Scale changes what you know
There is a specific kind of failure that I have seen in capable people, and that I have committed myself, often enough to recognize it on sight. Someone who has been genuinely good at a job gets a bigger one — more teams, more products, more geographies, more consequences — and, for a while, gets worse. Not because they lack ability, and not because the new job is beyond them. Because they bring the mental model of the previous level into the next one, and the model keeps producing answers that were right there and are wrong here. They can't see it, because they are looking from the wrong place.
I want to describe what this looks like from the inside, because from the outside it just looks like someone struggling, and that is not what is happening.
The thing that stops working
Every significant change of scale I've been through — and I've been through several, from one product to many, from one site to several, from building things to building the things other people build with — has had the same shape. The first few months felt like doing the old job with more of everything. More meetings, more decisions, more people. Tiring, but familiar. And then, at some point, a thing I had always been good at stopped producing the right result.
The first time, the thing was knowing the details. At a smaller scale, knowing the details was my advantage: I could answer any question, catch any error, make any call with confidence because I had the whole picture in my head. At the larger scale, trying to keep the whole picture in my head meant I was the bottleneck for every decision, and the decisions I made with partial detail — because detail was now partial, necessarily — were worse than the ones the people closer to the work would have made. My strength had become a tax on the organization, and I was the last to notice, because from where I stood it still felt like competence.
The second time, the thing was optimizing locally. In a single product, making the best decision for that product is the job. Across many products, the best decision for one is routinely the wrong decision for the whole: it creates an exception, a divergence, a thing that has to be special-cased forever. What had been speed became incoherence. The reflex that had made me effective — solve this, here, now, well — was now the thing I had to suppress, and suppressing a reflex that has been rewarded for a decade is not a matter of deciding to.
I could go on. Each shift had its own version. What they had in common was that the competence that failed was not a weak one. It was a strong one, applied one level too high.
Why it's invisible
This is the part I think is least understood, including by the people it is happening to. The failure is invisible from the inside for a precise reason: the old model still works for most things. Most decisions at the new scale are, in fact, similar enough to decisions at the old scale that the old reflexes give the right answer. So you get confirmation, constantly, that you know what you're doing. The cases where the model fails are a minority, and they don't announce themselves as model failures. They look like isolated problems: a team that isn't performing, a project that keeps slipping, a conflict that won't resolve. You address each one as a problem. You don't see that they share a cause, and that the cause is you — or more precisely, the version of you that was right at the previous level.
The people around you often can't tell you, either. The ones below you see the symptoms but don't have the altitude to name the cause. The ones above you see the outcomes but not the mechanism. And the ones at your level are usually going through their own version of the same thing.
So the discovery tends to come late, and it tends to come from the pattern: the third or fourth "isolated" problem that, looked at together with the others, turns out not to be isolated. The skill — and I do think it is a learnable skill — is shortening the time between the first symptom and the recognition.
What has to change
The word I keep coming back to is restructuring, as opposed to updating. Updating is what you do when the job gets bigger but stays the same shape: you learn the new names, the new systems, the new numbers. Restructuring is what you do when the shape changes: you have to take apart how you reason and reassemble it around a different question.
At one scale, the question is "what is the right thing to do here?" At the next, it is "what is the right way to decide things like this, so that people who aren't me get it right?" At the next, it is "what is the right structure, so that the right way to decide emerges without anyone having to be told?" Each of those is a real question with real answers, and each makes the previous question somebody else's.
That last part is the hard one. Restructuring means giving up the question you were good at. Not delegating it — giving it up, which means accepting that it will sometimes be answered worse than you would have answered it, and that this is correct. The alternative is to keep answering it yourself, which feels like diligence and is actually the failure mode described above, dressed as responsibility.
I have not found a way to make this comfortable. I have found that it goes faster when you expect it: when you take the bigger job already knowing that something you're good at is about to stop working, and you spend the first months watching for which one, rather than assuming your competence transfers intact.
Three questions
For what it's worth, these are the questions I now ask myself — and, when it's my place to, the people I work with — at every change of scale. They are not profound. They are just the ones that have found the problem fastest.
What am I still doing that someone else should now be doing? Not because I'm too busy, but because my doing it prevents them from building the competence the organization needs them to have. This finds the bottleneck-by-strength problem.
Which of my instincts is producing the right answer for the part and the wrong answer for the whole? This finds the local-optimization problem, and it is uncomfortable because the instinct in question is usually one you're proud of.
What am I measuring success by, and is it the measure for this level or the previous one? This one is the deepest. Every level has its own definition of "good," and the previous definition does not go away on its own. It stays in your head, feeling like the truth, long after it has become the wrong yardstick.
Not a slope
The common picture of professional growth is a slope: you accumulate, you get better, the line goes up. I think that picture is wrong in a way that costs people years. Growth, at least past a certain point, is not accumulation. It is a sequence of restructurings, each of which starts with something that used to work no longer working, and each of which requires giving up a kind of competence in order to build the next kind.
People who understand this treat the discomfort of a new level as information rather than as a problem. They go looking for the thing that has stopped working, instead of waiting for it to find them. And they are, in my experience, the ones who keep growing after most people have plateaued — not because they are more talented, but because they have stopped expecting the old model to carry them, and started expecting to rebuild it.