failing course from somebody else, with self-justification as the driver.

It is among the most cited results in the field and it fits this chapter almost too neatly, which is exactly when a book should slow down, because it has a replication problem. Martens and Orzen went back to it experimentally in 2021 and did not reproduce it — no increase in spending after bad news — and related work has reported the opposite effect, where a larger initial commitment made carrying on less likely. The field evidence for escalation is broad, and the 2012 review by Sleesman and colleagues supports the phenomenon in general, but the famous experiment is disputed.

So it does light work here, which is all I need it for: it is one possible explanation for something Chapter 5 already established on other grounds, that a large programme does not give back its share of attention when it starts going badly. I am not going to build anything on it. A famous but shaky finding does not get to carry weight in this book just because it points where I was already heading.

Nobody is promoted for a programme that is still running

Here is the connection the last two chapters have been walking toward, and it is mine rather than anybody’s finding, so take it as an argument. A career is built out of events, and that word is carrying more weight than it looks. A launch is an event. A turnaround is an event. A restructuring, an acquisition, a new function stood up, a programme announced with a video from head office — all events, all with a date on them, all attached to a person, and all of them things you can write in the box on the form that asks what you have delivered.

Keeping something alive is a state rather than an event. It has no date and it leaves nothing behind, and the nearest thing to evidence it produces is an absence, which is that the thing did not stop, and an absence cannot go on a slide because the slide would have to say that nothing happened.

So run the sum that a competent, decent, entirely unvain senior person runs without ever putting it into words. There are two options. Launch a new improvement programme, which is visible straight away, belongs to them personally, and gets judged on the launching rather than on the outcome, since the outcome is four years out and they will be somewhere else by then.

Or defend the budget line of the programme their predecessor launched, which is a cost with no matching benefit — Chapter 7 established that much — in a meeting where they would have to argue for something they cannot evidence, on behalf of somebody who has left, and which if it works will produce results credited to whoever ships the thing.

Both of those are open to a person of perfectly good character. One of them is a career and the other one is a favour.

And this, I think, is the answer Chapter 7 could not give you. Improvement Time did not die because the plant manager stopped believing in it. It died because launching it was the event, the event had already happened, and there was nothing left in it for anybody. He was promoted for the launch, moved sites about a year later, and by then it had already become a thing that used to happen. Nobody in the history of that company has been promoted for a programme that was still running, and nobody has ever been promoted for the absence of a decline.

Which gives the graveyard its shape: those programmes were not killed by the people who launched them, but launched by the same mechanism that then gave nobody any reason to keep them.

The three currencies close here, and they close in one sentence. Money bends toward the big because big decisions cannot be undone, attention bends toward the big because size pulls the room, and careers bend toward the big because a career is a list of things you started and only the big ones count as things you started. Three separate mechanisms with nothing in common, drawn from three bodies of research that have never been introduced to one another, all leaning the same way. That convergence is the argument of this book, and all three legs of it are now on the floor.

Why the fix has a minimum size

There is a consequence of all this that turns up in every large organisation trying to take innovation seriously, and Chapter 7 walked past it in the Google material without stopping. Twenty per cent time decayed partly through a steady pulling of innovation back into elite units — Chapter 5’s big factory reasserting itself inside the one company supposed to be immune to it. With this chapter in hand you can see why it was always going to happen, and the reason is not the one I expected when I started looking.

Start with the leader, sincerely wanting more innovation, whose own contribution will be judged on events. Distributed increments produce no event at all. Four hundred people each improving their own work by some amount nobody measured has no launch date, cannot be photographed, and would be credited to four hundred people, none of them the leader. There is no version of it that reaches a slide except as a claim nobody can check.

A centre is an event. It has an address, a head count, a budget, a name, a director and a ribbon, and it can be shown to the board, to journalists, and to a candidate the company is trying to hire. So the honest version is that nobody chooses the centre over the distributed thing, because only one of the two is on the menu. A leader can pick the centre. There is no procedure by which a leader picks the distributed increment, because there is nothing there to pick — no proposal arrives, no budget is requested, and the thing it would produce has no name.

That much I had before I started writing this section. What I had wrong was the assumption that the centre is simply the wrong instrument, a clumsy tool aimed at delicate work. It is worse than that.

A centre has a cost line. Head count, floor space, tooling, a director’s salary, and all of it visible in the plan every year. Chapter 7 established what happens to a cost line with no matching benefit line, so the centre knows from its first week that it will have to show what it is for. Which means it needs work large enough to be worth forty people, because work that is not large enough to be worth forty people is an argument for having fewer of them.

So it acquires a floor. Not a written one — nobody circulates a memo setting the minimum size of an acceptable topic — but a floor all the same, visible in what gets taken on and what gets politely redirected. Below the floor sits every improvement of the kind this book is about: worth a day a week to one engineer, invisible on any reported number, and impossible to describe as a reason for the centre’s existence.

Which is the trap from Chapter 7 wearing different clothes. The fix is subject to the gate, and here it is subject to it twice over, because an increment brought to the centre now has to justify itself and justify the centre.

Maria’s tool cost eighteen euros a month and six hours of a Sunday. It has no threshold to clear, because there is nothing underneath it demanding to be paid for. Move it into a unit with forty people and a floor, and it becomes too small to be interesting to the only part of the organisation officially responsible for being interested in it.

And the distance makes it worse rather than better. Tyre and von Hippel showed in Chapter 6 that this kind of improvement is situated — it emerges where the problem is, from people who have the clues in front of them.

A centre is staffed by people taken off the line, which is what lets it be a centre at all. The same property that makes it visible enough to survive is the property that puts it out of reach of the work it was built to capture.

