Productivity Was the Wrong Prize

machine that can observe, but cannot intervene. “What would have to be true for this outcome to keep repeating.” The answer is not that people do not care. The answer is that the enterprise is not controllable. A company can be productive and still be uncontrollable. It can produce more per labor hour and still be unable to change the next result before the cost is locked. It can know earlier and act later. It can have visibility without authority. It can have data without decision rights. It can have AI that predicts a deviation while the operating system forces the corrective action to climb a permission staircase that was built for a slower age. In that world, productivity becomes a local improvement inside a global delay. The market does not respond to your internal efficiency. It responds to what you can reliably do on purpose. The phrase we keep returning to is simple and it does not flatter anyone. Control means the ability to intervene and make the next outcome different on purpose, without convening the entire company to grant permission. That sentence is not poetry. It is a definition of power in operations. It is also a test. If we cannot do that, we do not have control. We have theater. The causal network on the screen is a confession that the old prize is not enough. It is a picture of dependency. It is a reminder that every “initiative” lives inside a web of constraints, incentives, decision lags, data lags, and approval gates. It is a warning that if we optimize the parts while the control loop stays broken, the whole will keep reproducing the same failure modes through different surfaces.

The last era trained us to chase productivity

Most of us were taught to think in the grammar of productivity because it was the right grammar for a long time. The work was physical. The constraints were visible. The cycle time lived on the floor. Waste lived in motion and inventory. The idea of improvement could be made concrete because the system state could be touched. If you reduced changeover, throughput improved. If you stabilized a process, scrap dropped. If you redesigned material flow, working capital freed. The results were measurable, repeatable, and close to the point of action. Even the best versions of those traditions, the ones that actually worked, were never only about “doing more with less.” They were about control. They were about distinguishing noise from signal, then intervening where intervention mattered. They treated improvement as a discipline of cause and effect, not as a motivational slogan. That is why those traditions still matter. What changed is not that productivity became wrong. What changed is that the dominant cost moved upstream into information, permission, and coordination. The enterprise began paying its largest tax in decision latency, not in machine speed. The work of mod ern operations became the work of keeping outcomes inside bounds while demand, supply, regulation, labor, and technology all pressed the system at once. A plant can run a beautiful lean system and still be trapped by a corporate approval chain that decides slower than reality changes. A supply chain can forecast well and still be unable to adjust

terms or volumes without multiple committees. A quality organization can identify drift early and still lack the authority to halt, correct, and prevent recurrence without political escalation. A safety team can see hazards forming and still wait for a bud get cycle, a staffing action, or a policy rewrite. The waste is no longer only in motion. It is in the gap between knowing and doing. That gap is not an accident. It is designed. We built permission systems to prevent error and protect accountability. We built governance to prevent rogue action. We built functional silos to create expertise and scale. Then we layered digital systems on top, which increased the amount we can observe without changing the structure that determines what we can do. So the enterprise learned to predict its own outcomes with greater precision while losing the ability to alter them in time. This is why “productivity” has become an easy place to hide. It is a score that can improve while controllability worsens. When controllability worsens, the enterprise becomes less able to correct drift. Drift becomes chronic. Chronic drift becomes surprise. Surprise becomes cost. Cost becomes a story told after the fact.

Visibility is not control, and prediction is not intervention

The slide on the screen is not a dashboard, but it exposes the same illusion. It suggests a question that should embarrass any serious operator. If we can see a deviation earlier than ever, why do we still live with the consequences as if we saw it late? Control has a strict meaning in the disciplines that actually care about it. A system is observable when you can infer what is happening. A system is controllable when the inputs you possess can move the state you care about. In an enterprise, the “inputs” are not only money and labor. They are decision rights, authority, information flow, escalation paths, policy constraints, and the ability to execute corrective action at the edge where the deviation starts. If those inputs cannot move the outcome before the cost is locked, the system is not controllable. It can be watched. It cannot be steered. Most modern operating models are built to maximize observability. They do this through reporting layers, analytics layers, data platforms, dashboards, and review meetings. They generate a high volume of explanation after the fact. They do not automatically generate authority before the fact. This is where AI often becomes a magnifier of disappointment. AI can increase observability fast. It can detect patterns. It can rank risks. It can predict likely outcomes. But AI does not grant permission. AI does not rewrite decision rights. AI does not compress the approval chain. If the enterprise routes action through the same permission staircase, AI becomes a better alarm inside a building with locked exits. The causal network makes this visible because it is not trying to score performance. It is trying to trace dependency. It puts causes next to each other in a way that makes excuses harder. It says, in

effect, that the outcomes you claim to want are downstream of the way you route decisions. The network looks like complexity, but the point is simpler. If your interventions cannot reach the state variables that produce the outcome, your improvement work will turn into talk. Control means the ability to intervene and make the next outcome different on purpose, without convening the entire company to grant permission. When we repeat that sentence, it tightens. Control means that the distance between signal and authorized action is shorter than the time it takes for the deviation to become irreversible. Control means that authority lives where the evidence first appears, not where the politics feel safest. When an enterprise lacks that, it does something predictable. It turns observation into ceremony. It turns ceremony into comfort. It turns comfort into delay. It calls the delay “prudence.” Then it pays the bill in scrap, expediting, customer churn, overtime, attrition, and incidents that were visible early but treated late.

