The machine that changed the world based on the Massachusetts Institute of Technology 5-million dollar 5-year study on the…

1. 1Write the bargain down as two explicit halves. Company gives: lifetime or long-term employment, pay graded by seniority rather than by specific job function, bonuses tied to company profitability, and access to company facilities. Employees give: flexible work assignments and active initiative in improving the system. The Toyota 1946 settlement is the template: "If we are going to take you on for life, you have to do your part by doing the jobs that need doing."
2. 2Make management's half credible before you ask for the workers' half. Put the security guarantee in place and name the sacrifices you will make to keep skilled workers when the market slumps. Lean production has no slack, so the guarantee has to extend job security to factory workers, something historically offered only to valued professionals.
3. 3Build the retention logic that makes a long-term commitment reasonable. Seniority-based wages that climb steeply make leaving costly. A forty-year-old does the same job as a twenty-five-year-old at much higher pay, and a quitter restarts at zero seniority, below even the younger worker. Check that your pay structure supports mutual long-term commitment.
4. 4Train every worker in all the jobs in the work group so tasks can rotate and people can cover for each other. Add simple machine repair, quality-checking, housekeeping, and materials-ordering. This is the concrete content of "flexible work assignments."
5. 5Give the team real delegated responsibility and a way to see problems. Post line status (breakdowns, personnel shortages, overtime) on lighted boards visible from every work station. Expect any employee who knows how to help to go and help when something goes wrong anywhere in the plant.
6. 6Ask for proactive thinking, not reactive fixing. Encourage workers to devise solutions before problems become serious. Treat this as the employees' side of the compact, in exchange for the control over their work environment and the creative tension that lean work gives them.
7. 7Test reciprocity on the shop floor, not on the org chart. Walk the plant and watch whether people actually cooperate. Changing the chart to show "teams" and adding quality circles does little on its own, so judge the compact by observed behavior and morale.
8. 8Monitor for reversion. If management fails to lead or the workforce feels no reciprocal obligations are in force, lean predictably reverts to mass production: people follow the rulebook and no one takes initiative. Treat falling initiative as an early warning of a broken compact.

Toyota, 1946. The company was in crisis, and Kiichiro Toyoda proposed firing a quarter of the workforce. The workers revolted and occupied the factory, and the company union was in a strong position to win the strike. Japan's government, under American prompting, had strengthened union rights and severely restricted owners' ability to fire workers, so power had shifted to the employees. The compromise: a quarter of the workforce was terminated as proposed, Kiichiro Toyoda resigned as president to take responsibility for the failure, and the remaining employees received two guarantees. One was lifetime employment. The other was pay steeply graded by seniority and tied to company profitability through bonuses. They also gained access to Toyota facilities such as housing and recreation. In return the employees agreed to be flexible in work assignments and to promote the company's interests by initiating improvements rather than merely responding to problems. The seniority wage structure, which other Japanese firms copied, made the expectation that workers would stay for their working lives reasonable. This formula remains the basis of labor relations in the Japanese auto industry.

• If you cannot or will not protect jobs in a downturn, then do not ask for flexible assignments and proactive improvement. Workers respond only when management is seen to value skilled workers, make sacrifices to retain them, and delegate responsibility.
• If workers are being treated as a variable cost or as interchangeable parts, then expect lean practices to fail. Reverting to mass-production behavior (heads down, minds elsewhere) is the predictable outcome.
• If you are tempted to introduce teams or quality circles as the main change, then first check whether the reciprocal obligation exists. Without it, formal apparatus produces little teamwork and low morale, as in the GM showcase plants.
• If workers already cooperate despite a rigid contract, then read it as evidence of trust in operating management. The Ford plants had a 1938 job-control contract yet showed lively teamwork because workers trusted managers who worked hard to learn lean principles.
• If the system carries no buffers (no slack, no safety net), then require every worker to try very hard and secure the compact before removing slack. Lean is fragile and cannot run on going through the motions.

• Cosmetic teaming: you redraw the org chart, introduce teams and quality circles, and sign a new team contract. The tell is little observed teamwork and low plant-floor morale, as in the GM showcase plants.
• One-sided demands: you ask for flexibility and initiative but offer no security or sacrifice. The tell is workers doing the minimum by the rulebook and no one taking initiative to improve the system.
• Security without skills or delegation: you promise jobs but never train workers across all jobs or give teams responsibility. The tell is narrow job assignments persisting and workers unable to rotate or cover for each other.
• Cutting workers first when sales dip: you treat the workforce as a variable cost and turn people away at the first downturn. The tell is workers who stop trusting management, and the reciprocal obligation collapses.
• Ignoring retention economics: wages are not structured to reward long tenure, so the long-term expectation is unreasonable. The tell is that leaving costs nothing in earnings, so staying for a working life is not a rational bet for workers.

• ▸Company guarantee offered: state the employment security commitment (lifetime or long-term) and its scope.
• ▸Downturn sacrifices: list what management will give up before cutting skilled workers.
• ▸Pay structure: describe how seniority grading and profit-tied bonuses reward staying, and what a leaver would lose.
• ▸Community rights: list facilities or benefits extended to employees (e.g., housing, recreation).
• ▸Flexibility asked of employees: list job rotation scope, cross-training skills (repair, quality checks, housekeeping, materials ordering).
• ▸Initiative asked of employees: define the improvement and proactive problem-solving expected of each team.
• ▸Delegated responsibility: state what decisions the team now owns and how problems are made visible (e.g., status boards).
• ▸Floor evidence check: record what you observed on a plant walk about real teamwork and morale, not the org chart.
• ▸Reversion warning signs: note any rulebook-only behavior or drop in initiative, and the management response.

1. 1Give every worker the authority to stop the line. Ohno placed a cord above every workstation and told workers to halt the whole assembly line immediately if a problem emerged that they could not fix. In mass-production plants only the senior line manager could stop it. Move that authority to the person who sees the defect.
2. 2Bring the whole team to the stopped station. When the cord is pulled, the team comes over to work on the problem, so the worker who found it is never left to fix it alone.
3. 3Trace each error to its ultimate cause with 'the five why's.' Ask 'why' as each layer of the problem is uncovered, then devise a fix so the error never occurs again. Train production workers in this method; it is not a job reserved for specialists.
4. 4Fix problems at the source rather than at the end of the line. A defective part installed upstream gets compounded by downstream workers. The problem is not found until the end, by which point many similarly defective vehicles have been built. Stopping early prevents this cascade.
5. 5Expect line stoppages and discouragement at first. When Ohno began experimenting, his lines stopped all the time and workers easily became discouraged. Keep going: as teams gain experience identifying and tracing problems, errors drop dramatically.
6. 6Retire rework areas and inspection as your quality strategy. Inspection, however diligent, cannot detect every defect that can be built into a complex vehicle. Measure progress by the falling amount of rework before shipment.
7. 7Supply the skills and the continuing challenge. Lean production gives workers the skills to control their work environment and the ongoing task of making work run more smoothly. Offer a continuing variety of challenges, so people do not feel they have hit a dead end and hold back know-how.
8. 8Back the system with reciprocal obligation. Management must fully support the factory workforce and, when the market slumps, make the sacrifices needed to ensure job security. That security has historically been offered only to valued professionals.

