The Carousel

When performance stalls, boards change leaders. Then they do it again. The problem is not always the person. It is the gate. The boardroom clock hits 6:42 p.m. and nobody looks at it because everybody already knows what time it is. It is late. It is always late. The decks are current, the numbers reconcile, the legal language is polished, and the operating review reads like a responsible document written by responsible adults. Still, the air carries that familiar dread that arrives when a team realizes the meeting is not about the meeting. It is about the explanation for why the explanation did not work. The CHRO, call her Dana, has the succession grid open on her laptop, but she is not looking at it. She is watching the CEO. The CEO is watching the CFO. The CFO is watching the head of the largest division, the one with the “reliable” results. Every face has learned the same modern skill. Speak in nouns that sound like control. Say “discipline,” say “accountability,” say “governance,” say “portfolio.” Nobody says the word everybody is thinking. The word is “stall.” The stall has a shape. A leader arrives, sometimes internal, sometimes imported. The board signals a new era without admitting the old one failed. The leader announces a plan with a name that photographs well. A new structure appears, then a new cadence, then a new set of KPIs, then a new bonus plan that “aligns incentives.” People work harder. Some costs come out. A quarter

or two looks better. The board relaxes just long enough to believe it again. Then the same problems return under a slightly different label, like a virus that learned the perimeter rules. That is when the carousel starts to spin faster. The division leader gets swapped. The COO role gets redefined. A high-visibility “transformation office” gets staffed. A few lieutenants get promoted because they can present cleanly and speak in metrics that feel safe. The organization learns the lesson it is always taught. Survive the measurement, not the mission. After a while, the only consistent outcome is rotation. What would have to be true for this outcome to keep repeating. The simple answer is also the most insulting answer, which is why so few boards say it out loud. The system is working exactly as designed. Most leaders believe a performance system is a mirror. It reflects who is strong and who is weak, who is ready and who is not, who should lead and who should follow. That belief is the source of the carousel. A performance system is not a mirror. It is a mold. It does not merely record behavior. It produces behavior. Once you accept that, the carousel stops looking like bad luck and starts looking like physics. You are paying for certain moves. People are making those moves. You are promoting the people who make them best. Those people arrive in bigger jobs and keep making the same moves because the system continues to reward them. The organization does not get better. It gets more skilled at presenting stability while pushing risk into places the dashboard does not see. The hard part is that the molds in most companies were built for a world that no longer exists. They were built for an era when outcomes were slower, information traveled with delay, and the organization could afford to treat “performance” as a quarterly artifact rather than a continuous loop. That world is gone. Pressure arrives faster now, not because physics changed, but because distance collapsed. Capital moves quickly, customers change quickly, activists strike quickly, and the tolerance for ambiguity inside and outside the firm has thinned. Boards respond to that environment with the tool they can most easily justify. They change the leader. The leader change is legible. It signals action. It satisfies the need to show control. It also avoids the harder admission, which is that the firm’s measurement and reward structure may be selecting for the wrong traits, the wrong skills, and the wrong time horizon. When a company rotates leaders without changing the mold, it is not solving a problem. It is feeding the mold fresh material. A blunt way to say it is this. The carousel is not a leadership problem. It is a performance definition problem.

Your Incentives Are Paying For the Carousel

Every board says it wants long-term value. Every compensation plan says it rewards results. Every performance management process says it differentiates the best from the rest. Then you look at the lived behavior inside most enterprises and you see the same pattern. People spend

