The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

When Productivity Decline Starts Funding Itself How recovery work consumes the capability, controls, and learning loops manufacturers need to reverse decline. By Michael Carroll Research Fellow, LNS Research | Founder | Investor | Board Advisor | Industrial AI, Causal Systems, and Enterprise Transformation

The enterprise can see the decline, but the capacity required to reverse it has already been pulled into the recovery.

The workbook is open on one screen and the causal map is open on another. The workbook looks ordinary at first. Rows, questions, gaps, percentages, leader and follower cuts. Forty companies classified as leaders. One hundred thirty-four classified as followers. A second cut, more revealing, separates companies that say they are consistently ahead from those that say they are competitive but not leading. The map looks nothing like a workbook. It is a field of causes, controls, operational challenges, mitigants, consequences, and feedback lines, all converging on the decline of manufacturing productivity. That is the moment when the question changes. The issue is no longer whether productivity declined. The government data already gives that concern weight. The issue is whether we have been looking at the wrong layer of the problem. Outsourcing, cost pressure, operational complexity, decision latency, workforce capability, capital gates, and fragmented learning do not sit beside one another like separate items in a survey. They can form a cascade. Once the cascade begins, the enterprise can spend the very capacity required to improve productivity on the work required to survive without it.

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

The spreadsheet gives the symptoms. The map gives the system. The hard part is admitting that the recovery work may have become part of the decline.

The decline did not arrive as collapse

There is a reason productivity decline is hard to confront inside serious companies. It rarely begins as failure. It begins as competence under pressure. A plant manager protects the customer with overtime. A supply-chain leader covers a service risk with premium freight. A finance team tightens capital gates after margin misses. A procurement team moves work to a lower-cost supplier because the internal cost structure no longer clears the hurdle. An engineering group postpones capability work because the line needs recovery now. None of those decisions requires stupidity. Most of them require responsibility. That is what makes the pattern dangerous. The organization can be filled with capable people making defensible decisions, and still produce a worse system. The action that protects the quarter can weaken the year. The action that lowers visible cost can raise hidden coordination burden. The action that moves work outside the company can move knowledge with it. The action that adds governance can slow correction. The action that explains variance can consume the same people who should have removed its cause. Manufacturing productivity has been treated for too long as if it were mainly an efficiency story. Better equipment. Better labor utilization. Better maintenance. Better scheduling. Better automation. Those matter, and no serious operator should pretend otherwise. But the more useful question is whether those improvements sit inside a system that allows them to compound. If capability does not become control, and control does not improve operating conditions, the gain remains exposed. It can show up in a metric, disappear in the next recovery cycle, and then return as another program. What would have to be true for this outcome to keep repeating. The answer is not that people forgot how to improve. The answer is that consequence-management can capture the system. Productivity decline creates visible damage. The organization responds. The response consumes attention, capital, engineering time, leadership bandwidth, frontline trust, and learning capacity. Those are the same scarce resources needed to build the controls that create productivity improvement. The enterprise does not merely lose productivity. It begins using the machinery of improvement to pay the bill for not improving. That is the capability inversion cascade.

The enterprise spent the capacity needed to recover on the work of recovery.

The old cascade created improvement In a healthy productivity system, capability comes first. Capability is not a slogan about talent. It is the accumulated ability of people, processes, equipment, knowledge, data, and governance to produce the desired outcome repeatedly. It includes operator judgment, engineering depth, maintenance discipline, supplier understanding, planning quality, process knowledge, decision rights, and capital allocation maturity. It is the part of the enterprise that knows how the work really behaves when the plan meets the floor. Capability enables controls. That word is often damaged by bad management practice, but the right controls are not bureaucracy. They are how the enterprise makes good work repeatable. Standard work is a control. Preventive maintenance is a control. Quality at the source is a control. Constraint visibility is a control. Decision rights are a control. Evidence thresholds are controls. Escalation rules, when designed well, are controls. The purpose of control is not to slow action. The purpose is to make the right action legitimate before the cost grows. Controls shape operating conditions. They reduce variation. They make flow more stable. They improve quality. They clarify which tradeoffs have already been decided and which need judgment. They reduce unnecessary handoffs. They Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

move correction closer to consequence. They let the plant, the supply chain, the commercial team, and finance operate from a more common view of reality. Better conditions do not guarantee productivity improvement, but poor conditions almost guarantee that improvement will not persist. That is the normal cascade. Capability enables controls. Controls shape operating conditions. Better operating conditions produce productivity improvement. Productivity improvement creates more capacity to strengthen capability. The system compounds because the work that improves the system is protected long enough to matter. The workbook makes this visible in the strongest companies. The larger separation is not found in whether companies have metrics. Many do. The separation appears where operations discipline is tied to financial outcomes at the board level, where knowledge technology makes tacit knowledge usable, where operating models are digitally reinforced, where decision quality and latency are measured, and where every decision can become part of a learning record. That is not more measurement. That is a different relationship between evidence and action. The companies that pull away are not merely better at doing work. They are better at preserving the conditions under which work teaches the enterprise.

