Lean Thinking

1. 1State the principle you are building to: everyone involved must be able to see and understand every aspect of the operation and its status at all times. Test each display against it. If a worker cannot read it from their station, it does not count.
2. 2Start with the workplace. Apply 5S: remove all debris and unnecessary items, and give every tool a clearly marked storage place visible from the work area. A missing tool then shows up as a gap.
3. 3Post the pace. Calculate takt time against current demand and post the production slots it creates (for example, ten per hour at a six-minute takt, or one per minute at a sixty-second takt). Use a simple whiteboard in the team area, and add andon boards where electronic display is warranted.
4. 4Post the work method. Display clearly posted, up-to-date standard work charts, written by the work team rather than a remote industrial engineering group.
5. 5Add status indicators and a progress board for work done out of sight of others. Divide the day into fixed cycles, show cycles to complete against time available, and have each person mark a completed cycle with a colored magnetic marker.
6. 6Make errors visible at the point of occurrence with poka-yoke. For example, put photo cells across each parts bin opening so a light flashes if a product moves on with a part left out. No defective part goes ahead unnoticed.
7. 7Display key measurables and the financial information on the cost of the process next to the operation. Let the team that does the work read the results and respond without waiting for a supervisor.
8. 8Extend the display across firm boundaries. Transmit takt and production status electronically to supplier and customer facilities, so each participant can verify that the others follow the agreed principles along the value stream.

Toyota's parts distribution centers (PDCs) show the method. After shrinking aisles and moving the most frequently demanded parts closest to the start of the picking runs, the team introduced standard work and visual control at the end of 1990. The workday was divided into twelve-minute cycles, the best compromise between walking distance and cart size. In one cycle an associate picks thirty lines of small parts, twenty of medium parts or twelve of large parts. A progress control board between the receiving and shipping docks showed the number of cycles to complete and the time available. Each associate had magnetic markers of a given color and placed one in the right square after every completed cycle. In a warehouse where everyone works out of contact with everyone else, the whole team could now see exactly how the work was going. The board removed the need for team leaders to 'supervise' their teams, because everyone could look at the board and judge progress themselves.

• If workers cannot see the status of their own work from their station, then add a display (whiteboard, andon board, or progress board) before adding supervision.
• If demand changes, then recalculate takt time and repost the slots at once. The display should tell the team whether they need more orders or need to remove waste to hit a shorter takt.
• If a defect could pass to the next step unnoticed, then build in a poka-yoke that makes it impossible for even one defective part to go ahead, rather than relying on inspection.
• If a tool, such as a large paint booth, must run in batch mode, then still make buffers and changeovers visible. Cut changeover times and batch sizes to the minimum the machinery allows.
• If work crosses firm boundaries, then transmit the same status display to suppliers and customers, so each party can verify the others' behavior.

• Displays that are out of date. Standard work charts or boards that are not current are ignored. The tell: workers follow a method that differs from the posted one.
• Standards and charts written by a remote industrial engineering group instead of the work team. The tell: the team treats them as someone else's paperwork.
• Transparency stopping at the plant wall. The tell: suppliers and customers find out about a problem only after the product is late.
• Financial transparency without a way to improve. Open-book numbers get posted, but no flow or pull techniques exist to improve them. The tell: the only response is sweat and longer hours.
• Using visual control as a supervisor's reporting tool instead of the team's self-management tool. The tell: leaders still walk around to 'supervise' work the board already shows.

• ▸Operation / area: name the workstation, cell, or value-stream step being audited.
• ▸Takt time and posted slots: state the current takt time and where the production slots are displayed.
• ▸5S status: list tools or items without a clearly marked, visible storage place.
• ▸Status indicator in use: andon board, progress board, or whiteboard, and where it is located.
• ▸Standard work chart: posted, current, and written by the work team (yes/no, date last updated).
• ▸Poka-yoke in place: describe the mechanism that stops a defective or incomplete part from moving on, or write 'none' and the gap.
• ▸Key measurables and process cost displayed: list the measures posted and who can read them.
• ▸Who can see it: employees, suppliers, customers (mark each yes/no and the means of transmission).
• ▸Action and owner: the one display to add or fix, who owns it, and the due date.

1. 1Pick one product or product family and fix the unit of analysis there. The lean enterprise follows a specific product's value stream, not a firm, a department, or a product line in aggregate. The alliance lasts as long as the product does: decades for a car size class, possibly under a year for application software.
2. 2Map the three streams for that product: the problem-solving task (concept to detailed design to launch), the information task (order-taking to scheduling to delivery), and the physical transformation task (raw material to customer). Put every independent firm touching any of the three on the map, including upstream firms you rarely meet.
3. 3Test whether anyone can narrate the flow. Ask each firm to put itself in the position of a design, an order, and the physical product, and describe what happens to it at each step. An awkward silence marks the point where the firm sees only its own piece.
4. 4Convene all concerned parties as a continuing conference. It is a voluntary alliance, not a new legal entity and not a form of vertical integration. Each firm remains independent and the stream stays disintegrated.
5. 5Agree on a short set of principles to regulate firm-to-firm behavior, and make the steps transparent. Each firm explains its activities to the others so every participant can verify that the others follow the agreed principles. Address the confidentiality fear directly: firms hold back because they expect others to use revealed information to drive a harder bargain.
6. 6Examine every value-creating step across firm boundaries and sort each into three types: steps that create value, steps that create no value but are unavoidable with current technology and assets (Type One muda), and steps that are immediately avoidable (Type Two muda). Cut the Type Two steps first.
7. 7Define value for the whole product, jointly with the end customer. Where each firm defines value to suit itself, the definitions do not add up. Rebuild the specification around what the customer receives from the full stream.
8. 8Carry the work from the initial radical realignment (kaikaku) into continuous improvement (kaizen). Re-map as the product moves through generations, and keep the alliance running for the product's life.

Pratt & Whitney mapped the value streams for its three families of jet engines and found waste in the tributary stream of four firms: the melter, the forger, the machiner, and the final assembler. None of the four had ever fully explained its activities to the other three. Part of the reason was confidentiality, since each feared that upstream and downstream firms would use any information revealed to drive a harder bargain. Part was obliviousness, since each looked carefully at its own affairs and never at the consequences of its internal activities for the others. Once they looked at the whole stream, they discovered massive waste. For example, ingots were being prepared in massive size because the melters were certain that was efficient, without attention to the shape of the finished parts. The melters were also preparing several different ingots, at great cost, to meet Pratt's precise requirements for each engine, which varied only marginally from other engine families and from competitors' needs. Many of these activities could be eliminated almost immediately with dramatic cost savings. The waste had gone unnoticed for decades in a supposedly sophisticated industry because no one had examined the stream across firm boundaries.

• If a proposed improvement stops at your firm's boundary, then extend the map to the suppliers and customers on both sides before acting, because major breakthroughs come from the whole value stream.
• If partners resist sharing process detail, then set transparency rules and shared principles first, and surface the fear of a harder bargain explicitly. Without verification no one can confirm that others follow the principles.
• If someone proposes a new legal entity or a merger to coordinate the stream, then reject it and keep a voluntary alliance of independent firms. A formal entity becomes vertical integration, the opposite of what is needed.
• If a step creates no value, then classify it as Type One (unavoidable with current technology and assets) or Type Two (immediately avoidable), and remove the Type Two steps first.
• If the firms' definitions of value do not add up to what the customer receives, then redefine value for the whole product with the customer before optimizing any single firm's step.

• Treating the lean enterprise as a legal structure or virtual corporation. The tell: talk of a new entity, ownership stakes, or consolidation rather than a voluntary alliance of independent firms.
• Mapping only your own firm's portion of the stream. The tell: the map starts at your receiving dock and ends at your shipping dock, and no one can say what happens to the product upstream or downstream.
• Using the program to squeeze partners, a "mean" version of lean that diverts revenue from downstream customers and extracts profit from upstream suppliers. The tell: partners withhold information because they expect it to be used against them in the next negotiation.
• Skipping the shared principles and transparency. The tell: participants cannot verify others' behavior, so cooperation depends on trust alone and lapses once the sponsoring effort fades.
• Letting each firm define value its own way. The tell: every firm reports success on its own metrics while the customer still receives a fragmented product, like the Crete holiday pieced together from nineteen separate parts.

