Knowledge

A system is a network of components working together toward an aim. Drop the aim and the components stop being a system — they become a heap. Most companies look organized but operate as heaps: sales, marketing, ops, support, each optimizing its own number, none of them aimed at the same place.

Deming's first move is always to ask: what is the aim of the system? If the team can't agree, the rest of the conversation is decoration. Once the aim is named, the components have a way to be judged — not by their own KPI but by whether they make the system better at the aim.

The dangerous corollary: improving a component in isolation usually makes the system worse. A sales team that hits its number by promising features the product can't ship has improved sales and hurt the system. The local optimum is almost always the systemic enemy.

Every process produces variation. Two days are never identical. Two units off the same line are never identical. The question Deming forced managers to answer is: is this variation routine — baked into how the system currently works — or is something unusual happening?

Routine variation he called common cause. Unusual variation he called special cause. The whole game is telling them apart, because the response is opposite. You react to a special cause: hunt it, fix it, document it. You do NOT react to a common cause one point at a time — you redesign the system that produces it. Treat common-cause variation as if it were special, and you make the system worse, every time.

Most managers do exactly this. A bad month produces a frantic plan. A good month produces back-slapping. Both reactions assume the variation meant something it didn't. Deming's discipline: stop reacting to noise. Build the chart that lets you tell signal from noise. Then react only to signal.

Deming's claim: there is no learning without a theory to learn from. Experience without a theory is a stack of anecdotes; you remember them, you repeat them, but you don't get better, because you have no model that can be wrong.

The mechanism for getting better is the Plan-Do-Study-Act cycle. State a theory. Predict what would happen if it's right. Run a small test. Compare the prediction to what actually happened. Revise the theory. Run it again. The cycle is not a bureaucratic ritual — it's the only way an organization can be sure it's learning rather than drifting.

Most companies have lots of activity and very little learning. People act on theories they never wrote down, get results they never compared to a prediction, and tell stories about it after. PDSA is the discipline of making the loop legible so it can compound.

Systems are run by humans. Humans have intrinsic motivation — the desire to do good work, to be seen, to be part of something — and most management practices are designed, accidentally, to crush it. Annual reviews, forced rankings, rah-rah quotas, pay-for-performance: all of them assume the worker is the problem and a sufficiently clever incentive will fix them.

Deming's reply was that the worker is almost never the problem. Ninety-something percent of the variation he saw came from the system, not the person. Punishing people for the system's outputs is both unfair and counterproductive: it teaches them to game the metric, hide bad news, and stop trusting management. The trust is what you actually needed.

The psychology lens asks managers to stop being judges and start being engineers of conditions. Make the work itself rewarding. Take fear out of the room. Give people the tools to do good work and then trust them to do it. The result is not soft management — it is rigorous management of the system rather than scapegoating of the people.

Toward improvement of product and service, with the aim to become competitive, to stay in business, and to provide jobs. The first point is the foundation: an organization with a clear, slow-changing purpose can absorb tactical setbacks because the long arc is decided. An organization without one re-litigates strategy every quarter and calls the noise progress.

Constancy means the purpose survives a bad month, a new investor, a competitor's launch. It is what the next CEO inherits. If yours fits on a coffee mug and changes when the strategy deck does, you don't have one yet.

We are in a new economic age, Deming wrote in 1982 — and the management practices that built the postwar boom (acceptable defects, blame the worker, reward the heroic firefighter) cannot survive in a market where customers compare on quality and walk on a single bad experience. He was right early; he is more right now.

Adopting the new philosophy means the leadership team itself has to change how it thinks, not roll out a program. Programs come from the same management style that produced the problem. Transformation comes from leaders absorbing the four lenses and slowly stopping the things they used to think were normal.

Inspection at the end of the line catches some defects but builds quality into nothing. By the time you find the bad unit, the system has already produced it. The cost is paid; you're just deciding whether to ship it.

The fix is to build quality into the process upstream — to design the system so that producing a good unit is the natural outcome and producing a bad one would be unusual. Mass inspection becomes an indictment, not a strategy. If you need a wall of QA at the end, the system upstream is broken.

Lowest-bid procurement seems prudent and is usually expensive. The cheapest input ships rework, returns, support tickets, and a relationship that breaks the moment something goes wrong. The bid was a number; the cost shows up downstream as a thousand smaller numbers nobody attributes to it.