Figure 1. The capability inversion cascade shows why consequence-management can become a cause of further productivity decline.

The reversal begins as mitigation

Once productivity declines, the operating logic changes. The enterprise still says it is pursuing improvement, but the daily allocation of capacity tells a different story. The best people are no longer improving the system. They are protecting it from the consequences of its own weakness. The maintenance leader is in recovery mode. The engineering leader is chasing repeat problems. The plant manager is explaining misses. The supply-chain leader is expediting. The commercial leader is negotiating promises around uncertainty. Finance is demanding proof before funding the very work that would reduce the uncertainty. This is not a character failure. It is a capacity failure. The organization has more consequence than it has correction capacity. When that happens, management does what management has to do. It triages. It patches. It prioritizes. It protects the customer. It reduces visible cost. It moves the issue to whoever can absorb it fastest. The problem is that absorbing a consequence is not the same as changing the cause. Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

A premium freight line can protect the customer while hiding a planning failure. Overtime can protect output while hiding instability. Added inspection can protect the shipment while hiding a process problem. A supplier change can protect cost while hiding the loss of internal process knowledge. A dashboard can improve visibility while hiding the fact that nobody has authority to act on what is visible. Each mitigant can be rational. Together, they can become the operating model. That is where the cascade turns. The consequence creates mitigation. Mitigation consumes capability. Weakened capability weakens controls. Weakened controls degrade operating conditions. Worse operating conditions create more productivity decline. The next cycle arrives with less capacity than the last one. The organization calls it pressure. The system calls it feedback. The most dangerous part is that the enterprise often feels busier and more governed while becoming less capable. There are more meetings, more reviews, more reconciliations, more supplier calls, more reports, more escalations, and more recovery plans. Activity rises. Control does not. The system becomes more crowded because the underlying productive architecture has lost strength. A CFO can see the cost, but not always the cause. The line item appears as premium freight, overtime, inventory, scrap, outside services, consulting spend, warranty, customer credits, or capital deferral. The ledger does not automatically show that the same engineer pulled into recovery would have been the person who removed the constraint. The P&L records the consequence. It does not record the work that did not happen because consequence captured the calendar.

The cost of decline is not only in the ledger. It is in the capability work that never gets done.

A cheaper input can create a more expensive enterprise

This is where outsourcing belongs in the argument, but it has to be handled with care. A weak version of the argument says outsourcing caused manufacturing productivity decline. That is too blunt. It is also unfair to the evidence. Outsourcing can lower input costs. Imported intermediates can support measured productivity. A supplier may genuinely perform some work better, faster, or at lower cost. A company that treats every make-vs-buy decision as a moral referendum will make bad decisions of its own.

Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

The stronger argument is not moral. It is causal. Outsourcing becomes a productivity problem when it moves the learning loop outside the enterprise. Make-vs-buy has usually been governed as a cost, capacity, and focus question. Can someone else make this cheaper? Can they provide flexibility we cannot? Should we reserve scarce capital for higher-return work? Are we keeping the right activities inside and moving the rest outside? Those are legitimate questions. In a slower environment, they may have been sufficient. In a faster one, they are not enough. The missing question is harder. If someone else performs this work, who owns the learning required to shape the next outcome? Who sees the process constraint first? Who understands the material behavior after the fourth exception, not the first? Who owns the operator judgment? Who knows which defect code is really a design issue, which supplier issue is really a specification issue, and which apparent labor issue is really a decision-rights issue? Who can correct before the customer teaches the lesson at full price? If the outsourced activity contains no strategic learning, the decision may be straightforward. If it contains process memory, constraint knowledge, quality judgment, manufacturing IP, rapid correction capability, or future productprocess optionality, then the decision is no longer procurement. It is architecture.

