The Metric That Exposes if Your Transformation Success is a Lie

Markets need a story that can be told in quarters, not years. It is entirely reasonable to look for proof in what was installed and what was used. The trouble begins when those proofs become substitutes for the thing they were supposed to serve. A dashboard can show where the problem sits and still leave the enterprise dependent on the same number of coordinators, schedulers, supervisors, analysts, expediters, and recovery teams to keep the plant from drifting off plan. A planning system can be called modern while the planners still spend their mornings adjusting yesterday’s answers. An AI pilot can be judged successful because it produced good screens and good summaries while the business still waits on the same people, in the same rooms, to decide what happens next. The machine may look newer. The burden is not. What would have to be true for this outcome to keep repeating. The answer is not flattering. The company would have to be measuring the presence of tools rather than the removal of work. It would have to confuse visibility with control. It would have to treat human compensation for weak design as normal operating expense rather than as evidence of architectural failure. It would have to assume that because the enterprise can explain more of what is happening, it must therefore be better at changing what happens. That assumption has carried more board packets, more operating reviews, and more capital plans than many leaders would care to admit. If the labor burden does not fall, the business has not changed. The strongest internal metric for real change is not fashionable, but it is brutally honest. It is percent cost improvement per labor hours consumed. Not labor cost in the narrow payroll sense. Not headcount. Not direct labor alone. The phrase matters because it reaches past the visible workforce and into the total human effort required to make the system function. Direct labor belongs inside it. So do indirect hours tied to execution, quality containment, rework, manual rescheduling, exception handling, data cleanup, supervisory intervention, reporting, expediting, and the recovery work that piles up whenever the process is too unstable to trust on its own. That definition changes the argument at once. The issue is no longer how many people the company employs. The issue is how much human effort the operating model still requires to produce acceptable results. That is a different question, and a more dangerous one, because it exposes the hidden tax most companies carry without naming. The deepest cost in many industrial systems is not labor rate. It is labor dependency. That dependency shows up where executives least want to look. It shows up in the planner who still needs two hours each morning to reconcile what the system said with what the plant can actually do. It shows up in the production manager who keeps a separate spreadsheet because the official one cannot be trusted after noon. It shows up in the quality team that catches defects late, then spends the week tracing, sorting, containing, and explaining. It shows up in the CFO who sees margins improve on paper while working capital absorbs the gain. It shows up in the plant

that claims higher efficiency while supervision rises, overtime grows, contractor reliance deepens, and every hard week ends with the same people standing in the same room trying to steady the line. What passes for proof in manufacturing is often proof of something else. Project count proves appetite. System deployment proves spend. Adoption proves compliance. OEE proves performance inside a lane. Revenue per employee proves little unless price, mix, acquisitions, divestitures, and business model differences are stripped out. Margin proves even less than many boardrooms assume. It can rise because price rose. It can rise because commodity costs fell. It can rise because a business was sold, a charge rolled off, or a weak quarter disappeared from the comparison. A company can post a cleaner earnings line and still require the same amount of human effort to operate. This is why so many transformation stories sound better than they feel. The narrative is built around the visible artifacts of motion. The plant feels the burden that remained. A stronger test begins with a simple distinction. Productivity is doing the work better inside the current model. Transformation is reducing the amount of work the model requires in the first place. Productivity matters. Nobody should pretend otherwise. A line that runs better is better. A schedule that holds is better. A plant that cuts scrap is better. But the enterprise can collect those wins and still leave the burden structure mostly intact. It can run harder inside the same architecture and call that a new era. It is not a new era if the business still needs people compensating for the same design weakness at the same scale. That is why public filings, read with discipline, are more revealing than the language that surrounds them. They do not disclose percent cost improvement per labor hours consumed directly. Almost no company does. But they leave a financial signature. That signature can be read. Not perfectly, because no outside reader sees the whole shop floor. Still, the record is richer than most commentary allows. It shows whether value conversion improved. It shows whether indirect burden appears to be easing. It shows whether inventory is acting as a cushion for indecision. It shows whether cash is following earnings. It shows whether failure costs are consuming the gain. Taken together, those signals tell a story about labor burden even when the company never uses those words. The most useful external approximation is a corrected operating score, not a single headline ratio. Start with adjusted operating margin, because reported margin alone is too dirty. Then strip out what belongs to price and commodity swings, because pricing power is not the same thing as labor conversion. Add the direction of SG&A burden, because much of the hidden work sits in coordination and administration rather than at the point of production. Add the direction of inventory days, because inventory is often the physical residue of weak synchronization and late decisions. Add cash conversion, because clean improvement should turn into cash rather than disappear into working capital and cleanup. Then subtract what the filing itself confesses about failure, restructuring, strikes, recall campaigns, and quality breakdowns. Those are not side issues. They are evidence that the business is still paying people to recover from itself.