Taiichi Ohno's Toyota plants are the passages' own case. In a mass-production plant, an error such as a badly installed part is treated as a random event: someone repairs it and hopes it does not recur. The defect is compounded by assemblers farther down the line and is not discovered until the very end, by which point a large number of similarly defective vehicles have been built. Ohno put a cord above every workstation. Workers stopped the whole line if they met a problem they could not fix, and the team then came over to work on it. He taught them to ask 'why' repeatedly to trace each error to its ultimate cause and devise a fix. At first the line stopped all the time and workers became discouraged. As teams gained experience, errors fell dramatically. Today in Toyota plants, where every worker can stop the line, yields approach 100 percent, so the line practically never stops. Mass-production plants, where only the line manager can stop it, still stop constantly, mostly over supply and coordination problems, and treat 90 percent yield as a sign of good management. Toyota assembly plants have practically no rework areas and do almost no rework.

• If a worker meets a problem they cannot fix, they stop the line and the team comes to work on it. Do not wait for a line manager's permission.
• If an error is fixed but its ultimate cause is not traced, expect it to recur. Ask 'why' at each layer until you reach the root, then change the process.
• If line stoppages are frequent and workers are discouraged early on, treat this as the expected learning phase. Keep the practice going and watch the error count for the fall.
• If your quality depends on end-of-line inspection and rework, treat that as a warning. Inspection cannot catch every defect in a complex product, so push detection back to the workstation.
• If people see no ladder of titles and no new challenges, supply a continuing variety of challenges. Otherwise they conclude they have reached a dead end and withhold their know-how and commitment.

• Giving the stop cord but not the team response. Workers who pull the cord and are met with blame or no help learn to stay quiet. The tell: pulls drop to zero while defects surface at the end of the line.
• Repairing symptoms and hoping they do not recur, the mass-production habit of treating problems as random events. The tell: the same error keeps returning and rework areas stay busy.
• Quitting during the early discouragement phase. Lines stop all the time at first and workers lose heart. The tell: the cord is removed or discouraged before teams have had the experience to drive errors down.
• Pushing responsibility down without managing the anxiety it brings. Responsibility gives freedom but also raises fear of making costly mistakes. The tell: workers hesitate to act on problems they could solve.
• Stripping slack without the creative challenge or job security. Removing slack without giving workers skills, control, and support produces mind-numbing stress rather than creative tension. The tell: people hold back know-how, and the main advantage of lean disappears.

• ▸Station and date: where the cord was pulled and when.
• ▸Problem observed: what the worker saw that they could not fix.
• ▸Who responded: the team members who came to the station.
• ▸Why #1 to #5: ask 'why' for each layer uncovered, writing each answer in turn.
• ▸Ultimate cause: the root cause the chain ends at.
• ▸Fix devised: the change that stops this error from ever occurring again.
• ▸Downstream check: whether defective units were built or passed along before the stop.
• ▸Recurrence log: whether the error has returned since, and the trend in error count over time.
• ▸Worker's next challenge: the new problem or skill assigned so the person keeps being challenged.

1. 1Pick three input-per-output ratios and measure them the same way at every site: labor hours per car, factory floor space per car (square feet), and inventory held on site. The passages' benchmark uses exactly these, so your numbers stay comparable.
2. 2Define the denominator before you collect anything. 'Hours per car' means total assembly hours divided by cars finished. Space means the whole factory footprint divided by annual output. Apply the same boundaries to every plant you compare.
3. 3Set your reference point from the best performer in a like-for-like group, not from your own history. The passages compare transplants against each other and against the same company's home plants, which shows what is achievable with the same product and a similar market.
4. 4Count the buffers. In mass production, extra supplies, extra workers, and extra space are added to protect expensive, disruption-intolerant machinery. List each buffer, then ask what disruption it insures against. Lean production removes the slack by removing the cause of the disruption, not by simply cutting the buffer.
5. 5Set the target at perfection, not at 'good enough': continually declining costs, zero defects, zero inventories. Treat any acceptable defect rate or maximum acceptable inventory level as a mass-production ceiling you have chosen to impose on yourself.
6. 6Staff for know-how, not nationality or cost. Put experienced managers who have run the system for years in charge of design, layout, and operations. The passages cite a senior executive's estimate that the system takes about ten years of practice under expert guidance to learn.
7. 7Give workers the skills and the authority to improve their own work, and back it with job security. The efficiency gain depends on workers using their know-how. If they hold it back, 'the main advantage of lean production disappears.'
8. 8Re-measure on a fixed schedule and compare any gap against the benchmark plant. When a gap persists, check management experience and layout first, before blaming the workforce.

The passages compare two Japanese-owned plants built in the United States. Company Y sent a large number of experienced managers from Japan to run its US plant, and got results 'exactly comparable' to its performance in Japan. Company Z delegated most operations, including design and layout, to Americans recruited away from Detroit. Its plant needed 23.4 hours per car and used more than 13 square feet of factory space per car, the least efficient use of space in the entire worldwide sample. Same national ownership, very different efficiency. The passages also note that Company Z trails Company Y in Japan too, so part of the gap is the company's own proficiency. The other part is that managers who do not fully understand and own the production system cannot introduce or sustain it in a new setting. Read it as a diagnostic: when hours per car and square feet per car are both far above the best peer, look at who designed the layout and how many years of lean experience they carry. Do not start with the workers or the machines.

• If a plant's hours per car and space per car are both well above the best like-for-like peer, then audit management experience and layout design before touching labor or equipment.
• If you are replicating a production system at a new site, then staff it with a large group of managers who have run the system for years. Do not delegate design and layout to outsiders unfamiliar with it.
• If an inventory, defect, or staffing buffer is justified as 'acceptable,' then trace it to the disruption it covers and attack that disruption. Do not just remove the buffer.
• If you propose cutting slack, then confirm workers have the skills, authority, and job security to solve problems. Otherwise you get stress without improvement.
• If you plan to expand to many sites quickly, then cap the pace at the number of experienced managers you can place. The passages say this supply constrains how fast foreign operations can be built.