Outsourcing becomes dangerous when it moves the learning loop with the work

This is why government measurement and enterprise measurement can tell different truths at the same time. Official accounts can recognize the efficiency associated with lower-cost imported inputs. Productivity measures can capture changes in output relative to inputs. Research can show that imported intermediate inputs contributed to manufacturing labor productivity over a period. All of that can be true. It can also be true that a firm weakened its own productive system by moving knowledge, feedback, and correction rights outside its walls. That is not contradiction. It is level of analysis. A national account can measure one kind of efficiency while the enterprise loses one kind of agency. The company sees lower cost. The operating system inherits longer correction paths. The board sees margin relief. The plant sees more dependency. Procurement sees savings. Engineering loses the failure record. Finance sees variable cost. Operations loses the right to learn from the work directly. A cheaper input can still produce a more expensive operating system. The workbook's market-shaping lens fits this point directly. Leaders were more likely to treat internal manufacturing as competitive strategy, protect manufacturing process IP, and invest in internal capability separately from maintenance capital. The consistently ahead cut was more revealing than the formal leader-follower cut because it exposed behaviors closer to the operating mechanism. Ahead companies were more likely to connect operations to margin, encode standard work and problem solving digitally, use AI in production decisions, and track decision quality and latency. The competitive but not leading cohort showed the failure pattern more clearly. Optionality was endorsed in concept, but every request still had to pass quantified ROI gates. Internal capability was eroding through cost-pressure-driven outsourcing. Tools existed. Programs existed. The loops did not close. That is the point. The weak enterprise does not necessarily lack tools. It lacks connection. Data does not become evidence. Evidence does not become decision. Decision does not become learning. Learning does not become control. Control does not become better operating conditions. And because the cascade does not complete, improvement remains episodic. The board thinks it approved discipline. The system experiences amputation.

Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

Supporting image. Make-vs-buy becomes dangerous when the enterprise prices the work but not the knowledge inside the work.

The counterargument must be taken seriously

The strongest counterargument is that this argument risks romanticizing internal manufacturing. Not every capability deserves to be owned. Not every supplier relationship weakens learning. Not every outsourced process contains strategic knowledge. Some suppliers invest more, learn faster, innovate better, and produce more consistently than the manufacturer could on its own. Outsourcing can let a company shed nonessential burden, gain access to specialized capability, and focus internal talent on what truly differentiates the firm. That counterargument is right. It should make the thesis more precise, not weaker. The issue is not whether work sits inside or outside the company. The issue is whether the enterprise retains enough control over learning, evidence, correction, and option value to shape future outcomes. A supplier can be part of a strong learning system if the relationship is designed that way. Shared data, joint problem solving, protected process IP, clear escalation thresholds, rapid feedback, evidence trails, and aligned investment logic can make external work part of a coherent operating model. The boundary of the firm is not the only boundary that matters. The boundary of learning matters more. But that is exactly why treating outsourcing as cost reduction is insufficient. The firm must know what it is moving. It must distinguish commodity work from learning-rich work. It must know when supplier dependence increases correction time. It must know when internal capability loss creates future capital cost. It must know when the apparent savings are funded by the decay of a capability that will be expensive to rebuild. The observed fact is that manufacturing productivity growth has weakened materially over the last long cycle. The inference is that fragmentation, decision latency, and capability erosion help explain why improvement has not compounded in many firms. The projection is that companies unable to govern make-vs-buy as a learning-system decision will keep funding recovery while competitors convert capability into market control. That projection can be wrong. It would be weakened if firms with heavy cost-pressure outsourcing also show faster correction cycles, stronger internal process learning, better decision latency, higher quality stability, and stronger boardvisible productivity-to-margin conversion over the next three years. That is a fair test. If outsourcing preserves or Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

improves the learning loop, the causal claim loses force. If it lengthens correction and erodes capability, the claim gets stronger. A serious thesis should be willing to face that test.

Figure 2. Outsourcing is not inherently wrong, but the decision changes when the outsourced work carries the learning loop.

Metrics do not rescue a broken cascade

The modern enterprise is not short on measures. It can measure uptime, yield, scrap, productivity, labor efficiency, cost of poor quality, on-time delivery, forecast error, inventory, margin, schedule adherence, asset performance, supplier performance, and customer service. Yet the existence of metrics does not mean the enterprise has a productive control system. A metric can describe a consequence without changing the cause. This is where many productivity programs stall. The company builds the dashboard, holds the review, assigns an owner, launches the program, and believes the mechanism is now in place. But the hard work is not the metric. The hard work is deciding what the metric is allowed to change. Can it change staffing? Can it change capital allocation? Can it change supplier strategy? Can it change decision rights? Can it change standard work? Can it change commercial promises? Can it change the make-vs-buy logic? Can it change who is allowed to act when evidence appears? If the answer is no, the metric is not a control. It is a report. That distinction matters because productivity improvement depends on the conversion of information into changed conditions. A quality metric that does not change process behavior is not productivity infrastructure. A maintenance metric that does not change asset strategy is not productivity infrastructure. A decision-latency metric that does not remove unnecessary approval gates is not productivity infrastructure. A capital metric that does not protect capability investment is not productivity infrastructure.

