The Staircase Is Cutting Your Margin In Half

The Permission Staircase Is Cutting Your Margin In Half! A New Operating Code for CEOs, COOs, and CFOs

It always starts smaller than the postmortem suggests. Not with a headline failure. Not with smoke in the parking lot. With a Tuesday. At Plant 14, Tuesday opens like any other. The high margin product that pays for half the capital budget is running on the main line. The product everybody points to when they talk about “our franchise.” The MES pings a little more often than usual, scrap ticks up from 1.5 percent to 3.5 percent, a few pallets are flagged for retest. The shift supervisor types exactly what you would expect into the log. “Keep an eye on this.” Maintenance has seen the vibration trend on the main drive start to walk away from normal. Not a cliff, a drift. They talk about it at the board and decide it will not survive a capital request queue that is already full of louder problems. Planning sees schedule adherence wobble and explains it away with mix, a call off, a rough weekend. Every signal is individually defensible. Together, they are a warning.

The warning does not trigger action. It buys a ticket to the staircase. It goes on the agenda for the weekly quality call. That call slides out a week because key people are traveling. When the call finally happens, the group does what high functioning corporate citizens are trained to do. Ask for more data. Put it on the divisional council agenda. “Let us look at this properly when all the right people are in the room.” The first slot with the full cast is ten days out. By the time they meet, four weeks of production have gone through Plant 14. The drive finally fails hard. You are in a war room staring at a month of product that might, or might not, behave in the field. Plant 14 runs about 250 million dollars a year of this product family. Roughly 62.5 million a quarter. At a healthy 25 percent EBITDA, that quarter should throw off about 15.6 million of earnings when things run as planned. When the “manageable excursion” is allowed to run four full weeks out of control, scrap, rework, premium freight, concessions, and lost volume quietly strip about 5 percent of that quarter’s revenue from the P and L. Roughly 3.1 million. That is how a 25 percent quarter falls to about 20 percent. The denominator has not changed. The 62.5 million is still there. You simply collected 12.5 million of EBITDA instead of 15.6. You bought four weeks of delay for about 3.1 million. Here is the part the staircase hides. The damage profile is time based. Every extra day you run out of control quietly prints more bad product, more bad shipments, more recovery overtime, more angry calls, and more capital trapped in the wrong form. If you cut the out of control window in half, you cut the area under that damage curve in half. Now replay the same physics with a two week operating code instead of a four week staircase. The asset still wears. The process still drifts. People still make mistakes. The difference is that guardrails and pre authorized moves stop the bleeding at roughly two weeks instead of four. If the cost of the event scales with how long you drift, the 3.1 million EBITDA hit collapses toward 1.5 to 1.6 million. Same problem. Half the time drifting, roughly half the damage. The title is not poetry. It is a description of the math. No one books a project called “Decision Latency. Minus five hundred basis points.” It just appears as freight, overtime, write offs, penalties, working capital that will not come home on time. The root cause report lists process drift, training gaps, supplier variation, asset health. All true. All incomplete. The real cause lives in the quiet gap between when the system knows and when the company moves. That gap is your operating model. That gap is decision latency.

When we say the staircase is cutting your margin in half, we are not claiming your total corporate EBITDA falls from eighteen percent to nine. We are pointing at a repeatable pattern. Every avoidable excursion that runs four weeks instead of two roughly doubles the cost of that event. Cut the drift time in half and you cut that avoidable damage in half. The math at Plant 14, and across the network, is simply that pattern written in dollars.