• Equating a label with performance, for example assuming a Japanese-owned plant is lean by default. Tell: you have no hours-per-car or space-per-car number, only reputation. The passages show leaders and laggards inside the same nationality.
• Delegating the production system to local hires who do not fully understand it. Tell: the layout and operating practices differ from the parent's, and the plant trails the parent's performance by a wide margin.
• Treating lean as simple cost cutting that strips slack from workers. Tell: stress rises, but workers have no tools or authority to improve their work and no job security.
• Accepting 'good enough' targets, such as a standard defect rate or maximum inventory. Tell: targets are set as limits to stay under, not as numbers that keep falling toward zero.
• Expecting fast transfer. Tell: a rollout plan assumes a system can be learned in a few months, when the cited estimate is about ten years of practice under expert guidance.

• ▸Plant / site and product: name the plant and the product it makes, so comparisons stay like-for-like.
• ▸Total assembly labor hours per finished unit: total hours divided by units completed, using the same boundary at every site.
• ▸Factory floor space per unit (sq ft): total footprint divided by annual output.
• ▸Inventory on site: record what is held, so you can compare against the best peer and track the trend toward zero.
• ▸Best like-for-like benchmark: name the peer plant and its figures for hours and space.
• ▸Gap to benchmark (hours, space, inventory): subtract the benchmark from your actuals.
• ▸Buffers in place (supplies, workers, space): list each one and the disruption it protects against.
• ▸Management experience: count managers with years of lean practice and name who designed the layout.
• ▸Next action and review date: one change aimed at the largest gap, with a re-measure date.

1. 1Define the three measures before you touch the line: assembly defects found, rework hours or units required, and defects customers report soon after purchase. The passages treat defects as the outcome of the whole process, so track all three together rather than one alone.
2. 2Set the target at zero defects, not 'good enough.' The passages contrast mass producers, who accept 'an acceptable number of defects,' with lean producers, who aim at zero. State the goal in writing as perfection, even though no lean producer has reached it.
3. 3Make parts interchangeable by gauging them consistently. Use the same gauging system for every part through the entire manufacturing process (Ford's 'working-to-gauge'), because this is what cut assembly cost and made assembly simple.
4. 4Design for manufacture: reduce the number of parts and make each one easy to attach. Ford's four-cylinder engine block was a single complex casting, where competitors cast and bolted four cylinders together. Fewer joins mean fewer places to assemble wrongly.
5. 5Locate where defects are discovered. In mass production, poor work 'was not discovered until the end of the assembly line,' where rework specialists fixed it. Count how many people inspect and repair downstream. That headcount shows how much defect cost you are carrying.
6. 6Move detection back to the assembler. In Ford's system the assembler did not inspect for quality or report a malfunctioning tool. Lean practice uses multiskilled teams and pushes responsibility down the organization. Give workers the authority and the information to flag problems where they happen.
7. 7Test the product systematically before release. Craft production failed on quality because cars were effectively prototypes and owners did their own on-the-road testing. Run reliability and durability checks as a system, not as ad hoc tinkering.
8. 8Close the loop with early customer reports. Log defects customers report soon after purchase and trace each one to the station, part, or gauge that produced it.

Take Ford's shift from craft to mass production as the contrast. In the 1908 craft plant, one skilled fitter gathered parts, got tools, repaired them, did the complex fitting for the whole vehicle, and checked his own work. Parts were not reliably interchangeable, and each car was in effect a prototype, so consistency and reliability were elusive and owners did the road testing. Ford insisted on one gauging system for every part throughout manufacturing, solved warping with machine tools that worked prehardened metals, and cut part counts (the single-casting engine block). Fitting dropped out of assembly: the Model T owner could buy a spare and 'simply screw or bolt it on.' The limit of this system is that the assembler no longer inspected, and defects surfaced at the end of the line, where reworkmen fixed them. Quality came from interchangeability, but detection stayed late and indirect. The lean contrast supplies the next step: fewer defects through multiskilled teams and a zero-defect aim.

• If defects are found only at the end of the line by rework specialists, then you are running the mass-production pattern; move detection upstream to the people doing the work.
• If parts need fitting or adjustment during assembly, then fix interchangeability first: apply one gauging system across all stages before adding inspection.
• If an assembly step has many joined pieces, then ask whether a single piece or a simpler attachment removes the defect opportunity, as with Ford's one-piece engine block.
• If a team says an 'acceptable' defect rate is the limit, then reset the goal to zero defects and treat the gap as the improvement agenda.
• If customers report defects soon after purchase that your assembly counts never showed, then your testing is not systematic enough; add reliability and durability testing.

• Counting only direct assembly defects: the tell is low line counts alongside a large rework area and many inspectors, repairmen, and rework specialists.
• Treating the moving line as the quality driver: the tell is investment in line speed while parts still need fitting. The passages say interchangeability and simplicity of attachment made the line possible.
• Keeping workers silent: the tell is that nobody reports a malfunctioning tool or suggests process fixes because those jobs belong to the foreman or industrial engineer, so defects recur.
• Settling for 'good enough': the tell is a defect target expressed as an acceptable number, not zero.
• Relying on owners or chauffeurs as testers: the tell is that reliability and durability problems appear first in customer reports because no systematic testing occurred before shipment.

• ▸Product/line and period: name what you are measuring and the dates covered.
• ▸Assembly defects found: count defects detected during assembly, by station.
• ▸Rework required: record units or hours reworked and where the rework happens.
• ▸Customer-reported defects (soon after purchase): count and list the defects reported in the early ownership window.
• ▸Point of detection: note where each defect was first found (at the station, end of line, or by the customer).
• ▸Interchangeability check: state whether one gauging system covers every part from start to finish, and list exceptions.
• ▸Fitting or adjustment steps: list assembly steps needing fitting, and note whether a design change could remove them.
• ▸Worker reporting channel: describe how an assembler flags a defect or faulty tool and who acts on it.
• ▸Zero-defect gap and next action: state the gap to zero and the one change you will make before the next audit.