• ▸Product or product family: name the single product whose stream the alliance will manage.
• ▸Expected product life: state how long the alliance must last (months, years, decades).
• ▸Firms on the stream: list every independent firm from raw material to end customer, noting which of the three tasks each touches.
• ▸Three-stream map owner: name who maintains the map of concept-to-launch, order-to-delivery, and raw-material-to-customer.
• ▸Shared principles: write the short set of rules governing behavior between firms.
• ▸Transparency commitments: list what each firm will disclose about its steps and how others verify it.
• ▸Value definition with the customer: state the whole-product value as the end customer would define it.
• ▸Muda register: list each non-value step with its firm, Type One or Type Two, and the action and date for removal.
• ▸Kaikaku and kaizen plan: record the first radical realignment and the cadence of follow-on improvement reviews.

1. 1Define muda against the customer, not the department: any human activity that absorbs resources but creates no value. Test each step with one question: would the customer pay for this, or would they miss it if it vanished?
2. 2Map the whole value stream for one specific product, from concept through design, from order entry through production scheduling to delivery, and from raw materials to the customer's hands. Do not stop at your firm's walls. Pratt & Whitney's waste sat between firms.
3. 3Sort every step into three categories: steps that create value, steps that add no value but are currently unavoidable, and steps that add no value and can be eliminated immediately because they are redundant. Remove the third category first.
4. 4Name each non-value step by type: defects and rework, overproduction, unneeded processing, purposeless movement or transport, waiting downstream for late upstream work, excess inventory, and goods or services that miss the customer's need. Naming the type points you at the cause.
5. 5Share the map with upstream and downstream partners. Waste hidden by confidentiality or obliviousness survives for decades, so make the steps transparent to everyone along the stream.
6. 6Remove obvious waste, then make the remaining value-creating steps flow. Replace batch-and-queue by department with continuous flow, and let customers pull. Flow exposes hidden muda, and pull exposes the impediments to flow.
7. 7Attack the muda in short, repeated kaizen events on the same activity rather than one exhaustive planning exercise. Revisit the activity after each round, because each round reveals more waste.
8. 8Set numerical targets with a deadline and the resources to meet them, for example inventory down 25 percent, escaping defects down 50 percent, and effort per unit down 20 percent in the first year. Measure against perfection, meaning zero muda, not against competitors.

Pratt & Whitney's tributary value stream for a jet engine ran through four firms: the melter, the forger, the machiner, and the final assembler. Once mapped end to end, it showed that the initial ingot weighed ten times the finished part, so 90 percent of very expensive titanium or nickel was scrapped. The forger duplicated at great cost metals work that the next firm downstream already did. The melters also prepared several different ingots to meet Pratt's slightly varying requirements for each engine family. The waste went unseen for decades because none of the four firms had fully explained its activities to the other three, partly from confidentiality fears and partly from obliviousness. Looking at the whole stream exposed massive waste, and many of these activities could be eliminated almost immediately with dramatic cost savings. A second Pratt case shows the same logic on the shop floor. A totally automated turbine-blade grinding system was replaced by an in-house U-shaped cell at a quarter of the capital cost. Production costs halved, throughput time fell 99 percent, and changeover dropped from hours to seconds.

• If a step is redundant and adds no value to the customer, then eliminate it immediately, before any flow redesign.
• If a step adds no value but cannot yet be removed with current technology or assets, then flag it as a target for lean techniques such as flow, pull, and kaizen rather than leaving it as accepted.
• If the waste appears to sit at a boundary between firms or departments, then widen the map to the full value stream and open the steps to the other parties.
• If you are tempted to benchmark against competitors, then compare against perfection instead. Being ahead invites complacency, and being behind distracts you with scale and cost-factor excuses.
• If a team is tempted to plan the perfect process before acting, or to stop improving once an activity is fixed, then run repeated short kaizens on the same activity, since gains kept compounding in the Ligonier case.

• Optimizing within the firm only. The tell: each company can describe its own operation but nobody can say what happens to the product after it leaves, so cost is duplicated downstream.
• Treating batching as efficiency. The tell: people and machines are all busy and utilization looks high, but the product spends long periods sitting while departments change over.
• Declaring the process fixed after one improvement round and moving to steady-state management of variances. The tell: gains stop after the first kaizen, although repeated events on the same part kept lifting productivity.
• Spending on automation instead of removing waste. The tell: expensive, fully automated systems with long changeovers, when a simple in-house cell could cut cost and time.
• Setting vague goals with no numbers or resources. The tell: no policy matrix, no deadline, and nobody agreeing that the targets are actually doable.

• ▸Product and value stream boundary: name the one specific product and the start and end points, from concept or raw material to customer.
• ▸Step description: record each activity in sequence, including those performed by other firms.
• ▸Category (1/2/3): mark value-creating, non-value but currently unavoidable, or non-value and immediately eliminable.
• ▸Muda type: defect/rework, overproduction, unneeded processing, motion or transport, waiting, inventory, or goods not meeting customer need.
• ▸Evidence of waste: state what is observed, such as scrap percentage, queue time, or duplicated work.
• ▸Owner and who else sees this step: list the responsible party and the up/downstream partners who have been shown it.
• ▸Action: eliminate now, or send to a kaizen event, flow redesign, or pull system.
• ▸Numerical target and deadline: state the improvement figure, the date, and the resources committed.
• ▸Re-check date: schedule the next revisit of this activity to find the muda exposed by the last round.

1. 1Audit the work against Csikszentmihalyi's five conditions before changing anything. For one role, score each as present or absent: clear objective, concentration that absorbs all attention, no interruptions or distractions, clear and immediate feedback on progress, and challenge where skills are adequate but just adequate. Batch-and-queue work typically fails all five: the worker sees a small part of the task, gets no feedback, uses little skill, and is constantly interrupted.
2. 2Make the whole flow visible to the person doing the work. Lantech's cell workers can see the entire flow from raw materials to completed machine, and the order office uses visual control so everyone knows where the company stands. Post the system status where the work happens, so each person knows how their piece fits.
3. 3Build immediate feedback into the task sequence. Use takt time, standard work, and posted work charts for every task so people know at once whether work is on pace. Group related steps so errors surface right away. Toyota's Miyata plant installs and tests the electrical system in one focused area, so the team learns immediately whether everything was done correctly.
4. 4Remove interruptions at the source: line stoppages, sudden demands to switch tasks to handle a crisis, and conflict over which task comes next. Let one employee complete a whole task where possible. Lantech's new order entry system lets one person do the entire job and get immediate results.
5. 5Match challenge to skill through multiskilling and job rotation, so each worker's full skill set is used. Treat the continuing elimination of muda as the challenge itself: keeping the system flowing with no interruptions is hard, and a team with the skills and a way of thinking can meet it.
6. 6Give workers a role in redesigning the work. Run frequent kaizen events so people take part in work design, and so thinking and doing, planning and acting, are joined, as at Lantech in both plant and office.
7. 7Measure actual human effort, not just whether a job fits the cycle time. Following Toyota's Motomachi plant, have work teams determine the fatigue and stress of each motion, total it per job, and kaizen any job with unacceptable levels, adding simple operator assist mechanisms. This keeps challenge from becoming overload.
8. 8Plan for the cost of constant change. Say in advance whether excess workers will be retained, how improvement contributions will be recognized and rewarded, and what improvement means for careers. Lantech's workers asked exactly these questions.

Lantech is the source's fullest case. Before the change, development work was cumbersome and most projects were never completed because market conditions changed first. After the change, workers in the manufacturing cells see the whole flow from raw materials to finished machine. Takt time, standard work, and posted work charts show immediately how work is going. Multiskilling and rotation use each worker's skills, and kaizen events let them help design the work. Line stoppages and sudden task switches are few. In the office, visual control in order-taking shows where the company stands, and one employee can handle a whole order for immediate results. Development staff work in the same space and projects move rapidly to completion, which gives real feedback. Bob Underwood says: "We were living in darkness and now we have come into the light." The cost shows up in the line "We just get something working smoothly when it's time to improve it again," and in worries about job security, recognition, and career direction. The source calls Lantech no paradise.