Once that correction is made, the contrast across public manufacturers becomes sharper and harder to evade. ITW stands out because the company’s recent results show the kind of consistency that usually accompanies a lighter burden structure. In 2025 it reported $16.0 billion of revenue and 26.3 percent operating margin, and management said enterprise initiatives contributed 130 basis points. Six of seven segments expanded operating margins, and three segments ran above 30 percent. The numbers matter less as boasts than as clues. They suggest a company whose system is not forcing the same degree of manual compensation every year. The plant may still work hard, but the enterprise appears to require less effort to hold the result together than weaker peers do. Parker Hannifin tells a similar story, though with its own operating language. Fiscal 2025 sales were $19.9 billion. Segment operating margin was 23.0 percent, or 26.1 percent adjusted, and cash flow from operations reached $3.8 billion, or 19.0 percent of sales. Management credited the company’s business system, but the proof is not in the phrase. It is in the pairing of margin quality and cash quality. When those two move together over time, the business is often doing more than controlling cost. It is reducing the amount of human drag needed to keep the machine running. PACCAR belongs near that group, though it is a less pure read because parts and financial services influence the result. Even so, 2025 revenue of $28.44 billion, net income of $2.38 billion, $4.42 billion of cash from operations, and disclosed investments in manufacturing efficiency point in the same direction. A company does not produce that kind of profile by chance. It usually arrives there through tighter conversion and lower friction across the operating model. The opposite cases are more instructive because they make the mechanism visible. Boeing’s filings read like a ledger of labor burden that the business could not fully escape. The company disclosed production quality issues, labor instability, supply chain delays, traveled work, higher production costs tied to slower recovery, and the consequences of long union strikes. Those are not abstract risks. They are labor hours being consumed in containment, rework, coordination, recovery, and delay absorption instead of clean value creation. The company had revenue recovery in 2025, but the filing itself still describes a system paying dearly for the gap between signal and control. Stellantis offers the same lesson in a different form. In 2025 it reported net revenues of €153.5 billion and a net loss of €22.3 billion, with negative industrial free cash flow and major restructuring costs tied mainly to workforce reductions in Europe. At the same time, management described broad rework in manufacturing and quality management and the hiring of more than 2,000 engineers. That combination matters. When a company is spending at that scale to reorder quality, reduce issues, cut roles, and rebuild capability, it is admitting that the prior system consumed too much labor and attention for the outcome it produced. Whirlpool sits in the middle, and middle cases are often the most useful because they prevent the thesis from becoming too tidy. It is not a company in public collapse. It is also not a company showing elite conversion of effort into result. Its 2025 report described about $16 billion of annual net sales, 41,000 employees, and a like for like ongoing EBIT margin around 4.7 percent,

while also recording restructuring charges tied to employee termination and asset impairment. That profile does not show a broken enterprise so much as one that still appears to require too much organizational effort relative to what the better operators can produce. The middle group proves why this argument must be made with restraint. Caterpillar, Deere, Honeywell, and Cummins do not fit into easy moral categories. Caterpillar’s 2025 results showed higher sales and revenue, but also unfavorable price realization, unfavorable manufacturing costs, and higher restructuring costs that weighed on the year. Deere operated through a weaker agricultural cycle, yet tied its performance to structural improvements, operational efficiency, inventory management, and cost control. Honeywell and Cummins each carried one time charges that make any simple reading of reported margins unreliable. These cases do not break the thesis. They sharpen it. They show why anyone serious about labor burden must correct for price, charges, and cycle rather than grabbing the first headline that supports a prior belief. The market rewards earnings. The plant pays for the way they were earned. A fair critic could say labor is not the largest cost in every industrial business, and therefore cannot be the master metric. That objection has force. In many sectors, materials, capital intensity, or freight may dwarf direct payroll. Yet the point is not that labor is always the largest line. The point is that labor burden is one of the clearest enterprise wide traces of instability, weak coordination, slow decisions, poor quality control, and operating mismatch. It is the place where many other weaknesses finally cash out. The company may not lose the most money there on paper. It often reveals the weakness there first. Another critic could say margin improvement can never be trusted as evidence of labor improvement. That is also true, which is why raw margin should not be trusted. A serious read strips out price. It corrects for commodity swings. It discounts gains flattered by one time items. It penalizes restructuring and failure costs. It asks whether the gain turned into cash. It asks whether inventory eased rather than merely moving the problem. The metric earns credibility only when it bears the weight of those corrections. Then there is the hardest objection. Some industrial systems improve by moving labor off the books. Contractors rise. Third parties take over. The expense shows up in another account and the company claims victory. That problem is real. It is one reason public inference will always be inferior to internal truth. An honest internal metric has to include contractor hours, temporary labor, outsourced operations support, and any third party effort that functionally replaces plant, quality, maintenance, planning, or recovery labor. Anything less is accounting, not measurement. Even with those limits, the standard remains more useful than the softer ones it replaces. In fact, its usefulness rises because it resists easy praise. It asks a management team to connect time, money, and architecture in one sentence. If the plant is better, why are planners still manually bridging the same gaps by noon? If quality has improved, why are containment labor, trace work, and recovery meetings still rising? If the business is more advanced, why does cash keep lagging