1. 1Fix the two measures before anything else. Total engineering hours is the sum of all engineering effort on the project. Total lead time in months runs from concept to market launch. The passages supply no benchmark figures for either, so your baseline has to come from your own projects.
2. 2Set the start and end events in writing. Start is the day a concept is formally authorized. End is the day the product reaches market. The IMVP framed the whole task as a chain: market assessment, product design, detailed engineering, supply-chain coordination, factory operation, and sales and service. Studies that looked only at the factory failed. Decide which links in that chain fall inside your count and apply the same boundary to every project.
3. 3Build the project record from primary sources. The IMVP did not rely on summaries. It recruited researchers who had worked in industry, such as Sheriff at Chrysler and Nobeoka at Mazda, and sent them into design shops for weeks or months to gather detail. Use timesheets, milestone dates, and interviews with the people who did the work.
4. 4Split hours by task type and by specialist group, and mark where work waits between groups. The source describes a lean-company development as a tightly knit team with clear leadership and many specialists contributing to a group effort. Hours or months lost between specialist groups show where the team is not working that way.
5. 5Separate all-new content from carried-over content. Sloan changed external appearance every year and added hang-on features (automatic transmissions, air conditioning, radios) to existing body designs, while the underlying mechanical components stayed in use across the product range for years. Tag each element of your product as new, modified, or reused, and log hours against each tag.
6. 6Run at least two or three comparable projects through the same record, ideally from different companies or divisions, as the IMVP's international comparisons did. One project tells you little. The comparison shows which gaps in hours and months come from process and which come from product content.
7. 7Test new-technology introduction as its own step. One IMVP researcher visited the major design and engineering centers to test the best ways of introducing new technologies, drawing on auto racing, where continuous technical leadership decides success. Record when each new technology entered the project and how many hours and months it added.
8. 8Feed results back to the team on a fixed cadence, since continuous feedback was one of the six elements the IMVP named as critical. Re-measure on each new project and compare it with the baseline.

Take Sloan's annual-change approach as the case, with the numbers left for you to collect because the passages give none. Sloan standardized mechanical components across the company's whole product range and produced them over many years with dedicated tools. Each year he altered the external appearance and added hang-on features that could be installed in existing body designs. Log one model year as an engineering-hours and lead-time record. Appearance changes and hang-on features go in the 'new' bucket. The shared mechanical components and existing body designs go in the 'reused' bucket. The hours and months in the 'new' bucket are the cost of keeping consumer interest alive. Now log a project that redesigns the body and the mechanical content together, using the same buckets and the same concept-to-launch boundary. The gap between the two records is your measure of how much reuse lowers effort and lead time. The passages also show the limit of this approach. Sloan's changes left the shop-floor model untouched, and the authors say work on the floor went from bad to much worse. A low-effort development record does not prove the wider system is healthy.

• If a project's hours or months can't be traced to a documented start and end event, then don't compare it with other projects. Fix the record first.
• If most hours sit in the 'new' bucket and little is reused, then check whether shared components could be carried across the product range, as Sloan did, before trying to cut effort elsewhere.
• If a delay shows up at the handoff between specialist groups, then the team structure is the problem. Look at whether leadership is clear and whether specialists share what they know, before adding engineers.
• If you are comparing your numbers with another company's, then confirm the other company's boundary covers the same chain, from concept to launch. Otherwise you are comparing different things.
• If a new technology adds visible months to a project, then record it separately and treat it as a distinct decision, since the source's racing example puts technical leadership at the center of introducing new technology.

• Counting only part of the chain, such as detailed engineering without design or market assessment. The tell: two projects with very different hours show the same lead time, or the figures leave out any task that the project actually performed.
• Using a single project as the benchmark. The tell: the team cites one number as 'our standard' with no second or third project to compare it against.
• Collecting figures from summaries and executive recollection rather than from people who did the work. The tell: dates are round, and timesheets and milestone records don't reconcile with them.
• Counting all hours as equal and ignoring reuse. The tell: a facelift and an all-new platform appear in the same average, so no one can say what drives the results.
• Treating a lower hours-and-months figure as proof the system works, as Sloan's innovations did for marketing while the production floor deteriorated. The tell: development speed improves while turnover, work rules, or quality problems get worse.

• ▸Project name and product type: one line naming the product and whether it is all-new, a modification, or an annual change.
• ▸Start event and date: the formal authorization of the concept, with the date.
• ▸End event and date: the market launch, with the date. Lead time in months is end minus start.
• ▸Scope boundary: list which links in the chain are included (market assessment, product design, detailed engineering, supply-chain coordination, factory operation, sales and service).
• ▸Total engineering hours: the sum of all engineering effort, taken from timesheets rather than estimates.
• ▸Hours by content tag: split hours into new, modified, and reused components.
• ▸Hours by specialist group: list each group's hours and note which groups handed work to which.
• ▸New technology introduced: name each technology, the date it entered, and the hours and months it added.
• ▸Leadership and team notes: one line on who led, whether the team worked as one unit, and where waiting occurred.
• ▸Comparison projects: list the other projects measured with the same boundary and what differs.

1. 1Split the construct into three gauges and report each separately: share (are you gaining and holding it), portfolio (how many distinct products, how current), and profit (profit and loss by unit). A single blended score hides which of the three is failing.
2. 2Build the Sloan reporting set for every profit center at a fixed, frequent interval: sales, market share, inventories, and profit and loss. Review capital budget requests against those numbers whenever a unit asks for central funds. The source describes this as managing 'by the numbers' from a small headquarters.
3. 3Count products the way the source counts them: by shared platform, not by badge. Vehicles sharing a wheelbase and many under-the-skin structural items are one product. Ford Taurus and Mercury Sable are one product, and so are the four GM-10 cars sold as Chevrolet, Pontiac, Oldsmobile, and Buick. Badge counts overstate your variety.
4. 4Compute average sales per product (total sales divided by platform count) for your company and each rival. This shows whether you are spreading volume thin or concentrating it. Compare like with like: use peak-volume years in cyclical markets, and when the current year is depressed, apply current market shares to a peak-year total, as the source does with 1986 volume for 1989.
5. 5Attribute sales by the region of the firm's headquarters, not the plant location. The source counts a Volkswagen Fox built in Brazil as European and a Ford Merkur built in Germany as American. This keeps share comparisons by producer clean.
6. 6Test the range against the whole market you intend to serve. Sloan's five-model ladder, from Chevrolet to Cadillac, was built to cover buyers of every income across their lives. For each segment and region, ask whether one standard product is serving it. Ford's Model T looked small to Americans, and the source notes that one standard product was not suited to all world markets.
7. 7Scan for niche attackers and rival-process advantage. Honda's aluminum-bodied NS-X attacked Ferrari's niche in ultra-high-performance sports cars. The source reasons that lean firms able to cut design and manufacturing costs and improve quality will threaten craft firms. Add a line to your review asking which rival is entering your niche and at what cost and quality.
8. 8Link the scorecard to the plant evidence. In the source's plant survey, the plants with the highest under-the-skin complexity also had the highest productivity and quality, and these were the Japanese plants in Japan. So a wide portfolio does not by itself mean poor profitability. Check whether profit shortfalls trace to plant performance rather than to variety.