The line item you never see

CFOs can recite their P and Ls the way some people recite scripture. They know the cost buckets, the variance drivers, the usual excuses. They know how much weather they are supposed to blame and how much “market conditions” they are allowed to hide behind. What is almost never visible in those numbers is the time it takes the enterprise to convert knowledge into irreversible action. The time from the first reliable signal to the moment you commit to a path that actually changes the outcome. That is decision latency. Not the time from meeting to meeting. Not the time from dashboard to deck. The time between knowing and doing. Once you look at the business through that lens, you notice something uncomfortable. You are not suffering from a lack of information. You are suffering from a surplus of permission. Planning sees the plan is wrong in real time and waits thirty days for S and OP to make it official. Maintenance watches patterns whisper “fix me now” and waits for approval chains designed for an era that did not have sensors. Quality teams see bands breaking and wait for signatures to touch a spec. Logistics can see bottlenecks forming in their network long before a truck moves but lacks authority to reroute until the call. Individually, these delays look like prudence. Collectively, they form a staircase of purchased delay. Each step bought with a story about risk, consensus, and governance. Each step quietly printing cost. You do not experience it as one large hit. You experience it as a thousand small, familiar nuisances. “We had to expedite a lot that month.” “We wrote off some inventory.” “We lost a little share in that channel.” Add them up and you are not talking about noise. You are talking about margin.

Focus, drift, and the architecture of attention

In our work across plants, enterprises, and The COO Council, one pattern keeps returning. Performance is not limited primarily by intelligence or effort. It is limited by attention.

Roughly twenty percent of outcomes come from choosing the right things to pursue. Another twenty percent comes from doing those things with competence and discipline. The remaining sixty percent is whether the organization can keep its attention on those things long enough for them to matter and compounding to work. That is the 20.20.60 problem. Strategy, execution, focus. Decision latency is how your architecture steals from the sixty. Every time a decision has to climb the staircase, you turn problem solvers into petitioners and leaders into interpreters. People spend their days negotiating permission, not fixing the problem. This is where we confuse virtue and weakness. We tell ourselves we are being careful, thorough, aligned. What we are often being is late. The question that follows is not rhetorical. How many decisions should be made at the edge. How many should be coordinated across plants and functions. How many genuinely require the C suite and the board. Most enterprises answer this accidentally, and they answer it badly. The actual ratio, if you reverse engineer calendars and email trails, looks something like this. A minority of decisions. maybe twenty percent. are truly owned at the edge. Another minority are genuinely run on shared platforms with clear rules. The great, soggy middle plus all the ambiguous risk gets pushed upward “just to be safe.” That inversion is why your people feel overloaded and your company feels slow. The alternative is not a motivational slogan. It is an operating code. Sixty. Twenty. Twenty. Sixty percent of decisions pre authorized at the edge inside explicit guardrails. Twenty percent coordinated across the network with shared platforms, playbooks, and agentic assistants. Twenty percent reserved for true enterprise trade offs that should command leadership and board attention. Everything else is theater. The exact percentages are design intent, not sacred math. In a real enterprise, the current split might look more like 15.25.60 or 20.20.60 once you trace where decisions actually land. The pattern is what matters. Too few decisions live safely at the edge. Too few sit on real shared platforms with clear rules. Too many climb the staircase under the banner of “alignment” and “governance.” Sixty.Twenty.Twenty is not a survey result. It is the operating code you write on purpose to reverse that pattern.

What the staircase really is

The staircase was not built by villains. It was built by previous generations who were trying to protect the company from irreversible mistakes in a world of scarce data and slow communication. You climbed the staircase to collect facts. You climbed to make sure everyone heard the same story. You climbed because the cost of a wrong move was high and you could not easily replay history. Today the facts arrive before the meeting invite is even sent. Your assets, networks, and customers are instrumented. Your data lake fills itself. The bottleneck has moved. The scarcity now is not information. It is permission. The staircase has not kept up. In that mismatch, you see the same pattern over and over. Operators and local leaders see the problem early and understand it early. Systems can show it, often in painful detail. The enterprise moves late. You are not running a control system. You are running a superstition. The superstition that risk lives primarily in the hands of the people closest to the work instead of in the architecture that forbids them from using what they know. The staircase is not a governance mechanism. It is a tax on the future.

The Plant 14 film. Once at four weeks, once at two

Return to Plant 14. First, watch the four week film again. Week one, the signals appear and are logged. No line is slowed, no asset is taken down, no product is quarantined. Week two, the pattern is obvious in hindsight, but still treated as an inconvenience. The problem is nominated for attention on the next quality call. Week three, the meeting slides. The problem continues. Product flows. Customers are served a little more risk with every pallet. Week four, the group meets, orders more data, drafts a plan, and then reality makes the decision for them. The drive fails. Now everyone is “aligned” around an outage, a write off, a recovery plan, and a set of apologies.