Metrics do not create productivity until they change how work is governed

Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

This is why the stronger companies in the workbook look different. They are not separated mainly by having more measures. They are separated by whether operations data feeds closed learning loops, whether decision quality and latency are treated as operating variables, whether capability gaps inform investment, and whether productivity is tracked to margin at the board level. That is the difference between describing productivity and governing productivity. A board should be able to ask a simple question: when this metric moves, what is allowed to move with it? If the answer is another review, the system is still consequence-level. If the answer is a change in decision rights, capital posture, operating standard, supplier integration, or learning record, the system is closer to causal control. The same question applies to outsourcing. When outsourcing reduces cost but increases correction time, what moves? When supplier dependency increases quality variance, what moves? When internal engineering loses process knowledge, what moves? When an outsourced component becomes strategic to the next product cycle, what moves? If the answer is only price negotiation, the enterprise is managing the commercial contract while the operating model loses control.

Figure 3. The productivity gap is not the presence of metrics. It is whether metrics are allowed to change the operating model.

The capability line belongs in capital governance

The capability inversion cascade cannot be fixed by operations alone. It reaches finance, because capability requires funding before its absence becomes obvious. This is where many companies trap themselves. They demand quantified return for work whose primary economic purpose is to preserve future control. They approve recovery spending because the consequence is visible and deny capability spending because the avoided consequence is not yet in the ledger. That is not financial discipline. It is timing error. The workbook's strongest signal may be the gap between companies that separate optimization from transformation and those that do not. When improvement, recovery, maintenance, and capability building all compete in the same capital logic, the work with the visible fire wins. The work that prevents the next fire waits. Over time, the firm funds consequence because consequence can defend itself. Capability cannot defend itself as easily, because its value often appears in the cost that never occurs. This is why capability capital needs different governance. Maintenance capital keeps the asset running. Growth capital expands the asset. Capability capital preserves or increases the enterprise's ability to learn, correct, and act under Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

changing conditions. It may not always show a clean one-year ROI. It may show up as reduced decision latency, fewer recurring defects, less premium freight, faster stabilization, lower customer concession cost, better schedule reliability, stronger process IP, and a shorter path from signal to correction. A CFO should not accept vague capability claims. But the answer to vague claims is better evidence, not a capital gate that starves the work by design. The company should ask what consequence the capability prevents, what control it strengthens, what decision it accelerates, what learning it captures, and what future option it preserves. That is not soft logic. It is economic logic with time put back into the model. Within three years, many manufacturers will have evidence of this distinction in their AI spending. Companies that use AI to generate more visibility without changing decision rights will add another layer of explanation to the same slow system. Companies that use AI to encode learning, support evidence thresholds, identify constraints, and shorten correction paths will separate. That prediction is concrete enough to be wrong. It will show up in whether AI reduces recurrence, decision latency, and consequence-management work, or simply creates better reports about all three. The factory will tell the truth before the board deck does.

Capital gates often approve the cost of recovery faster than the capability that would have prevented it.

The market-shaping question

A market-shaped company reacts to consequence. A market-shaping company protects the capability to intervene before consequence sets the terms. That is the connection between productivity, outsourcing, causality, and strategy. Market shaping is not a posture. It is not a slogan about being bold. It is the ability to create and stabilize outcomes competitors must respond to. That requires the company to learn faster than the market penalizes delay. It requires authority near consequence, coherence at the core, and evidence strong enough to make action legitimate before everyone is comfortable. It requires fewer degrees of separation between signal, judgment, permission, correction, and learning. Outsourcing can be compatible with that if the learning system remains intact. It is destructive when it turns the firm into a coordinator of other people's knowledge. A company can own the customer relationship and still lose control of the Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