• If a worker sees only a small part of the task and gets no feedback, then fix visibility and feedback first (visual control, charts, whole-task ownership) before adding incentives.
• If interruptions come from crises, stoppages, or competing task demands, then remove their source instead of asking people to concentrate harder.
• If a job meets the cycle time but fatigue or stress is unmeasured, then have the team quantify effort per motion and kaizen the high-stress jobs before treating the job as sustainable.
• If you add automation, then check that the plant can run at 100 percent output and that support costs are below the labor value saved. Toyota's Tahara plant failed both tests, and Miyata returned to lower automation with focused test areas that give immediate feedback.
• If improvement activity begins to cause anxiety, then answer the questions on job retention, recognition, and careers openly before launching the next round of change.

• Treating flow as a perk or morale program while leaving batch-and-queue structure intact. Tell: workers still see only a fragment of the task, get no feedback, and are interrupted constantly.
• Relying on financial rewards alone for feedback, as in open-book management, with no know-how to improve. Tell: the answer to better performance is sweat and longer hours. The source argues that immediate feedback from flowing work makes much of the carrot-and-stick apparatus unnecessary.
• Judging a job only by whether it fits the cycle time. Tell: critics say the pace is impossible, and no one can state the effort level objectively.
• Assuming flow work is a paradise and ignoring its costs. Tell: people say they just get things smooth when it is time to change again, and ask whether they are going anywhere or just flying a holding pattern.
• Over-automating until the technology, not the people, sets the work. Tell: high indirect technical support costs and a plant that cannot run at 100 percent output, as at Tahara.

• ▸Role/work area: name the specific job or cell being assessed.
• ▸Clear objective: state the objective the worker can name and whether it is visible at the workstation.
• ▸Whole-system visibility: record what share of the flow the worker can see and which visual controls exist.
• ▸Feedback delay: record how long after an action the worker learns whether it was done right.
• ▸Interruption log: list each interruption (stoppage, task switch, crisis, priority conflict) with its source.
• ▸Skill-challenge match: note which skills are unused, whether tasks are too easy or overwhelming, and what rotation or multiskilling is planned.
• ▸Effort/fatigue measure: record the stress and fatigue per motion and per job, and any kaizen or operator assist planned.
• ▸Worker role in redesign: record the next kaizen event and who participates.
• ▸Change-anxiety response: list the open questions on job retention, recognition, and career, each with an owner and a commitment.

1. 1Start with value, not the department. Name one specific product (good, service, or both), the customer, the price, and the time. Value is defined only by the ultimate customer. Providing the wrong good or service the right way is muda.
2. 2Map the whole value stream for that product across its three management tasks: problem-solving (concept to launch), information management (order-taking to delivery), and physical transformation (raw material to customer). Walk it in person. Ask: can you describe what happens to a design, an order, and the physical product at each step?
3. 3Take the problem-solving learning off the meeting calendar and into work. Replace weekly one- or two-hour improvement meetings that plan changes for weeks later with short training followed immediately by a hands-on kaizen exercise. Byrne used two-day lean-principles sessions for 150 employees, followed at once by three-day kaizen exercises.
4. 4Make the leader do the first round. The leader goes to the shop floor, the engineering area, and the order-taking and scheduling departments to see the waste first-hand. Byrne's rationale: a CEO who delegates never learns about change where value is created and goes on managing by the numbers.
5. 5Run a 'walk of shame.' Take managers and the union head through every part of the plant, engineering, and sales. Make the muda visible to them directly, then state the commitment: every process, including product development and order-taking, converts to continuous flow and pull.
6. 6Alternate kaikaku and kaizen on every stream. Use kaikaku (radical improvement) to rearrange activities from departments and batches into flow quickly, even in a day. Follow with repeated kaizen (continuous incremental improvement). Apply the same frequency to production, product development, and order-taking, with no exemption for office processes.
7. 7Build the team around the whole task. In product development, put the production/tool engineer on the original definition team. Set a target cost (estimated market price minus acceptable margin). Have the team do detailed part and tool design to that target. Move the team's desks to the factory floor at the end for process-at-a-glance and standard work with the production team.
8. 8Extend the learning beyond the firm's boundary. Once your own stream is understood, bring upstream and downstream parties into a continuing conference (the lean enterprise) so each can see the consequences of its activities for the others. Make the steps transparent so each party can verify the agreed principles are being followed.

Wiremold under Art Byrne. Earlier, Wiremold's TQM improvement teams met weekly for an hour or two, mostly planning changes to implement weeks or months later. Byrne replaced this. He wrote a training manual and personally ran two-day lean-principles sessions for 150 employees, each followed immediately by three-day kaizen exercises so people could use the skills right away. He then took managers and the union head on the 'walk of shame' through the plant, engineering, and sales: 'There was muda everywhere and my managers were now able to see it.' Hundreds of weeklong kaizen activities followed, and order-taking got the same kaikaku and kaizen as production. The new product-development team worked to a target cost, included the production/tool engineer from the start, and moved to the factory floor at the end. Wiremold's first product under this regime took six months, and tool costs came in at 60 percent of the original budget based on past experience.

• If improvement activity is delegated and the leader has never been on the floor, then the leader runs the first round personally, because otherwise they keep managing by the numbers and kill the effort.
• If a training session ends and application is scheduled weeks later, then collapse the gap: put a kaizen exercise immediately after the training so skills are used while fresh.
• If a problem's cause sits outside your department or firm, then widen the map to the whole value stream and bring in the upstream and downstream parties. Don't stop at your own boundary.
• If a process is incremental-resistant because of batches and departments, then use kaikaku to rearrange it into flow first, and kaizen afterward to refine it.
• If employees are treated as the problem (outside SWAT teams, job cuts), then stop. Redefine the work so it is in their interest to make value flow, or morale collapses and the organization regresses to the mean once the outside team leaves.

• Learning that never touches work: weekly meetings that only plan changes for later. Tell: improvements are listed in minutes but nothing physically changes within the week.
• Delegated change: the CEO sponsors but doesn't go to the floor. Tell: managers keep running by the numbers and improvement activities quietly die.
• Fixing within the firm's walls or by process fragments: reengineering order-taking in aggregate or stopping at your own boundary. Tell: local gains appear but waste along the whole value stream remains.
• Outside experts imposing change on employees: SWAT-team reengineering. Tell: morale collapses among survivors and performance drifts back to the mean after the consultants leave.
• Applying improvement to production only. Tell: product development and order-taking still run in batch and queue (e.g., a nightly batch order run and multi-day shipping lag), with no kaizen cadence of their own.

• ▸Product and customer: name the specific product, the customer, price, and time that define value.
• ▸Stream task in scope: problem-solving (concept to launch), information management (order to delivery), or physical transformation (raw material to customer).
• ▸Observed muda: list the waste seen first-hand on the walk, with location and step.
• ▸Root cause: state why each waste exists (batch, department boundary, forecast-based schedule, other firm's behavior).
• ▸Countermeasure: specify the concrete rearrangement or change to be made this week.
• ▸Kaikaku or kaizen: mark which type this is and the date it will be carried out (not planned for later).
• ▸Participants and leader: list the team, including who from production/tool engineering, and the leader who is present on the floor.
• ▸Result measured: record before/after on cost, time, errors, or scrap.
• ▸Next stream/neighbor to involve: name the upstream or downstream party or the next process to attack, and the date.

1. 1Define the product from the customer's side first. Walk one product the way the customer experiences it, from raw material to delivery, and write a box score: total elapsed time, time actually adding value, queue/wait time, number of lines or queues, number of handoffs, number of repeated inspections, total steps. The travel example scores 13 hours total, 7 hours (54 percent) moving, 6 hours waiting, 10 lines, 7 luggage pick-ups, 8 identical inspections, 23 steps.
2. 2Set baseline numbers for every metric you will claim: labor productivity (direct, managerial, technical), production throughput time, inventory in the system, errors reaching the customer, in-process scrap, job-related injuries, and space and tooling used. Verify them by walking the floor yourself, not from reports.
3. 3Set targets from the rules of thumb. Moving a classic batch-and-queue system to continuous flow with customer pull should double labor productivity, cut throughput time 90 percent, cut inventory 90 percent, and halve customer-reaching errors, scrap, and injuries. Treat these as the first-leap benchmark, not a ceiling.
4. 4Rearrange the activities for one specific product into continuous flow instead of departments and batches. Do it as a kaikaku event, which can take a day. Achieve it with quick tool changeovers and right-sized (smaller) machines so different process types (molding, painting, assembly) sit adjacent, with the object kept in continuous flow, usually without an assembly line.
5. 5Let the customer's pull set what is made. Make only the parts actually needed, and stop the line or machine automatically when a mistake occurs so bad parts never pass downstream (jidoka plus Just-in-Time).
6. 6Re-measure the same metrics against baseline immediately after the change, and record freed space, tooling, and equipment. Capital needed should be modest or even negative if equipment can be sold.
7. 7Build the support structure before you leave. Assign named line managers to maintain the new cells, provide day-to-day coaching, and close the loose ends from each event. Without this, blitz gains erode as people return to old ways.
8. 8Switch to kaizen after the radical realignment. Aim to double productivity again and halve inventories, errors, and lead times again within two to three years. Make results visible to everyone so employees get immediate feedback on their improvements.