That is how you lose three point one million of EBITDA in a quarter at a single plant without ever making one big, obviously stupid decision. You leak it through time. Now spool a different film. Same plant, same physics, different operating code. In this version, Plant 14 runs on sixty twenty twenty. The guardrails for the high margin family are defined in numbers, not hopeful prose. If scrap and rework cross a defined band, the supervisor has the authority to slow or stop the line. If vibration and amperage on that specific drive follow a specific pattern, maintenance has the authority to take it down on planned time, not just when it dies. If quality sees two defined triggers hit in a window, they have authority to contain, re sequence, and call for help without waiting for the meeting. Every action and its outcome is written to a ledger in the flow so that risk, audit, and leadership can see what happened without playing email archaeology. In week one of this film, the same signals appear, but they do not buy a ticket to the staircase. The line is slowed, suspect lots are contained, and the asset is taken down under controlled conditions. In week two, the cross functional group still meets, but they are meeting around a contained, local incident. They are validating and tuning what has already been done, not discovering the problem for the first time. Two weeks of damage, not four. Two weeks of scrap and rework and service headaches, not a month. In financial terms, the hit is still real. Perhaps one point five or one point six million of EBITDA. The asset still needed work. The process still drifted. People still made mistakes. But because the out of control period was cut in half, the area under the damage curve. the total pain. is cut roughly in half. That is how time and money actually talk to each other in your operation. You do not need an advanced degree in anything to see the pattern. Every extra day you run out of control prints more bad product, more bad shipments, more emergency overtime, more calls from customers, more capital trapped in the wrong form. Cut the days in half and you cut the damage in half. The staircase is four weeks. The operating code is two.

From one plant to the enterprise

Plant 14 is not alone. Every large manufacturer has a flock of Plant 14s. Ten sites here, twelve there, split across divisions and P and Ls.

If an excursion of this scale hits each critical plant three times a year. not a worst case, just the quiet background rhythm of reality. the numbers climb very fast. Three million of EBITDA lost per event, per plant in the four week world. Three events a year. Ten plants. You are now in the neighborhood of ninety million of earnings quietly leaking away. On a five billion dollar company aiming for eighteen percent EBITDA. nine hundred million of earnings. that ninety million is almost two full points of margin. You will never see a board pack with a line that says “Decision Latency. minus 190 basis points.” You will see a collection of familiar narratives. “Unexpected” quality costs. “Temporary” service issues. “One time” freight. “Unusual” scrap. The board hears a story about macro headwinds and competition. The real story is that your architecture routinely takes four weeks to respond to problems your own systems surfaced in week one and your own people understood in week one. Now take the same enterprise and run the two week film. Same incidents. Same plants. Same product families. But an operating code that refuses to let a problem drift for a month. The high value excursions are forced back into control inside a two week window. Ninety million of annual EBITDA loss becomes something closer to forty five. The company that seemed stuck at sixteen percent can suddenly print seventeen without a new product, a price increase, or a headcount reduction. Just by refusing to keep buying two extra weeks of false comfort. That is what we really mean when we say “cuts decision latency in half.” Not that every decision is twice as fast. That the time you spend out of control is cut roughly in half. Which is exactly where most of the money hides.

Where AI fits and where it does not

None of this requires artificial intelligence to begin. It requires courage, arithmetic, and a willingness to write down who decides what. But the moment you put guardrails, ledgers, and cadences in place, Industrial AI can finally be more than a storyteller. Most systems sold as “intelligent” today are interpretive. They describe the world with breathtaking detail then hand all the work back to humans. They produce “actionable insight” and “intelligent dashboards,” both of which are polite ways of saying the staircase remains in charge.