Causality is the bridge between productivity and controllability

The reason your image matters is not that it is a map. It is that it forces the right kind of thinking. A causal network is not primarily a description of what happened. It is a structure for asking what interventions are possible and what interventions matter. There is a hard boundary between correlating and controlling. Correlation can rank what tends to occur together. Control requires a claim about what happens when we intervene. That is why causal reasoning belongs at the center of the COO agenda now. Not as academic interest. As the only way to decide which parts of the operating system should be automated, which parts should be delegated, and which parts should be designed out. A productivity program often starts with measurement and ends with targets. A controllability program starts with the interventions you can take and ends with the outcomes you can change. It asks questions that are not comfortable because they expose the operating design. Can we alter the next outcome without escalation. Can we correct drift before it becomes cost. Can we take an action that changes what would have happened otherwise. Can we do it repeatedly, in a way that survives turnover, politics, and the next reorg. This is where the evolution from productivity to controllability becomes clear. Productivity focuses on the efficiency of effort. Controllability focuses on the ability to cause a different result. The market rewards the second, even when the first is admirable. The causal network on the screen also reveals a second evolution. Once you have controllability, you stop merely responding to a market. You begin shaping it. Not through slogans. Through behavior that other actors must adapt to. A market is not an abstract thing. It is a set of expectations and constraints held by customers, suppliers, regulators, competitors, and labor. Those expectations are formed by repeated experience. If your firm can correct drift faster than others, your lead times become more

reliable. Your quality becomes less volatile. Your service becomes less dependent on heroes. Your cost becomes less sensitive to shocks. Customers change how they plan around you. Suppliers change the terms they offer you. Competitors change the bets they make because the old vulnerabilities they exploited no longer work. That is what it means to shape the market. You are not persuading the market with words. You are forcing it to respond to your control. Control means the ability to intervene and make the next outcome different on purpose, without convening the entire company to grant permission. When we tighten it again, it becomes sharper. Control is the ability to force the outside world to react to your reliability, not to your excuses.

Permission is the hidden mechanism, and it is measurable

Executives often talk about culture because culture is a safe word. It points to behavior without naming architecture. It allows leaders to blame people when the system design is at fault. The causal network is more honest. It does not let you hide inside sentiment. It points to structure. Permission is structure. Permission is the way the enterprise decides who is allowed to act, when, with what evidence, under what policy, under whose signature. Permission is not only a compliance matter. It is a latency engine. It determines how long it takes to convert signal into action. Latency is not only a time problem. It is a finance problem. It determines how many opportunities expire before the enterprise can act. It determines how many losses become irreversible. It determines how often the enterprise must spend money to compensate for delay, through expediting, overtime, premium freight, rework, warranty, and settlement. It determines how much option value is destroyed by waiting. This is the bridge between controllability and market shaping. If you cannot act, you cannot shape. If you can act, you begin changing the landscape others must deal with. The market responds to whoever can change outcomes reliably. The reason this feels harder than productivity is that it runs into politics. A productivity project can often be kept local. A controllability project challenges who has authority. It challenges who gets to say yes. It challenges which committees matter. It challenges which functions serve as control towers and which functions serve as toll booths. So here is the question that should be asked without ceremony. Where does the enterprise force humans to translate, reconcile, and ask permission, even when the evidence is already sufficient? Where does it require escalation not because the risk is high, but because the organization is afraid of accountability? Where does it prefer procedure over intervention because procedure spreads blame? If you want a causal explanation for chronic drift, it is often here. The enterprise does not lack data. It lacks the ability to act on data without social negotiation.

We can pressure test this without moralizing. Measure time from first detectable deviation to first authorized corrective action. Measure how many sign offs it takes to stop the line, quarantine a lot, change a supplier term, adjust a schedule, or replace a standard. Measure how often the corrective action is delayed until the next meeting where the right people are present. Measure how often the root cause is known but the fix is deferred because it requires cross functional approval. If those times and counts are high, controllability is low, even if productivity is high. The enterprise is learning about its own problems faster than it can correct them. That gap is where value goes to die.