At Pratt & Whitney, a lean sensei-led improvement effort consolidated a value stream and cut the effort, space, and tooling needed by 75 percent. Mark Coran said the sensei's central contribution was to change permanently the sense of what was possible and in what time frame. The gains did not hold. Bob D'Amore's Continuous Improvement Office had neither the resources nor the authority to follow up on loose ends after each exercise or to coach line managers on maintaining the new compact cells. Many managers actively resisted, and the spectacular gains from the one-week blitzes were lost as managers and workers went back to old ways. Demand was also collapsing, with annualized shop hours falling from a peak of 11 million (June 1991 to July 1992) to 8.8 million by December 1992, so even real gains could not offset the loss of volume. The lesson: the 75 percent result proves the size of the available gain, but you have to pair it with a support structure and with a business sized to demand.

• If a metric's baseline has not been verified by direct observation, do not report an improvement against it. Measure first, then change.
• If the product is made in departments and batches with long queues, go for kaikaku (rearrange to flow in days) before kaizen. Incremental tweaks to a batch-and-queue layout will not deliver the doubling.
• If a gain depends on a central improvement team returning to fix loose ends, then it will not hold. Give line managers ownership and coaching before you declare the gain.
• If accountants or department heads argue for keeping expensive assets fully utilized, then test flow on one product first and show the throughput, inventory, and space results. Utilization logic is the main reason flow is resisted.
• If volume is falling while you improve, then resize the plant and headcount and the business structure as well. Operational gains alone will not restore profitability.

• Optimizing each firm or department's own efficiency while nobody looks at the whole product through the customer's eyes. The tell: every step is locally efficient, yet the box score shows large wait time, many queues, and repeated identical inspections.
• Treating the one-week blitz as the finish. The tell: impressive results at the end of the event, then metrics drift back within weeks because no one owns follow-up or coaching.
• Managerial resistance left unaddressed. The tell: old batch practices reappear in the new cells and managers describe the system as a program rather than how they run the line.
• Mislabeling cost-cutting as lean. The tell: the initiatives mainly eliminate jobs, divert revenue from customers, or squeeze suppliers, rather than improve flow, quality, and cost of specific products. That is 'mean,' not lean.
• Pushing hours and effort instead of method. The tell: performance targets are met through sweat and longer hours because nobody knows how to work smarter with flow and pull.

• ▸Product/value stream: name the one specific product being redesigned.
• ▸Customer-view box score: total elapsed time, value-adding time, wait time, number of queues, handoffs, repeat inspections, total steps.
• ▸Baseline metrics (verified on the floor): labor productivity, throughput time, inventory, customer-reaching errors, scrap, injuries, space and tooling used.
• ▸Kaikaku target: double productivity, cut throughput time and inventory 90 percent, halve errors, scrap, and injuries.
• ▸Flow design: changeover method, right-sized machines, and cell layout that puts process types adjacent.
• ▸Pull and stop rules: what the customer's demand triggers, and what automatically stops the line on a mistake.
• ▸Post-change results and freed assets: re-measured metrics plus space, equipment, and tooling freed or sold.
• ▸Sustainment owner: named line manager, coaching schedule, and loose-end list with due dates.
• ▸Kaizen target (2 to 3 years): productivity doubled again, with inventories, errors, and lead times halved again.

1. 1Specify value by specific product, not by company or category. Write one value statement per product family: what the customer needs, at what price, delivered at what time. The passages treat this as the first of five principles (specify value, identify the value stream, make value flow, let the customer pull, pursue perfection).
2. 2Talk to the people you are not serving. Doyle Wilson surveyed his own new-home buyers until he asked who bought the 78 percent of homes that are older. Build a list of non-buyers and lapsed buyers and interview them about why they chose the alternative.
3. 3Capture the pain in the customer's own terms. Wilson's older-home buyers named a 'hassle factor' in negotiating, long lead times to move in, an inevitable 'to-be-done' list after move-in, and 'phony choices' loaded as standard equipment. Record each complaint as a specific, countable friction, not a theme.
4. 4Test your offer against the three satisfaction conditions the passages name for Lantech: performance, price, and service. Mark each as met, exceeded, or missed for the specific product and customer. Note that Lantech's customers were already satisfied on all three, which made its lean transformation easier.
5. 5Remove sources of hassle by mapping the value stream. Map concept to launch, order to delivery, and raw material to finished good, and mark the unnecessary time, effort, and error sources that reach the customer. The passages call value stream mapping a way to raise consciousness about value and its components, leading to action.
6. 6Align internal incentives with customer value. Wilson eliminated individual sales commissions ('which destroy quality consciousness') and the builder bonus that superintendents earned for 'on-time completion' through side deals on a to-be-done-later list. Audit your bonuses for rewards that trade customer experience for a metric.
7. 7Shorten order-to-delivery so you can make to order instead of forecasting. Converting to continuous flow with customer pull typically halves errors reaching the customer and halves time-to-market, and lets you offer wider variety at modest additional cost.
8. 8Re-measure with customer surveys on the experience and with inventory turns, which the authors call 'the lean metric that cannot lie'. Repeat the cycle, because value is a moving target (the Lantech chairman starts rethinking product value again once the transformation is done).

Doyle Wilson Homebuilder ran a full Total Quality Management campaign starting February 1992. He taught TQM personally, collected data on every aspect of the business, dropped individual sales commissions, ended the builder bonus that superintendents earned through side deals on a 'to-be-done-later' list, and cut his contractor corps by two thirds. Customer satisfaction rose steadily, sales grew in a flat market, and he won the National Housing Quality Award in 1995 with the Baldrige as his next target. Then one fact bothered him: 78 percent of central Texas homes bought each year are older homes, so he had been competing for only the 22 percent seeking new construction. He began interviewing older-home buyers instead of new-home buyers. They told him they hated the negotiation hassle, the long lead time to move in, the post-move-in to-be-done list, and the phony choices of 'custom' homes padded with standard features of little value. The lesson is that surveying existing buyers measured satisfaction inside the old definition of value, while talking to non-buyers exposed a different definition and 'a complete rethinking of his business.'

• If satisfaction scores rise but share of the total market does not, then you are surveying the wrong people; interview those who bought elsewhere, as Wilson did with the 78 percent.
• If customers are already satisfied with performance, price, and service, then put effort into flow and cost inside your operation (Lantech's case) rather than redesigning the offer.
• If an incentive pays staff for hitting a completion or sales metric, then check whether it lets them defer or hide customer-facing problems; remove it if it does (commissions, on-time bonuses).
• If customers get high satisfaction only through high cost or effort, then redesign for high satisfaction at low cost (Toyota's door-to-door selling is the passages' example of a system that needs another leap).
• If you rely on forecasts, then cut order-to-delivery time and add or subtract capacity in small increments so you can make to order, since forecasts are reliably wrong.

• Defining value from the producer's side: adding features, 'standard equipment,' or 'custom' options the customer sees as phony. The tell: customers complain about choices that cost them money but give them little.
• Improving quality techniques without questioning the market. Wilson won quality awards yet was serving only 22 percent of buyers. The tell: awards and rising survey scores alongside a shrinking addressable market.
• Drowning in techniques without understanding the whole system. The tell: isolated tools (teams, target pricing, level scheduling) are in place, but the customer's experience of lead time and errors is unchanged.
• Incentives that reward the metric rather than the customer. The tell: superintendents earn on-time completion bonuses while customers carry a growing to-be-done list.
• Treating value as settled after one transformation. The tell: no one asks what the customer values now; Lantech's leadership returned to rethinking product value only after the flow conversion.

