Knowing when to leave
Leaving a role that isn’t working is rarely a failure of analysis. People are generally good at reading their own situations. What’s missing is permission to act on a conclusion they’ve already reached — and the psychology of withholding that permission is the difference between leaving at the right time and leaving years too late.
Most of the coaching conversations I have about leaving aren’t really about whether to go. By the time someone books the call, some part of them has usually already decided. What they’re actually looking for is permission — permission to act on a conclusion they’ve quietly reached and can’t yet say out loud, even to themselves.
That’s worth naming up front, because it reframes the whole problem. Leaving a role that isn’t working is rarely a failure of analysis. People are generally good at reading their own situations. It’s a failure of permission — and understanding why we withhold that permission from ourselves is the difference between leaving at the right time and leaving two years too late, ground down and out of options.
Why we stay too long
The psychology here is well-mapped, and it’s worth knowing the mechanism because the mechanism is what traps you.
The foundational work is Barry Staw’s 1976 study with the wonderful title “Knee deep in the big muddy.” Staw put people in simulated management roles, had them make an investment decision, then fed some of them results rigged to fail — and watched what they did next. The striking finding: the people who had personally made the original decision were far more likely to keep committing resources to the failing course than people who’d merely inherited it. The driver wasn’t economics. It was self-justification. Backing out would mean the original call was wrong, and that’s the thing we’ll spend enormous resources to avoid admitting.
This is escalation of commitment, and it’s the engine underneath staying too long. Notice what it implies for senior people specifically: the more of a situation you personally authored — the systems you built, the team you hired, the strategy you set — the more powerfully the bias acts on you. The thing that makes you valuable in a role, your ownership of it, is the same thing that makes it hardest to leave. You’re not staying because the situation is good. You’re staying because leaving would retroactively classify your investment as a loss rather than as something that might still pay off.
There’s a second mechanism layered on top. Research on escalation finds that negative feedback tends to provoke action — and in a role you’ve committed to, the available action is almost always “try harder,” because that’s framed as doing something while leaving is framed as giving up. So bad news, paradoxically, makes you dig in. The worse it gets, the more energy you pour into the thing, because de-escalating doesn’t feel like a choice. It feels like quitting.
Put those together and you get the characteristic shape of staying too long: a capable person, deeply invested in a situation they largely built, responding to mounting evidence that it isn’t working by working harder, because every signal to leave is processed as a signal to try again.
A cleaner way to make the decision
The antidote to a self-justification bias is to make the decision on variables that aren’t about self-justification. The most useful framework I know comes from Caryl Rusbult’s investment model, which predicts whether people stay in or leave a commitment based on four factors. It was developed studying relationships, but it transfers almost directly to roles.
Satisfaction — the honest balance of good and bad in the role as it actually is, not as it was or as you hope it’ll become.
Investment size — what you’d lose by leaving. This is the sunk-cost trap wearing a respectable name. The thing to notice is that investment size measures what’s already spent, which is exactly the thing that should be irrelevant to a forward-looking decision and exactly the thing that dominates it.
Quality of alternatives — whether you’d genuinely be better off elsewhere. People who feel they have no alternatives stay in almost anything, which is why the most important career work often happens before the decision: building the optionality that makes leaving thinkable.
Commitment — how much you’re willing to keep working to preserve the situation.
The trap is that investment size and commitment — the two backward-looking, self-justifying variables — tend to drown out satisfaction and quality of alternatives, the two that actually predict whether staying is a good idea. When you catch yourself justifying a role primarily by what you’ve put into it and how hard you’re willing to fight for it, rather than by whether it’s good and whether you have somewhere better to go, that’s the bias talking.
The signals worth trusting
Frameworks are for the slow, deliberate version of the decision. In practice, there are a few signals I’ve learned to treat as load-bearing — both watching clients and, once, learning it the hard way myself.
I stayed in a role longer than I should have. The signs were there well before I acted on them: the same structural problem kept recurring no matter what I built around it, the people who could have changed it weren’t willing to, and I’d started measuring my contribution by effort rather than effect. By the next time I was in a comparable situation, I’d learned to read those signs faster — not because the second situation was clearer, but because I’d stopped needing it to be unambiguous before I’d let myself act.
The signals that generalise:
The problem is structural and the people who could change it won’t. This is the single most important one. If what’s broken sits above your authority to fix and the people with that authority have seen the case and declined to act, staying longer doesn’t change the structure. It just spends your years.
You’ve started measuring yourself by input rather than output. When you find yourself citing how hard you’re working rather than what’s actually changing, that’s escalation of commitment reporting for duty. Effort is what we point to when effect isn’t available.
The recurring problem keeps recurring. Not new problems — those are just work. The same problem, surviving every intervention, is the structure telling you it has no intention of changing.
You’re waiting for a single unambiguous sign. The wish for one clear, undeniable reason to leave is usually a way of withholding permission from yourself. The clean break rarely comes. If you’re waiting for the situation to make the decision for you, you’ve already made it and are declining to act on it.
Leaving as a skill
The reframe I’d leave you with is that knowing when to leave is a skill, not an admission of defeat — and like any skill it gets better with deliberate practice and worse with avoidance.
The person who leaves a structurally broken situation at the right time, having done good work, handed it over cleanly, and kept their alternatives alive, hasn’t failed. They’ve exercised judgement that most people’s biases actively prevent. The failure mode isn’t leaving. It’s staying knee-deep in the big muddy, working harder, waiting for a sign that was never going to come, until the years you’ve spent become the main reason you give for spending more of them.
If some part of you has already decided, the useful question isn’t “should I leave?” It’s “what am I waiting for permission to do, and who am I waiting for it from?” More often than not, the answer is yourself.
This is the fourth in a short series on the structural patterns that quietly break technical organisations. Earlier pieces covered the hero hire, the bridge that keeps two teams apart, and what founders really mean when they ask for a product team. The final piece pulls it together: the kind of company where all of these patterns show up at once.