Take General Motors in the mid-1920s as the source describes it. Alfred Sloan faced four problems and built a measurement system around them. For management, he split the company into decentralized divisions: five car divisions plus component divisions such as Delco, Saginaw, and Rochester. Each was a profit center reporting on sales, market share, inventories, and profit and loss, and each reviewed against capital budgets. If the numbers were poor, the general manager was replaced. If they were consistently good, the manager became a candidate for promotion to vice-president at headquarters. For portfolio, he set a five-model range running incrementally from cheap to expensive, Chevrolet to Cadillac, to fit buyers at every income level. For funding, he arranged stable outside sources through DuPont and the Morgan Bank. For foreign operations, he applied the same decentralized logic, so units in Germany and Britain ran as self-reliant companies managed by the numbers from Detroit. Read as a Corporate Performance scorecard, share and profit come from the divisional reports, portfolio breadth from the model ladder, and sustainability from the funding. The source also notes that complete vertical integration brought bureaucracy on a vast scale, so the lesson is the measurement discipline, not the structure around it.

• If two products share a wheelbase and major under-the-skin structure, then count them as one product and judge variety on platforms, not badges.
• If a division's numbers on share, profit, and inventories are poor over time, then change its leadership. If they are consistently good, then treat the leader as a promotion candidate, as Sloan did.
• If one standard product is sold across several regions, then test fit region by region before assuming scale savings. The Model T was seen as small in America, and the Escort was Ford's first attempt at global standardization.
• If a rival with lower design and manufacturing cost and better quality targets a niche you hold, then expect to adopt its methods or lose the niche. The source says craft producers face this choice.
• If sales are cyclical, then compare per-product volumes at peak years and rebuild current-year figures from peak totals. Otherwise a downturn will look like a portfolio problem.

• Counting badges as products. The tell: your 'variety' figure is far higher than your platform count, and sales per product look healthier than they are.
• Reading profit alone as performance. The tell: margins hold while share falls, or you cannot say how many platforms you offer or how dated they are.
• Comparing years in a cyclical market without adjustment. The tell: a trend conclusion changes when you swap a trough year for a peak year (the source notes 1989 sales were down 9 percent from 1986).
• Assuming one product fits every market. The tell: a unit abroad sells a design drawn for another region and loses share, as the Model T did in markets where it seemed small.
• Ignoring niche attackers because the niche is small. The tell: your competitor list names only the large mass producers, while a rival like Honda targets your niche with a new product.

• ▸Period and comparison years: state the year reviewed and any peak year used to normalize cyclical volume.
• ▸Market share by region: share of each region, with firms attributed by headquarters region.
• ▸Share trend and driver: gaining, holding, or losing, and the one cause you can name.
• ▸Platform count: number of distinct products grouped by shared wheelbase and under-the-skin structure, not by badge.
• ▸Average sales per product: total sales divided by platform count, beside the same figure for your two main rivals.
• ▸Portfolio freshness: introduction year of each platform, so you can see which are oldest (your own addition to the source's variety measure).
• ▸Segment coverage: the buyer segments served, from cheap to expensive, and any gap in the ladder.
• ▸Profit and loss by unit: profit and loss, inventories, and capital requests for each profit center.
• ▸Niche threats and response: rivals entering your niches, their cost and quality advantage, and your decision to match, ally, or exit.

1. 1Start from the goal, not the org chart. The source says lean producers aim at perfection: continually declining costs, zero defects, zero inventories, endless variety. Mass producers accept a 'maximum acceptable level' of defects and inventory. Write the team's targets as zero-defect and zero-excess, not as tolerable levels. Otherwise you are only re-labelling a mass-production crew.
2. 2Define the team's collective scope in one sentence: the set of operations it owns, plus quality control, simple maintenance and process improvement. The source says lean pushes responsibility 'far down the organizational ladder'. List each responsibility currently held by a supervisor, inspector or specialist. Move it to the team explicitly, one item at a time.
3. 3Replace sequential single-skill roles with shared competence. The Panhard & Levassor fitters each filed one part to fit the last, so errors accumulated as 'dimensional creep'. Map which tasks each member can do and which only one person can. Rotate people until every critical operation has several qualified members.
4. 4Staff and run the team the way the IMVP ran its own: a tightly knit group, clear leadership, and specialists willing to contribute personal knowledge to a group effort. Name one leader. Require specialists to share what they know rather than guard it.
5. 5Give people the freedom that comes with responsibility. The source calls it 'freedom to control one's work' and a big plus. It also warns of anxiety about costly mistakes. Pair each transferred responsibility with a visible rule: who decides, what the team can stop or fix without asking, and how mistakes are reviewed.
6. 6Build the career path before you need it. The source warns that better teamwork can mean a narrower portable specialty, and that a flat structure with no steep ladder feels disappointing. Plan a continuing variety of challenges per person: new skills, new operations, new improvement problems. Review each person's next challenge on a fixed schedule.
7. 7Check for withholding. If workers feel they have hit a dead end early, the source says they hold back know-how and commitment, and lean's main advantage disappears. Treat falling improvement suggestions or narrowing participation as the signal. Respond by adding challenge, not pressure.

The source's clearest team case is the IMVP research team itself, which ran as the kind of group a team-based work system needs. The directors knew a study of the whole auto industry could not be done by narrow specialists, so they built 'a worldwide team with no geographic center and no one nationality in the majority.' They recruited people with industry experience: Richard Lamming had been a Jaguar parts buyer and Toshihiro Nishiguchi had worked at Pioneer Electric. These people also took on rigorous research and years of field visits. The team had a tightly knit project structure, clear leadership, and specialists who contributed their knowledge to a group effort. It also built in independence by capping each sponsor at 5 percent of the $5 million, in a single account. Contrast the 1890s P&L assembly hall, where fitters each filed one part to fit the previous one, owned nothing beyond their own fit, and let errors compound into dimensional creep. The lesson for a production team is the same as the IMVP's: mix experience, give the group the whole task, and make members accountable to each other. The passages do not describe a plant team's internal routines, so the specific mechanics have to be built from these principles.

• If a responsibility (quality check, simple maintenance, improvement) still sits with someone outside the team, then the team is not yet collectively responsible. Transfer it or drop the claim.
• If only one member can perform a critical operation, then add cross-training before adding headcount or targets.
• If members start withholding ideas or effort, then suspect a dead-end career feeling. Add a new challenge or skill before changing incentives.
• If responsibility is pushed down but errors trigger blame, then expect stress and caution. Fix the review rules first.
• If a target allows 'acceptable' defects or inventory, then rewrite it toward zero. The team structure only pays off against a perfection goal.

• Layering teams over a mass-production hierarchy. The source says superimposing lean methods on mass-production systems causes pain and limited progress. Tell: teams exist, but supervisors and inspectors still make the decisions.
• Calling it a team but keeping narrow roles. Each person still does one task and hands off, as the fitters did. Tell: absences stop the line, and no one can cover another's operation.
• Ignoring the stress cost of responsibility. People fear costly mistakes and become cautious. Tell: workers escalate every decision and hide problems.
• Promising no career path. The flat structure feels like a dead end with no ladder or titles. Tell: skilled members stop sharing know-how and commitment drops.
• Setting tolerable-level targets. Tell: defect and inventory goals are stated as maximums rather than as steps toward zero.