In a sixty twenty twenty world, AI becomes something different. It becomes an automated scientist living inside the operating model. At the edge, agents watch the same sensor data and ledgers as your people. They learn the patterns that matter in your context, propose moves inside guardrails, and execute small, reversible actions you have pre authorized. Not because a vendor promised “autonomy,” but because your operating code explicitly granted permission. In the middle band, agents run the network simulations human teams never have time to run. They test counterfactuals. If we had rerouted here, what would have happened to service, cost, and risk. If we shift this family there, what happens to working capital and resilience. They do not merely enrich the slide. They shape the decision. At the top, they sharpen judgment. They bring a causal map of second and third order effects into the boardroom so that when you move capital or change a promise, you do so with fewer blind spots and fewer self inflicted wounds. The line between fake intelligence and the real thing is simple. If your operating model cannot act differently tomorrow because of what the system learned today, you bought narration, not capability. Sixty twenty twenty is the difference. It gives AI a place to plug in where permission is already coded.

What a serious board actually does next

The easy move now is to treat this as another clever metaphor. Nod, agree, and wait for the next priority to knock it out of the agenda. A serious board does something harder. It picks one flow and insists on seeing the clocks. Take the franchise product that pays the bills. Take the plant that always shows up in explanations. Take the lane or customer that shows up in every “we had a rough quarter because” speech. Ask four questions. When did the system know the last time something went wrong. When did a human admit it. When did we decide. When did we truly recover. Do not accept narratives. Ask for timestamps. Then ask a fifth question at each step. What exactly were we buying with this delay.

If the answer is safety, regulatory exposure, or irreversible consequence, invest in better guardrails and better sensing. If the answer is habit, fear, or “this is how we have always done it,” you have found the staircase in its purest form. Then demand one concrete change. One decision written into local guardrails. One decision elevated properly to a shared platform. One decision formally reserved for the top of the house. You do not need an army of consultants or a new transformation office to start. You need the discipline to make permission visible and the courage to declare that four weeks of drift is no longer acceptable when the system knew in week one. Do that in a single value stream for a single quarter and your own numbers will tell you whether this is real. You will not need a white paper. What you will need, once you see it, is an answer to a final question. Are you comfortable continuing to pay one or two points of margin every year to a staircase built for a world that no longer exists. Or are you ready to give your enterprise a new operating code. One that lets it move at the speed of its own information, not the speed of its own calendar. Sixty. Twenty. Twenty is that code. It does not promise perfection. It promises something rarer. A company that stops pretending the problem is in its people, its tools, or its fate. A company that finally admits the truth. The real enemy has been the staircase all along. And the moment you cut it in half, so go half your excuses. References This argument draws on and extends our prior LNS Research and The COO Council work on decision velocity, the Industrial Productivity Index, and the architecture of permission rather than citing it verbatim, including The Line Item Every CEO Pretends Not to See, The Architecture of Permission No One Admits They Are Running, How to Spot Fake Intelligence Before It Destroys Your Company, From Data to Dominion. How Automated Scientists Will Redraw the Boundary Between Insight and Control, and our internal 20.20.60 focus model and Decision Velocity scaffolds for COOs. It is also informed by ongoing One Degree World work on access as the real architecture of control, the emerging One Degree for Everyone and Everything material, and draft COO Council guidance on purchased delay and the economics of escalation. Beyond our own work, the piece leans on and interprets ideas from Francis Bacon’s Novum Organum on causal knowledge, Judea Pearl’s Causality and The Book of Why on structural causal models and counterfactual reasoning, W. Edwards Deming’s Out of the Crisis and

Stafford Beer’s Brain of the Firm on variance, systems, and cybernetic control, Herbert Simon’s and James March’s work on bounded rationality and organizational decision processes, Richard Thaler’s Misbehaving and Nudge on choice architecture and hidden taxes on behavior, and Amartya Sen’s Development as Freedom on agency and institutional responsibility. It is further colored by contemporary operating and governance framings, including NIST’s thinking on Zero Trust as an access and permission model, emerging Industrial AI and agentic systems discourse, the EU AI Act and OECD AI principles, and classic civic framing from Abraham Lincoln’s 1862 messages to Congress and Winston Churchill’s war cabinet practice, which we reinterpret here as a practical operating code for CEOs, COOs, CFOs, and boards who are now competing on decision latency rather than on dashboards.