earnings? If scheduling is more intelligent, why does inventory still act like a hiding place for indecision? Those are not decorative questions. They are the kind a chief executive can read aloud in an operating review and know, almost from the room’s reaction, whether the company has been measuring motion or change. A second set of questions belongs at the board table. If the company says the operating model has improved, what burden disappeared, exactly? Which roles spend fewer hours compensating for uncertainty than they did two years ago? Which meetings no longer need to happen? Which reports no longer need manual cleanup? Which decisions now happen with less escalation, less interpretation, and fewer handoffs? If no one can answer that cleanly, the board is not hearing about transformation. It is hearing about activity. A system is not better because it explains more. It is better because it requires less compensation. This is where the language around AI has made the problem harder, not easier. The market is full of products that improve visibility, summarize issues, point to anomalies, and help people discuss the next move. Some of them are useful. Many are well built. But a useful assistant is not proof that the business has changed. If the same manager still must decide, route, verify, and stabilize the action before anything material happens, then the burden structure has not been altered very much. Better words have been added to the same chain of dependence. The same trap applies to automation. Companies often automate what is easy to see rather than what is costly to carry. A machine task disappears, yet the scheduling burden around it rises. A planning step is automated, yet manual exception work multiplies because the plant does not behave the way the model assumed. A quality control check becomes faster, yet the cost of late correction remains because the decision rights around intervention never changed. It is entirely possible to buy modernity and keep the same burden. That is why the strongest industrial operators rarely sound theatrical in their best years. They sound disciplined. The language is often plain. The record shows cleaner conversion, steadier cash, lower drag, fewer apologies, less need to explain away failure, and less evidence that the enterprise is spending human effort where architecture should have done the work. The weaker operators often sound the same at first. Their decks are current. Their language is current. Their systems are current. The difference appears lower in the filing, where the business admits what it is still paying for. The signature of progress is not more information. It is less effort spent holding the result together. There is a prediction here, and it is testable enough to be worth the risk. Over the next several years, the public companies that create the most durable industrial value will not be the ones with

the most pilots, the loudest digital language, or the broadest claims about intelligence. They will be the ones whose operating records show the cleanest fall in labor burden once price, cycle, and one time effects are stripped away. Their filings will show stronger cash conversion, steadier margins, fewer failure charges, less inventory drag, and less need to explain why progress has not yet reached the floor. If that proves wrong, then this thesis deserves to be discarded. If it proves right, much of the current discussion will look like what it has often been. A story told around the edges of the real issue. The real issue is not whether industry has more data than it once did. It does. The real issue is whether that data has reduced the amount of human effort required to make the business function. That answer cannot be guessed from adoption, praised from slideware, or inferred from one good quarter. It has to be read through the burden the system still imposes on the people inside it. A manufacturer has not changed because it bought software, connected machines, trained users, or built a command center. It has changed when the enterprise now needs materially less effort to produce equal or better results, with stronger quality, better cash, and less drift between signal and action. That standard is harder. It is also cleaner. It does not flatter management for installing things. It asks what the business no longer has to spend human energy doing. That is the metric most manufacturers still refuse to use because it exposes too much. It exposes where the burden really lives. It exposes whether productivity was local while drag stayed enterprise wide. It exposes whether the company has been digitizing its existing complexity instead of removing it. It exposes whether the system has become easier to run or merely easier to describe. Boards should want that exposure. Chief executives should want it more. The next decade of industrial value will not belong to the companies that can talk best about change. It will belong to the ones that can prove the operating model now requires less human compensation for weakness. That proof is harder to stage. That is why it matters. References This argument draws on the operating and financial record of public manufacturers that expose the mechanism more clearly than most strategy language does, including Illinois Tool Works’ 2025 results, Parker Hannifin’s fiscal 2025 results and fact sheet, PACCAR’s 2025 annual report and 2025 results, Boeing’s 2025 Form 10 K and related annual report disclosures on quality, labor disruption, and recovery, Stellantis’ 2025 full year results and annual report, Whirlpool’s 2025 annual report, Caterpillar’s 2025 full year results, Deere’s fiscal 2025 results, Honeywell’s 2025 results, and Cummins’ 2025 results, all of which help separate clean conversion from price effects, one time charges, inventory drag, cash realization, and failure burden. The conceptual ballast comes from W. Edwards Deming’s Out of the Crisis in 1986 on variation and

management failure, Herbert Simon’s work on bounded rationality from the 1950s onward, Ronald Coase on the cost of coordination inside the firm, Judea Pearl’s work on causality and intervention, and Michael Carroll’s own writing on operating architecture, decision latency, permission, and the difference between adding tools to a system and reducing the burden the system places on the people forced to carry it.