disproportionate time on what is easily counted, easily defended, and easily credited to one person’s effort. They spend far less time on what is harder to count, harder to defend, and harder to attribute. The first set of activities wins bonuses and promotions. The second set determines whether the company can still win three years from now. This is not a moral failure. It is an incentives theorem. Organizations seek information about what is rewarded, then do those things, often to the exclusion of what is not rewarded. It is also common for reward systems to “pay off” one behavior while the rewarder hopes for another. That is not a clever management quote. It is the operating manual of every organization with a stalled performance curve and a busy succession committee. Economists gave the same phenomenon a more technical spine. When a job includes multiple tasks, and some tasks are easier to measure than others, strong incentives tied to the measurable task pull effort away from the hard-to-measure task. The distortion is not hypothetical. It exists because measurement is not neutral. Measurement is a magnet. It pulls effort, attention, and identity toward what it can see. Replace any title with “plant manager,” “division president,” “head of sales,” or “product leader,” and you get the modern enterprise in one paragraph. If you pay heavily for quarterly output, people will find ways to produce quarterly output. If you pay heavily for margin this year, people will find ways to create margin this year. If you reward “hitting the number,” people will hit the number, including by pulling tomorrow into today, including by starving maintenance, including by cutting training, including by delaying quality fixes, including by avoiding hard customer conversations, including by under-investing in the capabilities that make next year’s number less fragile. Then the board looks up a year later and asks why resilience is down, why innovation is thin, why quality is noisy, why safety is jittery, why turnover is rising, why the next layer is not ready, why the organization feels tired. The board says it wants “better execution.” The truth is that execution was never the constraint. The constraint was the set of behaviors the system made rational. This is where many executives reach for the comforting story that the wrong individual was in the role. That story protects everyone from the uglier alternative. The uglier alternative is that the organization promoted and rewarded the very behaviors that produced the later stall. When that is true, replacing leaders becomes a way to avoid replacing the definition of performance. You can often spot this in the language leaders use when they want to feel responsible without changing anything. “We need more rigor.” “We need more accountability.” “We need more discipline.” Those are not wrong words. They are incomplete words. Rigor applied to a flawed target does not create excellence. It creates precision failure.

Promotion is Treating Past Output as Proof of Future Leadership

The carousel accelerates because most firms treat promotion as a reward instead of a job change. They treat it as the natural next step for someone who performed well in the last role. Then they

act surprised when the person struggles in the new role. The error is not that people fail. The error is that the firm keeps using the wrong evidence. If you want a clean empirical example, look at what happens when top individual contributors become managers. Detailed evidence in sales organizations shows a negative relationship between being a strong salesperson before promotion and being an effective manager after promotion. High salespeople are more likely to be promoted, but those promoted from the top of the sales distribution can be worse managers as measured by their impact on their team’s sales. The misallocation costs are not small. They are material. That is the Peter Principle with receipts, except the point is not that people rise to incompetence. The point is that organizations confuse one kind of competence for another because it is easier to measure. In most corporate settings, the promotable evidence is loud evidence. It is evidence that can be shown in a dashboard, attributed to an individual, and defended in a room where nobody has time to investigate causality. That evidence privileges short-cycle, individually visible wins. Leadership, in contrast, is largely the ability to build conditions under which other people can win repeatedly, under varying conditions, with fewer heroics. That skill is quieter, slower, and harder to credit. It is often invisible until it is missing. So the firm promotes the person who can win the old game. Then it asks them to win a different game. The person does what any rational person does. They keep playing the game they know. They keep leaning on the behaviors that got them promoted. They keep managing up to the measures that shaped their identity and paycheck. Over time, the organization becomes a museum of misapplied excellence. This is why the carousel is often filled with genuinely capable people. The board is not selecting idiots. The enterprise is selecting people who are highly adapted to the measurement environment. It is the environment that is mismatched to the job the enterprise actually needs done. Ask yourself a question that most boards avoid because it sounds like philosophy until it becomes bankruptcy. Are we promoting people because they create durable capability, or because they create photogenic outcomes. Are we rewarding the removal of risk, or the concealment of risk. Are we paying for decisions that expand future options, or for decisions that harvest options early and then call the harvest “performance.” If those questions feel hard to answer, that is not because the answers are unknowable. It is because your current system was not designed to answer them.

You Are Measuring What You Can Defend, Not What Matters

The modern organization is full of metrics and starving for meaning. Leaders drown in dashboards and still end the quarter with the same argument. Nobody trusts the numbers.