Deming's alternative is long-term relationships with suppliers chosen on total cost of use. Move toward single-source partners who share the aim. The procurement department's job stops being to drive the price down and starts being to drive the system's total quality up.

The system that produces today's output will not produce tomorrow's growth. Every process is a candidate for improvement, every day, by everybody who runs it. The work is continuous because the world is continuous: customers change, suppliers change, the team changes, and a system that stops improving falls behind without anyone noticing.

Concretely: every team has a way to surface and try improvements; every improvement runs through PDSA; every learning is captured so the next person doesn't repeat the experiment. This is not a campaign. It is the steady-state metabolism of a healthy company.

Workers are usually trained by the previous worker, who was trained by the worker before them. Variation accumulates: by the third generation everybody is doing the job slightly differently and nobody can tell you which version is actually correct.

On-the-job training, done well, is grounded in the operational definitions of the work — written down, demonstrated, supervised, signed off. It is one of the highest-leverage investments management makes and one of the most under-invested. New hires deserve to be set up to succeed; they almost never are.

Supervision and leadership are not the same thing. Supervision watches the worker; leadership improves the system. Most managers were promoted for being good individual contributors and were never taught how to do the second job. They keep doing the first one with a bigger title.

A leader's job is to make the system the people work in better. That includes understanding variation, removing barriers, advocating for the resources the team needs, and protecting them from drive-by demands from above. It does NOT include grading them on outputs the system controls.

Fear is information loss. A team that's afraid to surface bad news doesn't surface it; the bad news compounds in silence and lands as a crisis later. Fear is the most expensive cost most companies refuse to measure, because measuring it would require leadership to admit to being a cause of it.

Driving fear out is mostly a leadership behavior, not a policy. Don't shoot messengers. Reward the early call about a problem more than the heroic save once it has metastasized. Make it safe to disagree with the boss. The information you get back when fear is gone is the information you needed to run the system.

Sales books a deal the product can't deliver. Engineering ships a feature the support team can't explain. Marketing runs a promotion legal hasn't reviewed. Each department, run as a silo, optimizes its own metric and externalizes the cost onto the next one in line. The customer pays the integration tax.

The fix is not a reorg. It is rituals — joint planning, shared dashboards, problems escalated by the people closest to the seam — that put the aim of the system above the aim of any one function. The org chart can stay; the wall has to come down.

“Do it right the first time.” “Zero defects.” Posters in the breakroom that ask the worker to do better, while the system that produces the work hasn't changed at all. Deming had no patience for this. Slogans imply the worker is the problem. The worker, by his measurement, almost never was.

If you want better outcomes, change the system. If the system is the same and you've taped a poster to it, you have insulted the people doing the work and improved nothing. Most slogans are the leadership team's way of avoiding the harder job.

A quota — “close 20 deals this quarter”, “ship 8 features by Q3” — substitutes a number for an understanding of the work. People hit the number by gaming, hiding, or sandbagging, and the long-term capability of the system declines as the quarterly figure improves. Deming saw this everywhere; he loathed it.

The alternative is leadership: understand the system, predict its capability, work to improve it, and trust the resulting numbers because they're the natural output. Numbers are a thermometer; quotas are an attempt to make the thermometer read what you want by yelling at it.

Most workers want to do good work. They came in wanting it. Then they discover that the tools are broken, the process is contradictory, the manager grades them on something they don't control, and the only feedback they get is once a year, written by someone who hasn't seen them work.

Pride of workmanship is not a perk. It is a precondition for sustained quality. The leader's job is to remove the barriers that take it away — broken tools, ambiguous expectations, performance theatre, ratings that punish the system's variation. What's left is people doing work they can sign their name to.

Training (point 6) makes you better at today's job. Education makes you better at tomorrow's. Deming was emphatic that the difference matters: an organization that only trains gets stuck at its current sophistication; an organization that educates compounds.

Concretely: time and money for people to learn things adjacent to their work. Statistics, design, psychology, the field's classic books, the operational research you'll need in five years. Most companies cut this first under pressure. The ones that don't are the ones that are still around.

The first 13 points are not management's private project. The transformation is the entire organization's work, and the only way it actually happens is if every person — line worker, manager, exec, board — sees themselves as a participant, not a target.

Practically, this means committees, working groups, shop-floor improvement projects, regular forums where progress is reviewed and obstacles are surfaced. It means leadership signals, repeatedly, that this is the actual job — not a side initiative — and stays at it long after the novelty wears off. Five-year transformation, not a Q3 program.