• ▸Team scope: list the operations the team collectively owns.
• ▸Quality control: state what the team inspects and stops on its own.
• ▸Simple maintenance: list the tasks the team performs itself.
• ▸Process improvement: name the improvement the team owns this period and how it is tracked.
• ▸Skill matrix: for each operation, list the members qualified to do it, flagging any with only one.
• ▸Team leader and contribution rule: name the leader and say how specialists share their knowledge.
• ▸Perfection targets: state defect, inventory and cost goals as steps toward zero.
• ▸Mistake-review rule: define how errors are reviewed so responsibility does not become fear.
• ▸Next challenge per member: record each person's next new skill or task and the date it is reviewed.

1. 1Start by finding where defects are discovered today. In Ford's mass-production line, work done improperly 'was not discovered until the end of the assembly line,' where rework men with fitter skills repaired it. If your defects surface at the end, or at the customer, you have an inspect-and-repair system, not a built-in one. Mark the point of detection on a process map.
2. 2Remove the separation between doing the work and checking it. Ford's assembler 'didn't order parts, procure his tools, repair his equipment, inspect for quality,' and a separate inspector, repairman and rework specialist handled those jobs. Give each station's worker the duty to verify their own task and to reject a bad input from the station before.
3. 3Give every worker the authority to stop the line (andon). The Ford model assumed workers 'wouldn't volunteer any information on operating conditions — for example, that a tool was malfunctioning.' Reverse that assumption in writing: a stop is expected, and it triggers help, not blame. Remove the foreman's role as a 'semiskilled checker' who spots slacking and make it a responder role that fixes the problem.
4. 4Stop production at the first defect and correct it before the part moves on. Rework at the end of the line lets the same fault repeat for every unit built in between. Do not restart until the defect is contained and the part or process is fixed.
5. 5Trace the defect to its root cause with five whys. Ask 'why' of each answer until you reach a condition you can change, such as a gauge, a tool, a design or a method. Stop at a cause you can act on, not at a person. Craft production shows the pattern: parts don't fit, because they warped in the oven, because the tools couldn't cut hardened steel, and because contractors used different gauges.
6. 6Fix the cause at the source, not at the symptom. Ford did not hire better fitters. He insisted on the same gauging system for every part through the whole process, and he used machine tools that could work prehardened metal, so the filing step disappeared. Judge a countermeasure by whether it removes the need for the downstream correction.
7. 7Test systematically before the customer does. Craft production failed on reliability and durability because of 'the lack of systematic testing,' and owners and their chauffeurs did the on-the-road testing. Build tests into the process and record every stop and its root cause so you can see whether recurrence has actually fallen.

The passages give the cause chain behind Panhard & Levassor's quality problem. Symptom: the final assembly hall received parts whose 'specifications could best be described as approximate,' and two cars built to the same blueprints came out different. Why? Fitters filed each part to match the previous one, producing 'dimensional creep.' Why was filing needed? Parts warped in the hardening oven and needed more machining. Why did they warp? The 1890s machine tools could not cut hardened steel, so parts were cut soft and then hardened. Why did the parts not fit even before that? Contractors used different gauges and there was no standard gauging system. This chain ends in conditions that can be changed. Ford changed them by using one gauge system for every part from start to finish and by adopting machine tools that could work prehardened metal. That removed the filing and let him eliminate the skilled fitters, and task cycle times fell from 514 minutes to 2.3 minutes. The passages do not say P&L or Ford used 'five whys' by name, so treat this as a reconstruction. It shows the method: the correction targets the cause, not the part in front of you.

• If a defect is found anywhere past the station that caused it, then count it as a detection failure and move the check upstream to the cause, not just the repair downstream.
• If a worker sees a defect or a malfunctioning tool and the process offers no way to halt work, then add a stop mechanism before adding any inspection staff.
• If the 'why' chain ends at a person's carelessness, then you have stopped too early. Ask why the process allowed the error, and continue until you reach a gauge, tool, design or method.
• If a fix needs a skilled person to compensate each time (filing, rework), then it is a workaround. Fix the source, as Ford did with gauging, so the compensating skill becomes unnecessary.
• If the same defect recurs after a countermeasure, then the root cause was wrong or the fix was local. Reopen the why chain and check whether the same condition exists at other stations.

• Inspect-and-rework in disguise: you add inspectors and a rework area and call it quality. The tell is that defects are still found at the end and a rework group is still growing.
• Stop authority exists on paper only: nobody pulls the cord. The tell is zero stops, while defects still reach the end of the line. Workers who 'keep their head down' have not been given safety to speak.
• Whys that stop at blame: the answer to 'why' is 'operator error.' The tell is that the same defect returns under a different operator.
• Quick-fix restarts: the line restarts once the single part is repaired, without a root-cause record. The tell is repeated stops for the same reason.
• Quality left to the customer: reliability is learned in the field because internal testing is thin. The tell is that early use reveals faults, the pattern that burdened craft-built cars and their owners.

• ▸Date/time and station: where the line was stopped.
• ▸Who stopped it: name the worker, and record that the stop was expected.
• ▸Defect or abnormality: describe what was seen in one factual sentence.
• ▸Where it was caused vs. where it was found: note both stations, and the gap between them signals a detection failure.
• ▸Immediate containment: what was done to prevent the defect moving forward and who corrected it.
• ▸Why 1–5: write each 'why' answer on its own line until you reach a cause you can change.
• ▸Root cause (changeable condition): name the gauge, tool, design or method, not a person.
• ▸Countermeasure at the source: the change that removes the need for downstream correction, with owner and date.
• ▸Recurrence check: the date and result of confirming the defect has not returned, and whether other stations share the cause.