• ▸Product family and customer segment: name the specific product and who buys it.
• ▸Value statement: what the customer needs, at what price, by when.
• ▸Who I surveyed vs. who I did not: list current buyers and the non-buyers or competitor-buyers to interview.
• ▸Reported hassles (countable): list each friction in the customer's words (negotiation, lead time, post-delivery fix list, phony choices).
• ▸Performance / price / service rating: met, exceeded, or missed for each, with evidence.
• ▸Value stream sources of delay and error: steps from order to delivery that add time, effort, or defects the customer sees.
• ▸Incentives that could conflict with customer experience: bonuses or commissions that reward a metric over satisfaction.
• ▸Current order-to-delivery time and inventory turns: baseline numbers to track.
• ▸Next action and re-check date: one change aimed at the biggest hassle, and when you will re-ask customers.

1. 1Define value from the ultimate customer's side before touching cost. Write down the specific product (good, service, or both), the specific customer, the specific price, and the specific time. If you cannot state all four, you are still working from your own organization, technology, and core competencies, not from value.
2. 2Run the executive test. Ask: can you put yourself in the position of a design moving from concept to launch, an order moving from request to delivered product, and the physical product moving through the chain? If the answer is no, you do not yet see where growth is leaking.
3. 3Separate cost-cutting from lean. Screen every initiative: does it eliminate jobs, divert revenue from downstream customers, or squeeze upstream suppliers? Then it is 'mean,' not lean. Keep only initiatives that convert muda into value for the customer.
4. 4Convert batch-and-queue to flow along the value stream and measure the effect on customer-facing outcomes: throughput time, defects reported by customers, on-time shipment to the agreed date, inventory, and space. Track inventory turns as the metric that cannot lie.
5. 5Protect the people who free up capacity. Promise that no one is let go because of the lean conversion, and move freed-up workers (the best, not the worst) onto a kaizen team that improves other activities. This keeps improvement from being read as a threat.
6. 6Use the freed capacity to grow. Because capacity rises without added headcount, price aggressively to capture share and launch new products, rather than banking the gain as layoffs.
7. 7Make profit visible and shared. Show everyone the financials so the link between their work and profitability is clear, and tie a fixed share of pretax profit to pay.
8. 8Extend the work upstream. Cut the supplier base to a number you can actually teach, start with the most critical suppliers, and improve their performance instead of only negotiating their margins.

Lantech converted its whole production system from batches to single-piece flow by fall 1992, including the cell for its $50,000 H model, built at one per week. Headcount stayed at three hundred while machines shipped doubled between 1991 and 1995. Customer-reported defects fell from 8 per machine to .8. Production throughput time fell from sixteen weeks to fourteen hours. Shipments on the date agreed with the customer rose from 20 to 90 percent. The plant, once bulging with inventory, had 30 percent excess space despite doubled output. The growth came from market recovery, aggressive pricing to capture share, and new products. Pat Lancaster paid for that capacity by promising no one would be let go and by creating a kaizen team of freed-up workers led by the skeptic Bob Underwood. Wiremold shows the profit side: profit sharing rose from 1.2 percent of wages in 1990 to 7.8 percent in 1995 as financials were shown to all employees.

• If a savings program works by cutting jobs, squeezing suppliers, or diverting customer revenue, then reject the 'lean' label and redesign it around removing waste from the value stream.
• If lean frees up capacity, then commit it to growth (share capture, new products, pricing) rather than headcount reduction, and say so to the workforce first.
• If you cannot name the specific product, customer, price, and time, then stop improving processes and define value first.
• If suppliers are numerous and bought only on competitive margin, then consolidate to a number you can teach and start with the most critical ones.
• If employees cannot see how their work drives profit, then open the financials and tie pay to profit share before expecting engagement.

• Labeling quarterly cost-cutting as lean. The tell: the program's results are headcount and supplier price reductions, with no change in lead time, defects, or on-time delivery.
• Defining value from internal capabilities. The tell: presentations center on technology and core competencies, and nobody can name the specific product, customer, and price.
• Banking efficiency as layoffs. The tell: employees hide improvements or resist kaizen because gains threaten their jobs, and the removed incentive stalls further progress.
• Optimizing a local metric that creates queues, such as low freight cost per part. The tell: a month of parts queued at suppliers, again at inspection, again awaiting a full truck, with inventory growing as part numbers grow.
• Improving only inside the plant. The tell: internal gains plateau while purchased materials remain a large share of cost and supplier performance has never been addressed.

• ▸Specific product and customer: name the good/service and the customer who defines its value.
• ▸Price and time: state the price and delivery time at which the product meets the customer's need.
• ▸Baseline throughput time: record order-to-delivery or concept-to-launch time before the change.
• ▸Customer-reported defects per unit: record before and after.
• ▸On-time to agreed date (%): record before and after.
• ▸Inventory turns and space used: record before and after; note excess space freed.
• ▸Capacity use plan: list where freed capacity goes (pricing for share, new products, other lines) and confirm no job loss.
• ▸Profit-sharing link: state the share of pretax profit paid out and how financials are shown to staff.
• ▸Supplier plan: count current suppliers, target count, and the first critical supplier to teach.

1. 1Accept the premise first: customers 'are certainly not interested in the security of your job.' Job security therefore cannot come from defending existing activities. It has to come from growth and sustainable value for customers, employees and owners. Write this down as the logic of the program before you start.
2. 2Define value from the customer's side with a small cross-functional team, not inside departments. Wiremold sent a marketer, a designer/product engineer and a production/tool engineer to talk directly with building-design and construction customers and ask the 'value question.' This is the engaging-work half of the construct: people get closer to the customer and see their contribution.
3. 3Pick a starting activity that is important, visible and performing very poorly. Involve the direct work group, every management level between you and them, the senior executives you want to convert, and your sensei. Include the people whose jobs the change will touch from day one, so it is not done to them.
4. 4Name the security people feel in the old arrangement. Wiremold old-timers asked why a tool designer was in the field, because the 'departmentalized, everything-in-its-place' way of organizing work felt safe. List who holds that feeling and what they fear losing, then answer it with a concrete new role.
5. 5Before the improvement lands, decide where freed people and capacity will go. The construct's commitment is to redeploy resources toward growth rather than lay off people whose jobs the improvement eliminates. The passages do not give a mechanism, so you write one: a named destination (new product, new capacity, new work) for each eliminated task.
6. 6Build multifunctional teams that sit together and learn on the floor. Wiremold co-located the product team (team leader, production planner, production/tool engineer, buyer) and, late in the process, moved their desks to the factory floor for process-at-a-glance and standard work exercises with the production team. Shared space turns specialists into people with broader, more engaging work.
7. 7Demand immediate, verified results, then publish them to the people involved. Wiremold's first product under the new regime took six months and tool costs were 60 percent of the original. Visible wins build morale and make the redeployment promise credible.

Wiremold's Tele-Power Pole development shows the engaging-work side. Steve Maynard formed a three-person team (marketer, designer/product engineer, production/tool engineer) and sent it to talk directly to prospective customers. They returned with a clear definition: a pole that fits any ceiling height, comes in a wide range of colors and is unobtrusive. Old-timers were amazed that a tool designer was in the field, and Maynard recalls how secure many felt in the departmentalized structure. A co-located team of team leader, production planner, production/tool engineer and buyer then worked to a target cost, and moved to the factory floor near the end. By mid-1992 the first product was ready in six months, with tool costs at 60 percent of the original. The passages do not report what happened to headcount, so the redeployment commitment is not shown. That is the step you must add yourself: before the next project, name where any freed capacity will go.

• If someone argues for an improvement because it protects jobs, then reframe it around customer value and growth. The customer will not pay for job security, so security has to follow from growth.
• If an improvement will eliminate tasks, then state the redeployment destination before work starts, not after the savings appear.
• If people resist because the current departmental structure feels safe, then put them on a cross-functional team with direct customer contact so the new work is visibly larger than the old.
• If you must choose a pilot, then choose the important, visible, poorly performing activity, where improvement potential is large and failure is unaffordable.
• If lean is applied only to an easy corner of the value stream (for example flow in one part of one plant), then expect limited results and little change for people. Extend to the whole value stream.