Everybody trusts the politics. The performance process becomes a trial, not a learning loop. That is how the mold hardens. The most dangerous feature of many performance systems is not that they measure the wrong things. It is that they teach the organization to treat measurable proxies as reality. Once that happens, the organization begins to confuse the artifact with the outcome. Managers begin to manage the artifact. Employees begin to perform for the artifact. Executives begin to defend the artifact. Boards begin to govern the artifact. The actual enterprise becomes the thing the artifact describes, which means the enterprise starts living inside a model that was never built to carry the full truth. This is the same failure mode across industries. A company claims it rewards customer obsession. Then it pays for cost cutting that degrades service. A company claims it rewards quality. Then it pays for throughput that drives defect escape. A company claims it rewards innovation. Then it punishes failed experiments and pays for predictable releases. A company claims it rewards collaboration. Then it ranks people against one another and pays for individual heroics, contribution theater, and political dominance. These are not contradictions. They are consequences. Deming’s point was unforgiving. The system produces the results. People work inside that system. When measures, incentives, and management rhythms are misaligned, the organization does not get better behavior. It gets fear, gaming, and internal competition. It gets exactly what it pays for, then calls the outcome a culture problem. Companies that miss this keep trying to fix people. The leak is in the system. If the system does not change, the outcome will not. Those companies are destined to bleed trust, talent, and customers while insisting they are rewarding the right things. The academic literature on incentives does not say “pay for performance is bad.” It says something more precise and more useful. If performance is multidimensional, and measurement is noisy or incomplete, then high-powered incentives tied to a narrow measure can make the organization worse by pulling effort away from important dimensions. Simple contracts fail in complex settings for a reason. The organization is not a single-variable machine. Boards feel this in their bones, even if they do not name it. They see a leader who “hit the number” while the business became more brittle. They see a leader who “improved efficiency” while culture degraded. They see a leader who “delivered a transformation” while the underlying operating system accumulated debt. Then they wonder why the company is back in the same room, with a new deck, having the same fight. Here is the part that separates serious governance from ritual. If the same problems recur, it is rarely a people problem. It is almost always a system problem. People can cause acute failures. Systems cause persistent failures. So the right question is not “Who failed.” The right question is “What did we reward that made this failure rational.”

Your Talent Data Is Lying to You About Causality

Now we hit the most uncomfortable truth in corporate life, which is that most firms treat their HR data as if it were a causal map when it is often a record of selection. That distinction sounds academic until it explains why you keep promoting the wrong people. This is where a seemingly distant piece of research becomes directly relevant. Recent work on elite human performance has argued that young exceptional performers and later adult world class performers can look like two discrete populations over time. It also argues that early exceptional performance is associated with extensive discipline-specific practice and fast early progress, while adult world-class performance is associated with more multidisciplinary practice and more gradual early progress. Many leaders read claims like that and immediately import them into corporate life. They tell themselves the early stars are a different species. They tell themselves slow starters can become the true greats. They tell themselves specialization is dangerous. They tell themselves breadth is the secret. Then an important critique appeared. It argued that a negative association between early and adult performance in elite samples can arise naturally from collider selection bias when selection into the elite sample depends on both early and adult performance. The critique’s point was not that development does not matter. The point was that associations estimated within selected elite samples can be descriptively accurate for that selected group while still being causally misleading if the selection process is not modeled. If you think that is a sports science debate, you are missing the corporate punchline. Your internal promotion and performance datasets are also elite samples. They are selected by your own filters, your own definitions, and your own prior incentives. You do not have a neutral view of talent. You have a view of talent conditioned on who survived your measurements, who adapted to your culture, who learned the political game, who stayed long enough, who had the right sponsor, who did not take the risky role, who did not have the unlucky quarter, who did not dissent at the wrong time. That means your correlations can be true and still mislead you. You can “prove” that certain traits predict leadership success, when what you really proved is that those traits predict survival in your environment. You can “prove” that your highest performers should be promoted, when what you really proved is that your current performance measures select for the behaviors your current performance measures reward. You can “prove” that the new leader failed, when what you really proved is that the system creates brittle outcomes and then scapegoats individuals when the brittleness shows. This is why the carousel is so durable. The data appears to justify it. The data was produced by the same mold. A board that wants to end the carousel has to treat its own internal people analytics with the same skepticism it would apply to a vendor pitch. What is the selection mechanism. What is being filtered out. What traits win inside our system but fail in the real job. What would we see if we measured the things we currently ignore.

Those are governance questions, not HR questions. They belong in the boardroom, not buried in a talent review.