1. 1Start with the scope, not the factory floor. The source warns that many efforts to understand this industry failed because they never looked beyond the factory. Map the whole chain (design, supply, assembly, sales) and mark where parts wait. JIT only works if the stage that pulls and the stage that supplies share one logic.
2. 2Name the 'next stage' for each part. A pull system produces only what the downstream stage needs. For every part, write down who consumes it and what event tells them they need more. If you cannot name the consumer, that part is being pushed.
3. 3Replace the schedule-driven push with a signal. The definition calls for kanban-style signals from the consuming stage to the supplying stage. Define one signal per part flow, and make the rule that no signal means no production and no delivery. The source passages do not describe card mechanics, so take the sizing and format from your own plant data and Toyota-origin practice.
4. 4Smooth the load before you cut stock (heijunka). A pull system with lumpy demand passes the lumps upstream. Level the mix and volume you release to the line so downstream consumption is steady enough for the signals to work.
5. 5Remove inventory deliberately and expect it to expose problems. Cutting buffers is the mechanism, not the risk. When a stage stalls because a part did not arrive, log the cause (supplier, machine, quality, scheduling) and fix it at the source instead of restoring the buffer.
6. 6Extend the pull upstream to suppliers. The source treats supply-chain coordination as a separate element of lean production. Apply the same signal logic to purchased components, and expect to change how suppliers are organized and not only their delivery dates.
7. 7Learn it at the source and audit against the best. The IMVP trained its factory specialist (John Krafcik) at NUMMI and at Toyota plants in Toyota City, and it surveyed ninety assembly plants in fifteen countries. Benchmark your plant against world-class performance and use the gap to find the lagging practices.
8. 8Secure a mandate for the disruption. The source reports that layering lean methods onto a mass-production system causes great pain and dislocation, and that without a threat to survival only limited progress seems possible. Get explicit leadership backing before inventory is cut.

The source's clearest real example is General Motors. GM was, in the authors' words, the best at mass production, a system it helped create. In the age of lean production it found itself with too many managers, too many workers, and too many plants. Superimposing lean methods on the existing mass-production system caused great pain and dislocation. Because GM had not yet faced a life-or-death crisis, only limited progress seemed possible. For JIT, the lesson is that pull signals, low inventory, and leveled production are a system. Adding them to a plant built around large buffers and high-volume assumptions does not give the system, it gives disruption. The contrast the source offers is Toyota, where Eiji Toyoda and Taiichi Ohno pioneered the approach after World War II, and NUMMI, the Toyota-GM joint venture, where Krafcik learned the fundamentals at the source. The practical sequence is to benchmark against the best plants, learn the method where it is practiced, and agree up front that the problems exposed when stock falls are expected. The passages give no kanban quantities or lot sizes, so you must set those from your own data.

• If a stage can produce without a downstream signal, then it is pushing. Fix the signal before you touch inventory levels.
• If demand released to the line is uneven, then level the mix and volume (heijunka) before tightening pull. Otherwise the lumps travel upstream.
• If cutting a buffer halts a stage, then treat the halt as the finding. Trace and fix the root cause instead of restoring the buffer.
• If you are applying JIT only inside the factory walls, then widen the scope to suppliers and design. The source says factory-only views of this industry failed.
• If there is no crisis or mandate behind the change, then expect limited progress (the GM pattern). Secure leadership commitment first or pilot in a contained area.

• Cutting inventory without leveling the schedule. The tell is chronic starvation at some stations and pile-ups at others, followed by pressure to rebuild buffers.
• Treating JIT as a Japanese cultural trait instead of a transferable logic. The source says lean principles are universal and cultural features are of secondary importance. The tell is the claim 'it won't work here because of our workforce'.
• Layering pull signals onto a mass-production structure (too many managers, workers, plants) without changing the structure. The tell is dislocation and pain with no sustained gain, as at GM.
• Hiding the problems the system is meant to expose. The tell is expediting, workarounds, and unlogged stoppages, so root causes never get fixed.
• Limiting the effort to the assembly floor. The tell is that supplier deliveries and design changes keep breaking the flow even though in-plant signals work.

• ▸Part / flow: name the part and the stage that produces it.
• ▸Consuming stage: name the exact downstream stage that uses it. Write 'none' if unknown (a push flag).
• ▸Pull signal: state what tells the producer to make or deliver more, and who sends it.
• ▸Current buffer: record how much stock sits between the stages today, in your own unit.
• ▸Level of demand released: state whether the mix and volume released to the line are steady or lumpy, and the evidence.
• ▸Target buffer after reduction: set the new buffer you will test, with a review date.
• ▸Problem exposed: log each stoppage or shortage after the cut, with date and stage.
• ▸Root cause and owner: name the cause (supplier, machine, quality, scheduling) and the person fixing it.
• ▸Supplier extension: note whether the same signal reaches the supplier and what must change there.

1. 1Diagnose your current system first. Ask whether the relationship is run on power-based bargaining: drawings go out, suppliers bid on a set quantity, quality (max defects per 1,000) and delivery date, and the low bidder wins. If yes, you run a mass-production supply system, however much you say 'partnership.'
2. 2Drop the make-or-buy debate as the organizing question. The passages say the real question is how assembler and supplier work together to reduce costs and improve quality, whatever the legal relationship. Frame each component by that question, not by in-house versus outsourced.
3. 3Restructure into tiers. Give larger, more capable first-tier suppliers responsibility for engineering whole components, not just building to your drawings. Hand them design and engineering discretion you currently keep in a central staff.
4. 4Replace bid-and-switch with longer-term contracts and more frequent delivery intervals. Commit to the relationship so the supplier can plan, instead of guessing your next move.
5. 5Substitute an agreed rational structure for power bargaining: jointly analyze costs, jointly determine prices, and agree how profits are shared. Contracts are based on falling costs, so the supplier keeps part of the gains it produces.
6. 6Keep performance pressure on, openly. Compare suppliers against each other continuously and set much higher quality standards and much lower cost targets. Cooperation here is not a cozy atmosphere.
7. 7Shrink the purchasing function to match. Fewer, deeper relationships should need fewer buyers; the GM 6,000 versus Toyota 337 purchasing staff gap is the benchmark for whether you have truly changed.
8. 8Audit for 'new song, old tune' behavior. Check whether you cut across-the-board prices, rationalized supplier counts only for scale economies, or kept the power-based bargaining behind the partnership language.

The IMVP supply research, led by Richard Lamming (formerly a Jaguar parts buyer) and Toshihiro Nishiguchi (formerly at Pioneer Electric), compared Western and Japanese systems. At GM and Ford, central engineering designed most of a vehicle's 10,000-plus parts, gave drawings to suppliers, and awarded business to the low bidder, switching firms on short notice. Toyota's approach had suppliers take responsibility for designing and engineering their own products under cost-based contracts, with prices and profits set through joint cost analysis. The outcome is visible in purchasing headcount: in 1987 GM had 6,000 employees in parts purchasing against Toyota's 337. Lamming's interviews also showed Western firms mimicking only the surface. During the 1980s they cut suppliers from 2,000-2,500 to 1,000-1,500 and aimed for 350-500 per plant, but Chrysler in 1981 simply imposed across-the-board price cuts. That is rationalization under pressure, not lean partnership, because the power-based bargaining stayed intact.