• Selling lean as job protection to customers or executives. The tell: the business case cites internal comfort rather than customer value.
• Improving first and deciding about people later. The tell: freed capacity appears and the only available option is headcount reduction, and staff stop volunteering improvements.
• Leaving old specialists out of customer contact. The tell: engineers and tool designers still say 'division of labor requires I stay at my desk,' and morale gains stay confined to a few.
• Treating a small, easy pilot as proof. The tell: the plant still carries about eight weeks of parts and builds to a wrong forecast, because supplier and upstream behavior never changed.
• Claiming morale gains without verifying them. The tell: no before/after data and no direct observation of the people doing the work.

• ▸Improvement activity: name the specific value stream or activity being improved.
• ▸Customer value statement: one sentence of what the customer wants, from the customer's side.
• ▸Tasks/roles likely eliminated: list each task and the people currently doing it.
• ▸Redeployment destination: for each task, name the growth work (product, capacity, new role) that absorbs the person.
• ▸Engaging-work change: describe what the person does differently (customer contact, cross-functional team, floor time).
• ▸Security concerns heard: record the fears of those used to the old departmental arrangement and the response given.
• ▸Team composition and co-location: who sits together, where, and for how long.
• ▸Commitment owner and date: the executive who signs that no one is laid off as a result of this improvement, and the date.
• ▸Result check: before/after measure (time, cost) verified on the floor, plus what each redeployed person is now doing.

1. 1Name the single product or service you are specifying. Value is specified by specific product, not by company, division, or aggregate portfolio. If you cannot name one product, split the portfolio until you can.
2. 2Pick the ultimate customer and write down who they are. In the Wiremold case, leading customers were major contractors, so ask who actually installs, pays for, and lives with the product, not who buys it from you.
3. 3Set up a joint dialogue between customer and producer, staffed by a cross-functional team that stays with the product for its whole production life. Wiremold used a marketer, a product engineer, and a tooling/process engineer. Do not run a departmental hand-off of survey, then design, then prototype, then machine design.
4. 4Ban the old products, solutions, and competitor offerings from the conversation. Ask what the customer really needs, not which of your existing features they prefer. Existing assets, plants, and engineering preferences are not inputs at this stage.
5. 5Capture value in the customer's terms and test it against the whole product. Record capabilities, such as looks and installation speed as well as ruggedness, safety, and cost per foot. Record the price and the time (when and how fast it must be available). Include what other firms along the value stream contribute, because each firm tends to define value only for its own piece.
6. 6Ask the customer for trade-offs, not wish lists. Find out what they will give up (cost per foot) to get what they want (appearance, quicker installation), and quantify it in their economics, such as bid price and total installed cost.
7. 7Set a target cost after the product is defined. Take the specification and capabilities that competitors currently offer, then ask what the cost would be with all currently visible muda removed. Do not work backwards from a price the market will bear.
8. 8Treat the first pass as kaikaku for value, then schedule recurring product-team reviews that ask whether this is still the best answer. That is the kaizen analog for value specification.

Wiremold, 1992. Product development ran through departments: marketing commissioned competitor-comparison surveys, product engineering designed to a gap, a prototype group tested, and machine designers tooled. The result was designs that lacked imagination and that customers often ignored. Speeding this up through simultaneous engineering or adding variety would only have delivered bad designs faster. Steve Maynard instead formed a team per product (marketer, product engineer, tooling/process engineer) and held a dialogue with major contractors in which all old products and solutions were ignored. Wire guides had been specified almost entirely on ruggedness, safety, and cost per foot delivered to the site, a view that suited the engineers who dominated development. The dialogue showed that customers also wanted a product that looked nice and could be installed quickly. They would trade cost per foot for appearance, which raised the construction bid price, and for quicker installation, which cut total cost. Once all product families got this treatment, sales of these conventional products rose more than 40 percent within two years and gross margins soared.

• If your value statement is built from your existing features, plants, or technology, then discard it and restart with the customer's stated need, with old products excluded from the discussion.
• If the specification is in the engineers' vocabulary (ruggedness, refinement, complexity) and no customer evidence backs it, then treat it as the German-style engineer-defined distortion and go get the evidence.
• If the specification is driven by aggregated financial targets or what the market will bear, then reverse it: define the product from the customer's view first, then set target cost from the current competitor offering minus visible muda.
• If your product is only one part of what the customer uses, then define value for the whole product, jointly with the other firms along the value stream.
• If the specification was done once and then left alone, then schedule a recurring product-team review, because value specification needs its own kaizen.

• Surveys and competitor comparisons replace dialogue. The tell: your opportunities are gaps in the market or competitor weaknesses, and customers ignore the resulting designs.
• Engineers define value. The tell: the justification is "the customer will want it once we explain it," and failures are blamed on customers not being sophisticated enough.
• The financial mind-set displaces the customer. The tell: executives cannot describe what happens to a design, an order, or a product at each step, and the discussion drifts back to aggregated financials.
• Speeding up or broadening an unexamined product. The tell: faster development and more variety, but still ignored designs, which the authors call pure muda.
• Value is specified only for your own firm's piece. The tell: each firm along the stream defines value separately, such as a car company selling a car and a dealer selling services, instead of the jointly provided personal mobility the customer wants.

• ▸Specific product or service: name one product, not a family or portfolio.
• ▸Ultimate customer: who installs, uses, and pays in the end, and who participated in the dialogue.
• ▸Team and dialogue: names and roles on the cross-functional team, and the date and format of the customer sessions.
• ▸Excluded baseline: list the old products, solutions, and assets explicitly set aside in the discussion.
• ▸Capabilities required: what the customer says they need, in their words, beyond the traditional specification.
• ▸Trade-offs quantified: what the customer will give up for what, with the effect on their bid price or total cost.
• ▸Price and time: the price point and the timing or speed at which the customer needs it.
• ▸Whole-product and partners: other firms along the stream that contribute to the value and what each must change.
• ▸Target cost: the competitor-equivalent specification minus all visible muda, with the resulting cost figure.
• ▸Next review date: when the product team will ask whether this is still the best answer.

1. 1Pick one product family and map only that. The passages frame the value stream 'for each product,' running from raw material to finished good, order to delivery, and concept to launch. Name the family and the customer before drawing any box.
2. 2Walk the stream yourself, from the customer back upstream to the supplier base. Authors who 'walked the production floor' and spent extended periods in engineering, sales, customer support and purchasing found what summaries missed. Record what actually happens, not what the procedure says.
3. 3List every step across the three task types: problem-solving (concept to launch), information management (order to delivery, including scheduling signals such as MRP production orders), and physical transformation (raw material to customer). Do not map only the factory floor.
4. 4Mark each step with one of three labels: value-creating, Type 1 muda (non-value-creating but unavoidable under current technology and assets), or Type 2 muda (non-value-creating and avoidable now). Define 'value' first, by specific product, as the customer would define it.
5. 5Capture the numbers for a Current State box score: total time through the process, work-in-process count, space needed, changeover downtime and cost. These are the baseline against which any future state is judged.
6. 6Draw the Future State. Remove Type 2 muda first, convert batches to flow, and replace push scheduling with pull. In the passages' example, MRP is disconnected and kanban signals from a heijunka box pace a 'pacemaker process,' with pull loops running upstream to stamping and the steel-coil supplier.
7. 7Finish the maps in a week or two, with no simulations or benchmarking detours. Then assign one named person to manage and improve the value stream and write an implementable plan with steps and dates.
8. 8Start on a troubled activity, usually in physical production, and expect dramatic results in the first week. After the first round of improvements, link adjacent parts of the stream. Once the Future State is achieved, relabel it the Current State and map again.

North Haven's turbine blade fabrication shows what mapping exposes. The process mapping team found that the Blohm machining centers caused 'the great majority of the cost in the total process,' and that their erratic performance thwarted smooth flow in the rest of the line. Hiding them behind a curtain wall had not worked. The team replaced each Blohm with a cell of eight simple three-axis grinding machines using quick-change fixtures, one worker per cell, and changeovers in under two minutes. Actual processing time rose from three minutes to seventy-five minutes, yet total time through the process fell from ten days to seventy-five minutes. Changeover downtime dropped by more than 99 percent, parts in process fell from about 1,640 to 15, and space fell by 60 percent. Total manufacturing cost was projected to fall by more than half. The lesson for your map is to judge a step by its contribution to the whole stream's average velocity, not by its 'point velocity.'