productive intelligence that makes the promise credible. It can own the brand and lose the process. It can own the margin target and lose the correction path. It can own the board deck and lose the operating reality. The causal map shows why this matters. Triggers and controls do not flow into productivity decline once and stop. Operational challenges create consequences. Consequences require mitigants. Mitigants can feed back into the system. When they weaken capability, they become new causes. The enterprise is no longer managing a productivity problem. It is managing a productivity-producing system that has started producing decline. That is a different problem. The Monday morning question is not whether the company should insource everything, outsource nothing, buy more technology, or run another productivity program. The question is where the cascade has inverted. Where are the best people spending their time? Which recurring consequences are consuming the work that should remove their cause? Which outsourced processes contain learning the company still needs? Which capital gates protect cash while starving future control? Which metrics describe pain but cannot change the operating conditions that create it? A board could ask those questions without a new model, but most will not ask them until consequence has become visible enough to be politically safe. By then, the best answer is often more expensive than it needed to be. Capability has to be rebuilt. Supplier dependence has to be renegotiated. Process knowledge has to be recovered. Workarounds have to be unwound. Decision rights have to be rewritten. The company pays twice, first to lose the capability and then to reacquire what it should have protected.

A market cannot be shaped by a company that no longer owns its correction path

The real productivity question is not how many improvements a company can launch. It is whether improvement becomes a property of the operating system. That requires causality. Correlation can tell the company what strong performers tend to do. It can suggest hypotheses. It can show patterns. It cannot tell the enterprise which intervention will change its own system. For that, the company has to know what causes what, what happens if it intervenes, and what evidence will prove whether the intervention worked. Productivity is not restored by naming the consequence more precisely. It is restored by changing the causes that keep producing it.

The bill for the softer explanation

The softer explanation is always available. Productivity declined because labor markets changed. Productivity declined because supply chains became more complex. Productivity declined because the easy gains were taken. Productivity declined because demand changed. Productivity declined because technology has not diffused evenly. Productivity declined because companies outsourced too much. Each explanation may contain part of the truth. None is sufficient if it leaves the cascade untouched. The harder explanation is that many companies trained themselves to survive productivity decline in ways that made productivity harder to recover. They used mitigation as management. They used outsourcing as relief. They used dashboards as control. They used ROI gates as discipline. They used escalation as speed. They used heroics as operating capacity. Each response had a reason. The system absorbed the reasons and converted them into recurrence. That is why this is not a story about failure at the edge. The edge often knows first. Operators know when workarounds become normal. Supervisors know when the schedule is no longer believable. Engineers know when the same defect is being explained instead of removed. Planners know when lead time is hiding uncertainty. Supply-chain teams know when supplier calls have replaced supplier capability. Finance knows when volatility has become guidance risk. The problem is not that nobody sees. The problem is that seeing is not the same as being allowed to correct.

Michael Carroll | The Future Was Spent on the Recovery

Michael Carroll | The One-Degree Dispatch | Industrial Productivity, Causality, and Market-Shaping Enterprises

If the enterprise wants a different outcome, it has to stop treating productivity as an efficiency project and start treating it as a causal control system. That means protecting the capabilities that make controls possible. It means designing controls that improve conditions rather than merely slow embarrassment. It means governing outsourcing by learning content, not only cost. It means funding capability before consequence can defend the spend. It means measuring decision latency because time is where the hidden bill accumulates. It means asking whether every recurring mitigant is actually a clue to a missing control. This is not a call to nostalgia. It is not an argument to bring everything back inside the walls. It is an argument to know what must never be lost, because once it is lost, the company will spend the future trying to buy back its right to correct. The tragedy was not simply that manufacturing productivity declined. The tragedy was that the consequences of decline captured the very capability required to reverse it. That is the bill now sitting in the system.

References

This article draws on the Bureau of Labor Statistics' 2026 Monthly Labor Review analysis of manufacturing productivity growth across recent business cycles, the Federal Reserve Bank of New York's 2024 and 2025 work by Danial Lashkari and Jeremy Pearce on the broad slowdown in U.S. manufacturing productivity and the R&D puzzle, the Bureau of Economic Analysis guidance on how imported inputs and offshoring enter GDP and productivity measurement, Susan Houseman's Federal Reserve and Journal of Economic Perspectives work on offshoring bias in manufacturing productivity and value added, and the BLS research by Lucy Eldridge, Michael Harper, and coauthors on imported intermediate inputs and their contribution to productivity. The causal and organizational logic is informed by W. Edwards Deming's work on improvement as a learning system, Herbert Simon's bounded rationality, Cyert and March's behavioral theory of the firm, W. Ross Ashby's Law of Requisite Variety, and Judea Pearl's distinction between observation, intervention, and counterfactual reasoning. The article also builds on Michael Carroll's prior One-Degree Dispatch work on decision latency, the permission staircase, non-earning complexity, market-shaping enterprises, the learning speed limit, and the principle that data does not become evidence until there is a question.

Michael Carroll | The Future Was Spent on the Recovery