The same logic explains how a protection that already exists gets quietly starved without anybody starving it. Chapter 7 established that nobody decided to end Improvement Time, and this chapter explains why nobody had to. Ending it took no act at all. Keeping it would have taken a repeated act, every planning round, with nothing coming back the other way.

The case for the bold ones

I want to put the other side at its strongest, because it is stronger than this chapter has so far allowed and leaving it out would make the argument cheap. Confidence is not a defect. An organisation facing a real threat needs somebody willing to commit serious resources before the evidence is in, and the person able to do that is precisely the person who scores high on the traits described above. A firm whose leaders only ever acted on complete information would not be careful, it would be dead, and it would die slowly enough that everyone involved could explain why each separate delay had been prudent — which is more or less the thesis of the companion volume to this book.

Chatterjee and Hambrick’s average result points the same way. No worse on average, whatever else the big-bet temperament is, means it is not reliably destructive, and a book that read their finding as an indictment would be misreading it.

Adam Grant adds a wrinkle that matters here, with the same trade-book caveat attached. In Give and Take he separates people who claim more than they contribute from people who do the reverse, and finds the self-promoters doing well at getting picked and doing well in the short run, while over longer stretches the advantage turns around.

His most useful category is the person who is prickly and difficult and nonetheless acting in other people’s interests, who he argues is the most systematically undervalued profile in working life, because the visible signal, which is the manner, points the opposite way to the thing that matters.

Which is this whole chapter again, arriving from the other side. None of it makes the trait the problem. The problem is that the organisation has one way of choosing people, it runs on a signal available in an hour, and it therefore cannot separate four different people who look identical across a table. Which is why the remedy cannot be better people. It has to be a second channel, and Chapter 12 is where that gets built.

Where the evidence stops

Four limits, and the first is the biggest leap in the book.

Almost all of this evidence is about chief executives, and this chapter is about a plant. Jensen, Malmendier and Tate, Roll, Hayward and Hambrick, Chatterjee and Hambrick — every one of them studies the very top of very large listed companies, mostly through acquisitions, in samples weighted toward finance and technology. Nobody in that literature has looked at a promotion round in a manufacturing company, or a plant manager’s spending choice, or the person chairing a use-case review in a factory. The step from overconfident chief executives buy more companies to the man across the table on Tuesday will turn her tool into a programme is a leap. I think it is a fair one, because the mechanism underneath it — that the reward attaches to visible events with a name on them — is if anything stronger in the middle of a hierarchy than at the top, where there is at least a board and a share price pulling in the other direction. But it is a leap, it is doing real work here, and if you want to attack this chapter, that is where to attack it.

The evidence about who gets picked and the evidence about what they then do come from two different literatures. Grijalva and Anderson are psychologists measuring who rises, in samples that are largely not senior executives. Malmendier and Chatterjee are finance and strategy researchers measuring what senior executives do. This chapter joins the two: selection favours the trait, the trait produces big visible bets, so the deciders end up tilted. No study makes that join. It is the same kind of assembly as the three-currency argument, and it deserves the same standing — a strong case rather than a proven chain.

Staw is disputed, and the personality measures are stand-ins. The escalation caveat is above. Separately, the clues Chatterjee and Hambrick used — photograph size, counting the word “I”, pay ratios — are ingenious and they are still stand-ins, and it is fair to ask whether a large photograph in an annual report measures a person or measures the house style of a communications department in a particular decade.

And here is the study that would settle it. Take a large organisation with ten years of promotion records. For everyone promoted into a decision-making role in that time, score two things from the record: how many things they launched, and how many things they inherited, kept running and handed on working. Then test which of the two predicts promotion. If keeping things alive predicts advancement at anything like the rate launching does, this chapter is wrong, Chapter 7’s graveyard needs another explanation, and Part III loses a leg. My expectation is that launching predicts promotion strongly, that stewardship predicts nothing whatever, and that most organisations will find they cannot even fill in the second column because they never recorded it — and that last part is not an obstacle to the study but the first result of it. Every large company already holds the data for this. As far as I can find nobody has run it, and it would not require anybody to be asked anything.

Who she would be surfacing to

Eight chapters in, the room is finally fully described, and the thing to notice about it is that there is nobody in it to be angry with. The gate was reasonable, in Chapter 4. The caution about money was rational, in Chapter 5. The programme’s death needed no decision at all, in Chapter 7. And the person at the head of the table was chosen by a process doing its honest best with the only signal it could see in the time it had, and is now behaving exactly as chosen, which is what choosing means. Every step defensible, every step correct from where the person taking it was standing, and the whole thing adds up to a machine that reliably turns a working tool into an eighteen-month programme and reliably fails to notice that it has.

This is why the problem survives being well understood, which is the question Part III exists to answer, because a problem made of bad actors gets solved by removing them and a problem made of reasonable decisions has nothing to remove.

It also puts Maria’s Tuesday into its final shape. Her risk is not rejection — rejection would leave her exactly where she is, with a tool that works and eighteen euros a month. Her risk is adoption, that the tool gets taken up by the person best placed to take it up, who will do the only thing his position allows him to do with something he likes, which is to give it a sponsor and a scope and a name, and make it visible enough to be worth his time, and in doing so turn it from a thing that runs on a Sunday into a thing that has to justify a forecast it cannot produce.

She has watched him do exactly this once before, correctly, with a real problem, seven years ago. He was right that time, and that is what makes it so hard to argue with.

There is one thing left, and it is the piece this book has been walking around since Chapter 1. Three chapters have gone on what the organisation does to the idea, and none of them has asked what it does to the person — not on the Tuesday, but over eleven years, and not as a mood but as something she learned and has had no reason since to unlearn. That is Chapter 9, and it is the hardest one.

Next: Chapter 9 — what all of this does to the person, which turns out not to be a feeling.