• If a step is non-value-creating and you can remove it with current technology and assets, then label it Type 2 muda and eliminate it first.
• If a step creates no value but the process cannot run without it today, then label it Type 1 muda, keep it, and schedule it for reduction later.
• If a single step dominates cost and disrupts flow everywhere else, then redesign or replace it rather than working around it.
• If you are tempted to simulate, benchmark, or extend the planning exercise, then stop and start removing muda. Delay used to avoid action is itself muda.
• If the Future State map exists but no one owns the value stream and no plan exists, then treat the mapping as unfinished and name the owner and plan before doing anything else.

• Beautiful maps, no change. The tell: a walk along the value stream shows no actual Future State, and nobody can show an implementable plan or name the responsible manager.
• Mapping only physical production. The tell: the map has no information-management flow (MRP orders, scheduling) and nothing on concept-to-launch.
• Judging steps by point velocity or hourly labor cost. The tell: you target the cheapest-per-hour workers or the fastest machine instead of average velocity and total stream length.
• Over-planning and simulating. The tell: a complex 'what if' model whose unsettling predictions mean nothing ever moves.
• Calling avoidable steps 'necessary' to avoid disruption. The tell: almost nothing is labeled Type 2, and the first week produces no dramatic gain such as halved effort or 90 percent less work-in-process.

• ▸Product family and customer: name the one family being mapped and what its customer values.
• ▸Stream segment: mark each step as problem-solving, information management, or physical transformation.
• ▸Step description: record what actually happens, as observed on the walk, not the written procedure.
• ▸Classification: label the step Value-creating, Type 1 muda, or Type 2 muda.
• ▸Evidence for classification: state why the customer would or would not pay for the step, and why it is or is not avoidable now.
• ▸Current State metrics: record total time through the process, WIP count, space, changeover downtime, and cost.
• ▸Future State action: write the change (eliminate, flow, pull, replace) and the target metric for each step.
• ▸Value stream owner: the named person accountable for the Future State.
• ▸Plan and dates: list the implementable steps, due dates, and the date the Future State becomes the new Current State.

1. 1Pick one product or product family and define the flow from the product's point of view, not the department's or the machine's. Use the newsletter test: ask how the item would get finished fastest with the least effort. Then compare that path with the current batch-and-queue path (fold all, label all, seal all, stamp all).
2. 2Write out every value-creating step in sequence, such as molding, painting and assembly. Mark each place the product waits for a batch to fill or for a department to pick it up. Those places are where you break the batch.
3. 3Rearrange the steps into a cell in true process sequence, so the product moves one piece at a time. Ohno's group did this in low-volume work, in most cases without assembly lines. Kaikaku-style rearrangement can be done in a day.
4. 4Right-size the machines. Where a large shared machine forces batching, replace or supplement it with smaller machines that can sit directly beside the next step. Then objects can move from one process type to the next without a queue between them.
5. 5Cut changeover time so small lots are economical. Ohno and Shingo's central insight was quick tool change from one product to the next. Pratt & Whitney took changeover from hours to seconds, which let it make exactly what the customer ordered.
6. 6Standardize the work and maintain the tools in the cell. Big batches and mountains of inventory hide slow tool changes and tool neglect. Wiremold's tools had deteriorated unnoticed behind its buffers. Once you remove the buffers, a defective tool stops the line right away.
7. 7Pair flow with level scheduling and customer pull before you cut inventory. Wiremold cut inventories and lot sizes without reducing changeovers or building to a level schedule, and customer service collapsed. Measure the result: labor productivity, throughput time, inventory, errors and scrap.
8. 8Treat the conversion as the start, not the end. Once the kaikaku rearrangement is done, run kaizen. Flow exposes hidden muda, and that is the signal for the next round of improvement.

Pratt & Whitney had a totally automated grinding system for turbine blades. Its own engineers replaced it with a U-shaped cell, designed and installed in a short time at a quarter of the capital cost of the system it replaced. The cell cut production costs by half and throughput time by 99 percent. Changeover fell from hours to seconds, so Pratt could make exactly what the customer wanted on receiving the order. The conversion was expected to pay for itself within a year, even if the automated system fetched only scrap value. The case shows the pattern: put the steps in sequence in a compact cell, right-size the equipment, and make changeover trivial.

• If volumes are low and product mix is varied (dozens or hundreds of units, not millions), then do not copy Ford's dedicated high-speed line. Build a product-family cell with quick changeover and right-sized machines.
• If a large shared machine forces you to batch, then look for a smaller machine that can sit next to the adjacent step before accepting the batch.
• If you plan to cut lot sizes or inventory, then first reduce changeover time, level the schedule, and fix tool maintenance. Otherwise you repeat Wiremold's service collapse.
• If flow exposes new problems after conversion, then treat them as the hidden muda flow is meant to reveal. Hold course and move to kaizen rather than reverting to batches.
• If the people doing the work insist that batching is more efficient, then run a one-piece trial and measure it against the batch path. The newsletter exercise shows intuition favors batching, so rely on measured results.

• Cutting inventory and lot sizes without enabling flow. The tell: late deliveries, expedited freight and extra customer-service staff, as at Wiremold between 1989 and 1991.
• Rearranging by department convenience or equipment utilization, not by the product's needs. The tell: the product still travels between departments and waits for a full batch.
• Copying the Ford special case where it doesn't apply. The tell: a dedicated fast line that only works with identical parts, one model and enormous volumes.
• Ignoring tool condition once buffers are gone. The tell: defective tools discovered at installation and parts running out, because maintenance was skipped under the old regime.
• Reaching for automation instead of a simple cell. The tell: large capital spend on a system that still batches, when a simple U-shaped cell at a quarter of the cost could do better.

• ▸Product family: name the product(s) that share the same processing steps.
• ▸Current path: list each step and mark every wait for a batch or department handoff.
• ▸Current batch size and throughput time: record the baseline.
• ▸Proposed cell layout: list the steps in true process sequence, such as a U-shape.
• ▸Machines to right-size: name each large shared machine and the smaller replacement or alternative.
• ▸Changeover time today vs. target: record the current time and the goal, in hours or seconds.
• ▸Tool maintenance check: note tool condition and who inspects it, so buffer-hidden defects surface.
• ▸Scheduling and pull link: note how the cell is leveled and what customer demand signal drives it.
• ▸Results to measure: record labor productivity, throughput time, inventory, errors and scrap before and after.

1. 1Do not start with pull. Start by making the product flow. The sequence in the source is dedicated product teams and flow first, then pull. Pull needs short throughput times to work. Ask: can this process make any product in any combination on short notice? If not, you are not ready to cut inventory.
2. 2Set throughput targets before you touch inventory. The benchmarks are 50 percent less time in product development, 75 percent in order processing, and 90 percent in physical production. If you cannot reach these quickly, the source says you are doing something wrong. Record the current times as the baseline.
3. 3Remove the changeover barrier. Pratt & Whitney's U-shaped cell cut changeover from hours to seconds, which let it make exactly what the customer wants on receiving the order. Large lots and slow tool changes only work because inventory hides them. Reduce changeover time before you reduce lot size.
4. 4Build to a level schedule. Wiremold's VP of operations cut inventories and lot sizes but could not introduce flow and pull, because no one knew how to reduce changeovers and build to a level schedule. Level the schedule first, so the upstream process sees a steady signal rather than spikes.
5. 5Find the hidden buffers before you remove the visible ones. Wiremold's MRP carried a 50 percent extra margin in its safety-stock calculation, and big batches let tool maintenance slide. List every buffer: safety stock margins, batch sizes, tolerated tool defects, and slack in maintenance. Fix the capability behind each buffer, then remove the buffer.
6. 6Replace forecast-driven scheduling with a signal from the customer end. Pratt replaced MRP across the company with a pull system from final assembly. Make final assembly the only point scheduled from actual orders. Every upstream step then makes only what the next step takes.
7. 7Stop pushing demand. Drop the periodic discounting campaigns meant to move goods already made. The source says demand becomes more stable when customers know they can get what they want right away.
8. 8Track inventory turns, which the source calls the lean metric that cannot lie. Treat each stockout or late delivery as a sign of an impediment to flow that pull has exposed. Fix the impediment and keep pulling. Do not add the buffer back.
9. 9Plan for a visible dip and hold course. Pratt delivered only 10 percent of engines on time in spring 1994, and it took more than three years before turning back became unthinkable. Agree with senior management in advance that setbacks are expected.