Boards Keep Changing Captains While the Map Stays Wrong

The pressure environment makes this worse. When investors demand action, boards reach for actions that are legible and defensible. Leadership change is the cleanest lever because it signals accountability without requiring the board to admit the system is mis-specified. Shareholder activism makes that lever even more tempting. Activism is often framed as a force for discipline. Sometimes it is. Sometimes it is also a force for impatience that pushes boards toward the short, visible move rather than the deeper, system-level correction. CEO succession data reflects the same reality. Large-cap succession research in the postpandemic period has documented rising CEO turnover even among stronger-performing companies, with turnover among top performers increasing and narrowing the historical gap between top and bottom performers. That matters because it suggests a world in which even delivering acceptable results may not buy a leader time. When time shrinks, the temptation to optimize for the metric that saves you this quarter grows stronger. The mold gets harder. The carousel spins faster. Here is the bitter irony. Boards often replace leaders because performance stalled. But the stall is often a lagging indicator of earlier decisions made under earlier incentives. By the time you see it in the P&L, the causal work is already done. The leader you replace may have inherited structural debt. The leader you hire may be forced to harvest quick wins to satisfy the same impatient environment that triggered the previous replacement. The board then concludes, again, that the leader was the issue. The system survives another year. This is how you create an organization that looks perpetually busy and feels perpetually stuck. It has constant motion and low progress. It has constant messaging and declining trust. It has constant restructuring and increasing fragility. The board believes it is demanding change. The organization experiences it as churn. The core governance mistake is treating recurring outcomes as proof of recurring leadership failure. Recurrence is proof of architecture. So the board needs a different set of questions, and it needs to ask them before it reaches for the human sacrifice. When the same operational failure returns, is it returning because the responsible leader lacked grit, or because the system made the failure the rational byproduct of incentives. When the same customer complaint returns, is it returning because the frontline does not care, or because the company’s decision path makes care expensive. When the same safety event returns, is it returning because people are sloppy, or because the organization keeps paying for output while starving the conditions that produce safe output. If those questions sound harsh, remember the alternative. The alternative is another leader swap, another reorg, another “new operating model,” another year of value drained into churn, and another board meeting at 6:42 p.m. where no one looks at the clock.

A Different System, or the Ride Never Ends

Ending the carousel does not require magic. It requires a willingness to redefine performance in a way that matches how performance actually works. That means treating outcomes as the downstream result of decisions, capabilities, and time, not as isolated quarterly events that can be attributed to a single leader. Start with a principle that boards rarely write down because it forces accountability upward. If you reward short-cycle outputs and punish long-cycle investments, you will get short-cycle outputs and long-cycle decay. That is not cynicism. It is an equilibrium. Then adopt a second principle that feels obvious once you say it. Any meaningful executive job is multidimensional. That means your reward and evaluation system must either measure the dimensions that matter or explicitly protect them from being sacrificed to the dimensions you can measure. When the work is multidimensional, the enterprise must prevent strong incentives tied to narrow measures from destroying the hard-to-measure work. Boards can do this through governance, not only through HR policy. If you cannot yet measure the quality of decisions, then do not build a comp plan that pays only for the number. If you cannot yet measure risk removal, then do not punish leaders for surfacing risk early. If you cannot yet measure capability growth, then do not treat training cuts as “discipline” without charging the leader for the future cost. This is where many executives protest that what matters cannot be measured. That protest is half true and fully weaponized. Many things can be measured poorly, which is why measurement must be treated as an engineering discipline, not as a compliance exercise. But even when measurement is imperfect, governance can still protect what matters by changing what gets punished and what gets promoted. Consider promotion itself. The evidence on salespeople and management is a warning label. Past individual output can be a poor predictor of managerial impact, and promoting based on the wrong signal can impose large costs. Boards should assume this problem exists in every function where the job changes from doing to enabling. The fix is not to stop promoting top performers. The fix is to stop pretending that past output alone is proof of future leadership. A serious organization treats promotion like an acquisition. It performs diligence on the new job, not only the old job. It defines the mechanisms the new job must produce, not only the outcomes it must report. It tests for the ability to build repeatable performance through others. It builds apprenticeship into the role so the organization learns whether the leader can operate the new system before it places the full burden on them. It designs incentives for the role that fit the role’s multidimensional work, not incentives inherited from the prior job. Now apply the same logic to the top of the house. A board that wants to end the carousel needs to adopt a small set of enterprise-level measures that make system performance visible, not just quarterly results. The point is not to add another dashboard. The point is to measure what the current dashboards allow you to ignore.