The fair counterexample, and why it proves the point

There is a legitimate objection that must be faced, not waved away. In some environments, speed is the enemy. In high hazard operations, aviation, nuclear, chemical processing, emergency medicine, the wrong intervention can kill people. Deliberation can be a control function. Authority without discipline can be catastrophic. So it is fair to say that removing permission gates is not automatically good. It is also fair to say that some decisions must be slow because the cost of error is extreme. But that counterexample does not defeat the argument. It clarifies it. Those industries do not succeed by being slow. They succeed by being controllable. They pre define actions, triggers, and authority so that the right intervention can be executed without debate when the signal appears. They invest in training so that the person closest to the hazard can act within clear bounds. They design escalation paths that are fast when time matters and deliberate when time allows. They build systems where control is not improvisation. It is design. That is the point. The goal is not speed for its own sake. The goal is the ability to make the next outcome different on purpose, safely, repeatedly, and in time. When a firm uses safety as a reason to keep every decision centralized, it often creates a worse risk. It creates a system that cannot act when it must. It creates drift that goes uncorrected until it becomes crisis. Then the enterprise acts in panic, which is the least safe state of all.

The questions that expose whether you can shape anything

If we stripped the adjectives away, what would we actually ask in the executive meeting when this causal network is on the screen? We would ask questions that can be answered in the record. When a deviation appears, who can take the first corrective action without permission. What evidence is considered sufficient to act. How many approval steps exist between the person who sees the signal and the person who can authorize the fix. How often d o we postpone action until a weekly review because that is the only place authority convenes. How many outcomes do we accept as “unavoidable” that are actually outcomes we were not allowed to intervene on?

Those questions are not about culture. They are about controllability. Now we ask the market facing version, because this is where the argument stops being internal. What promises do customers believe when they plan around us. Do they treat our lead time as a commitment or a guess. Do suppliers treat our forecasts as signals worth honoring or as noise. Do regulators see us as a firm that prevents recurrence or a firm that explains it well. Do competitors fear our ability to correct drift faster than they can exploit it, or do they assume we will do what we did last time? If you cannot answer those questions with evidence, you are not shaping the market. The market is shaping you. Here is the testable prediction, stated as a risk, not as a boast. If you measure your own operating system, you will find that your biggest recurring costs live in the space between detection and authorized intervention, not in the space between authorized intervention and execution. If that is wrong, we should be able to prove it in your records. If it is right, then every productivity project that does not reduce that delay will produce diminishing returns, even if local metrics look better. That prediction should be embarrassing if wrong because it is not philosophical. It is a measurement claim.

What the causal network is really telling the COO

We can now say what this image is doing without romanticizing it. It is not a diagram to explain the world. It is a device to force discipline about cause and effect. It asks the COO to decide what is controllable, what is not, and what must be redesigned so that it becomes controllable. The evolution from productivity to controllability is the evolution from managing effort to managing outcomes. The evolution from controllability to market shaping is the evolution from reacting to outside pressure to creating outside pressure for others. That is how advantage is created when products are similar and technology is available to everyone. The board does not ultimately care how many initiatives are in motion. It cares whether the enterprise can prevent harm, reduce waste, hold quality, meet carbon obligations, and protect margin without needing heroics and escalation each time. Those are control outcomes. They are not slogans. They are the practical definition of whether the enterprise can be trusted with its own complexity. Control means the ability to intervene and make the next outcome different on purpose, without convening the entire company to grant permission. Tighten it one last time and it becomes a verdict. If you cannot do that, you are not running the system. You are watching it. The market has a simple habit. It rewards whoever can act in time. What would have to be true for your enterprise to keep repeating the same failures with better explanations.

References This narrative rests on a control tradition that predates modern digital management and clarifies the difference between observing and steering. Shewhart’s Economic Control of Quality of Manufactured Product in 1931 framed quality as a problem of statistical control rather than inspection after the fact. Deming’s Out of the Crisis in 1986 carried that same discipline into executive responsibility for systems, variation, and repeatability. Wiener’s Cybernetics, first published in 1948, established feedback as the core mechanism of control in complex systems, and Ashby’s An Introduction to Cybernetics in 1956 sharpened it into the law of requisite variety, which maps cleanly onto modern enterprises that face more disturbance than their decision systems can absorb. Kalman’s 1960 work on the general theory of control systems gives the formal definition of controllability that motivates the central claim, and Galbraith’s information processing view of organization design in the 1970s explains why coordination cost and decision rights, not only labor efficiency, determine performance under uncertainty. Coase’s The Nature of the Firm in 1937 anchors the argument that internal governance has a cost and that cost shows up as delay. Pearl and Mackenzie’s The Book of Why in 2018 explains the boundary between correlation and intervention that separates dashboards from control. Dixit and Pindyck’s real options work in the 1990s, along with later syntheses, grounds the claim that delay destroys option value when action is irreversible. Weick and Sutcliffe’s Managing the Unexpected in 2007 provides the fair counterexample and the corrective point that high hazard work wins through designed authority and feedback, not through centralization as a reflex.