• If your lever for cost reduction is an across-the-board price cut or a rebid, then you are running mass-production supply, not partnership, whatever you call it.
• If you reduce supplier count only to capture scale economies, then you have rationalized, not partnered; also transfer design and engineering responsibility for whole components to the remaining first-tier firms.
• If a supplier is asked to commit to falling costs, then agree up front how the resulting profits are shared, or the cooperation reverts to bargaining.
• If you cannot show joint cost analysis behind a price, then do not call the price 'agreed'; use the analysis to set it.
• If your purchasing staff stays large after you consolidate suppliers, then the relationship is still transactional; expect headcount to fall toward the lean benchmark.

• Treating trust as the mechanism. Tell: you talk about partnership and love between firms, yet the passages find no evidence Japanese suppliers love assemblers more than Western ones; the structure of incentives does the work, not goodwill.
• Supplier-count cuts without role change. Tell: fewer suppliers, but your central staff still designs the parts and hands out drawings for bids.
• Using single-sourcing as leverage. Tell: the sole supplier faces unilateral price cuts, as in Chrysler 1981, and guesses your next move.
• Equating cooperation with comfort. Tell: no ongoing comparison with other suppliers and no falling-cost targets, so performance pressure disappears.
• Keeping the old purchasing machine. Tell: a large buyer organization still negotiating each deal, like GM's 6,000, despite 'strategic' supplier language.

• ▸Component family and current supplier count: list the part group and how many suppliers serve it.
• ▸Selection method today: bid-on-drawings, low bidder, or joint analysis; state which.
• ▸Design/engineering owner: who designs it, your central staff or the supplier?
• ▸Contract length and delivery frequency: current terms versus the target of longer terms and more frequent deliveries.
• ▸Price-setting basis: unilateral cut, bid, or joint cost analysis; note the evidence.
• ▸Profit-sharing rule: how supplier gains from cost reduction are split; write the agreed rule or 'none'.
• ▸Performance pressure: comparison with other suppliers, quality standard (defects per 1,000) and cost-down target.
• ▸Purchasing staff assigned: current headcount against the intended level after change.
• ▸Reversion check: list any power-based moves used in the last 12 months (price cuts, short-notice switches).

1. 1Inventory what your new product shares before you design the team. List the models, mechanical components and factories it has in common with other products. The source's logic: Model A shares a transmission with Model B and is built alongside Model C in the same factory, so a fully isolated team would work at cross-purposes with the others.
2. 2Staff the team from every function the product touches. The source's list for a car is marketing, power train engineering, body engineering, chassis engineering, process engineering and factory operations. Name one person per function with real authority to commit that function, so collaboration is intensive and continuous rather than a series of handoffs.
3. 3Set up explicit dual reporting. Each person reports both to their functional department and to the development program. Write down which decisions the program leader owns (the product's content and timing) and which the functional head owns (technical standards, shared parts). The source calls managing this matrix the leadership challenge, so give one strong leader that job for the life of the project.
4. 4Keep engineers tied to their functional home. Isolated engineers lose contact with the technical frontier, and their designs stop being state-of-the-art. Schedule recurring contact between project engineers and their functional departments so they stay current on shared-component developments.
5. 5Bring first-tier suppliers in as an integral part of the development team, not as bidders on a finished drawing. Assign each a whole system, such as brakes, steering or electrical, that must work in harmony with the other systems.
6. 6Brief each supplier with a performance specification. State the performance target, the physical envelope, the unit price and the delivery point, and leave materials and design to the supplier. Ask for a prototype for testing. A working prototype earns the production order.
7. 7Let suppliers talk to each other about improving the design process. This is comfortable when each specializes in a different component and does not compete with the others in the group. Have each first-tier supplier organize its own second tier beneath it.
8. 8Set the calendar from the competitive cycle, not from habit. GM's executives concluded in the 1980s that they could wait no longer than 1986 for a new model, because running the usual ten-year cycle would leave them badly behind Ford and the Japanese. Work back from the date the market forces and size the team and the supplier hand-offs to meet it.

Toyota's brake assignment shows how a first-tier supplier is briefed. Toyota gave the supplier a performance specification: brakes that stop a 2,200-pound car from 60 mph in 200 feet, ten times in succession without fading. The brakes had to fit a space of 6" x 8" x 10" at the end of each axle and be delivered to the assembly plant for $40 a set. The supplier was then told to deliver a prototype for testing, and if the prototype worked it received a production order. Toyota did not specify what the brakes were made of or how they worked, because those were the supplier's engineering decisions. The brief fixes the interfaces (performance, space, cost) and leaves the design open. The supplier therefore works as a member of the product-development team, and the brakes still have to function in harmony with the steering and electrical systems.

• If the new product shares a transmission, components or a factory with other models, then do not build a fully self-contained team. Use a matrix with dual reporting and a strong program leader.
• If a component is a whole system (brakes, steering, electrical), then assign it to a first-tier supplier with a performance spec covering performance, envelope, price and delivery, and let the supplier choose materials and design.
• If a supplier's prototype passes testing, then issue the production order. If it fails, the order is not guaranteed, so the prototype gate stays real.
• If suppliers in the group do not compete on the same component, then encourage them to share design-process improvements. If they do compete, do not expect candid sharing.
• If project engineers are drifting from their functional department, then restore contact, because isolated engineers lose the technical frontier and the design stops being state-of-the-art.

• Building a fully isolated team for each product. The tell: its transmission or factory plans collide with other teams' plans, and the design is not state-of-the-art because engineers lost contact with their specialty.
• Running the matrix without a leader who owns it. The tell: people satisfy their functional department or the program but not both, and the same conflicts return to meetings unresolved.
• Handing suppliers a finished drawing instead of a performance spec. The tell: the spec dictates materials and mechanisms, the supplier has no engineering decisions to make, and design improvement stays with you.
• Treating suppliers as outside the project, brought in after design. The tell: component systems arrive that do not work in harmony with each other, and the problems surface late.
• Letting the cycle drift by default. The tell: you are running the old ten-year cycle on one model line while competitors replace theirs sooner, as GM feared in the 1980s.

• ▸Product and target date: Name the product and the date the market forces, and note what competitors' cycles imply.
• ▸Shared resources: List the models, components and factories shared with other products, and who owns each.
• ▸Functional members: Name one empowered person for each function (marketing, power train, body, chassis, process, factory operations).
• ▸Dual reporting rules: State which decisions the program leader owns and which the functional head owns.
• ▸Program leader: Name the single leader accountable for managing the matrix for the project's full duration.
• ▸Supplier systems: List each system assigned to a first-tier supplier and the supplier's name.
• ▸Performance spec per system: Record the performance target, physical envelope, unit price and delivery point. Leave materials and design to the supplier.
• ▸Prototype gate: Set the test, the date and the condition for issuing the production order.
• ▸Functional-frontier link: Say how and how often project engineers meet their functional departments to stay current.