One measure is how quickly the organization detects and corrects reality. Another is how often the same failure returns after you “fixed” it. Another is whether decision paths are shrinking or growing, which you can observe through cycle times, rework loops, and the number of approvals required for material actions. Another is whether the company is building future options or consuming them, which you can observe through capability growth, bench readiness, quality stability, customer retention, and safety performance that does not rely on heroics. You do not need perfect precision on these measures to get value from them. You need consistency and honesty. The moment you start measuring recurrence, you stop celebrating onetime wins that leave the underlying problem alive. The moment you start measuring decision latency, you stop congratulating governance that produces late decisions. The moment you start measuring option creation, you stop paying people to burn the future for a clean quarter. This is also where you learn whether you are serious about learning. If your internal data “shows” that early high performers become the best leaders, ask what selection created that conclusion. If your data “shows” that slow starters never become stars, ask what your system did to them. The debate around elite performance development is a reminder that selection can manufacture patterns that feel causal. Corporate life is full of the same trap. So here is the diagnostic test that will tell you, in plain terms, whether your enterprise is built to spin the carousel. When a leader misses a number, do you learn which upstream decision or constraint created that miss, or do you punish the person until they learn to manage the optics. When a leader hits a number, do you examine what they borrowed from the future, or do you celebrate the artifact and promote the behavior. When the same issue returns, do you treat recurrence as a governance failure, or do you treat it as proof you need a fresher face. If you want a second test, ask it in one breath and listen to how defensive the room gets. In our company, who gets promoted, the person who prevents problems or the person who resolves crises. In our company, who gets rewarded, the person who tells the truth early or the person who tells the best story late. In our company, who gets protected, the leader who invests in capabilities or the leader who cuts them to hit the quarter. If you cannot answer those questions without arguing, you already have your diagnosis. The mold is selecting for presentation, not performance. The carousel is not a mystery. It is a payout schedule. None of this means leadership does not matter. Sometimes a leader truly is the constraint. Sometimes ethics fail. Sometimes competence fails. Sometimes character fails. The point is that swapping leaders is not a strategy when the underlying system keeps manufacturing the same failure mode. It is a ritual that converts systemic causality into individual blame. The board that ends the carousel is the board that changes the mold. It changes how performance is defined, what is rewarded, what is punished, and what is treated as non-negotiable evidence of real capability. It stops treating quarterly outcomes as proof of leadership and starts treating decision quality, recurrence reduction, and capability growth as the real indicators of whether the enterprise is compounding or consuming itself.

Here is a prediction that will be easy to test and hard to forgive if it is wrong. Take any enterprise that has rotated two or more senior leaders in the same business within five years. Measure, with brutal honesty, the time it takes that business to detect a material issue, decide on a corrective action, and implement the correction. Then measure how often the same class of issue returns. If the carousel was driven by individual failure, those system measures should improve with each swap. If the carousel was driven by the mold, those measures will stay stubbornly flat, and the new leader will inherit the same physics with a new title. That is the choice in front of every board that keeps a polished succession plan but keeps living the same year. Fix the system that produces the carousel, or keep buying the illusion of progress one leader at a time. And pay for it forever. References This essay is grounded in works drawing on Science review on world-class performance development and its claims about early versus adult elite trajectories, published December 18, 2025. Public critique arguing collider selection bias can generate negative early versus adult associations within elite samples, published December 20, 2025. Holmström and Milgrom’s multitask principal-agent work on incentive distortion and task measurability, Journal of Law, Economics, and Organization, 1991. Steven Kerr on reward systems paying for one behavior while hoping for another, Academy of Management Journal, 1975. Empirical evidence that top sales performers promoted to management can perform worse as managers and that misallocation is costly, Quarterly Journal of Economics, 2019 and related working paper materials. CEO succession research indicating rising turnover even among stronger-performing companies and narrowing gaps between top and bottom performers, The Conference Board and partners, November 2025. Shareholder activism reporting describing record campaign volumes in 2025, including Reuters and Lazard’s annual review released in early January 2026.