Wiremold and Pratt & Whitney show pull done in the wrong and the right order. In 1989 Wiremold's VP of operations returned from Japanese factories enthusiastic about Just-in-Time. He pulled down inventories and reduced lot sizes. Nobody knew how to shorten tool changeovers or build to a level schedule, so flow and pull never arrived. Customer service, in Orrie Fiume's words, 'went completely to hell.' Wiremold found a 50 percent extra margin buried in its MRP safety-stock calculation. It also found that mountainous inventories had let it tolerate slow tool changes and skimp on tool maintenance, so tools had deteriorated badly. From 1989 to 1991 the company slid from record profits to breakeven. Costs rose through express freight and added customer service staff, while sales fell only a few percent. Pratt & Whitney did the capability work first. It replaced an automated turbine-blade grinding system with a U-shaped cell at a quarter of the capital cost. Production costs fell by half, throughput time by 99 percent, and changeover from hours to seconds, so it could make exactly what the customer ordered. A pull system from final assembly then replaced MRP across the company. In spring 1994 only 10 percent of engines were delivered on time, but management held course. After the new final-assembly management and quality initiatives took hold, results turned quickly.

• If changeovers still take hours, do not cut inventory or lot size. Reduce changeover time first, as Pratt did, or you repeat Wiremold's collapse in customer service.
• If the schedule is driven by a forecast or by MRP with padded safety stock, then replace it with a pull signal from final assembly. Strip out the safety-stock padding as you do so.
• If a stockout appears after you cut stock, then treat it as a revealed impediment: look for a deteriorated tool, a slow changeover, or an unlevel schedule. Fix the cause and keep pulling.
• If finished goods are being moved with discounts, then stop the campaign and shorten the lead time instead. The source ties stable demand to quick availability, not to promotions.
• If throughput time has not fallen by about 50 percent (development), 75 percent (order processing), and 90 percent (production), then flow is not yet real. Do not declare the pull system finished.

• Cutting inventory as the first move and calling it JIT. The tell is falling customer service, rising express freight, and extra customer service staff, as at Wiremold.
• Pulling without leveling the schedule. The tell is that upstream processes swing between idle and overloaded, and the lot sizes you cut creep back.
• Leaving the buffers inside the planning system. The tell is that the shop floor shows low stock while MRP safety-stock formulas still add large margins.
• Ignoring tool and equipment maintenance. The tell is that defective tools surface only when you run out of parts, because the old stock hid the deterioration.
• Quitting at the first bad results. The tell is that on-time delivery drops (Pratt hit 10 percent) and leadership asks for a rollback before the pull system has run through the whole chain.

• ▸Product/value stream: name the one product family and its customer-facing endpoint.
• ▸Baseline throughput times: record current days or hours for development, order processing, and production.
• ▸Throughput targets: write the 50% / 75% / 90% reduction targets as actual numbers for this stream.
• ▸Changeover time per process: list each step's changeover time and the target. Flag any step still measured in hours.
• ▸Schedule leveling status: state whether final assembly is built to a level schedule, and by what rule.
• ▸Hidden buffers found: list safety-stock margins, batch sizes, and maintenance slack that currently mask problems.
• ▸Pull trigger point: name the single point (such as final assembly) where actual customer orders set the schedule.
• ▸Forecast/MRP and discount dependencies to retire: list each one and the date it is switched off.
• ▸Impediments revealed: log each stockout or delay, its root cause, and the fix. Do not log a restored buffer as a fix.
• ▸Inventory turns: record the baseline and the current value each review.

1. 1Set the standard at zero muda, not at competitors. Define the target as the elimination of all activity that absorbs resources but creates no value (mistakes needing rectification, unwanted items, unneeded processing steps, needless movement and transport). Drop benchmarking as the yardstick: a lean firm that finds itself ahead relaxes, and a mass producer that finds itself behind gets distracted by factor costs, scale, or 'culture'.
2. 2Make perfection a standing assumption, stated aloud by senior management. Use Wiremold's version: every value stream can always be improved and must be improved. Pair it with a time expectation: 'if you can't get a major improvement in three days you are doing something wrong.'
3. 3Reorganize around the value stream so improvement has an owner. Dissolve functional departments and 'process villages' and reassign people to dedicated product teams that hold all the resources for one product family. Wiremold created six teams, one per product family, and eliminated the separate purchasing, scheduling, engineering, stamping, rolling, molding, painting and assembly groupings.
4. 4Build an improvement engine with defined roles. Set up a promotion office (Wiremold's JIT Promotion Office). The product team leader and the office jointly assess the value stream and decide which kaikaku (radical) and kaizen (incremental) activities to run and when. Assign each improvement team a team leader from the product team and a facilitator from the office. The facilitator owns the follow-up work, because the team leader returns to the regular job when the week ends.
5. 5Run short, repeated kaizen events on every stream, involving practically everyone. Revisit the same activity again and again. Freudenberg-NOK's vibration-damper work at Ligonier gained 56 percent labor productivity and cut factory space 13 percent in the first event, and five further three-day events kept finding muda in effort, time, space and errors.
6. 6Teach and reteach the principles and techniques. Cover value stream, flow, pull and perfection, plus standard work, takt time, visual control, pull scheduling and single-piece flow. Middle management feels threatened by the loss of safety nets and will drift back to batches and inventory unless teaching is paired with continuous hands-on improvement exercises.
7. 7Remove the fear that stalls improvement. Guarantee that no one loses a job because of kaizen, and tie that guarantee to flexibility in work rules. Otherwise the workforce cannot tell layoffs from weak demand from layoffs from kaizen, and resistance becomes chronic.
8. 8Feed each cycle's learning back into the other four principles. Faster flow exposes hidden muda, harder pull reveals impediments, and dedicated product teams in dialogue with customers specify value more accurately. Extend the target outward: Lantech is now rethinking its own product for the small-lot, single-piece-flow world its customers are heading toward.

Joe Day, president of Freudenberg-NOK General Partnership, began introducing lean thinking in 1992 and noticed that no matter how many times employees improved an activity, they could always remove more muda from effort, time, space and errors. At the Ligonier, Indiana, plant, reorganizing vibration-damper manufacture produced a first kaizen event with a 56 percent labor-productivity gain and a 13 percent cut in factory space. The team then returned to the same activity in five additional three-day kaizen events, and each pass found further waste. The activity also grew more flexible and responsive to customer pull. Wiremold institutionalized the same pattern: hundreds of weeklong kaizen activities involving practically every employee, repeated on every value stream, backed by the rule that a major improvement should arrive within three days. Art Byrne supported it with six dedicated product teams, a JIT Promotion Office whose facilitators finish the follow-up work, and a guarantee that no job is lost to improvement.

• If your performance beats competitors' and the team is easing off, then reset the target to perfection (zero muda) instead of the benchmark, because relative standards invite relaxation.
• If a kaizen event on an activity yields a gain, then schedule it again: the first pass is never the last, and Freudenberg-NOK needed five more three-day events after the first.
• If an event ends without a major result in about three days, then treat it as a design or execution problem with the event, as Wiremold does, and not as evidence that the process is already lean.
• If improvement efforts stall or middle managers drift back to batches, then increase teaching and hands-on events and check whether job security and work rules have been settled.
• If a team leader will return to the regular job after an event, then assign a facilitator who owns the follow-up work, so the weeklong effort is completed.

• Treating perfection as a slogan or a one-time project. The tell: after the first kaizen event, no one schedules a second pass over the same activity.
• Benchmarking against competitors and stopping when you lead. The tell: improvement targets are expressed as gaps to rivals, and effort fades once the gap closes.
• Improving without job security. The tell: workers resist or hide gains, and cannot distinguish layoffs from weak demand from layoffs from kaizen, so resistance to improvement is chronic.
• Leaving improvement to functional departments instead of value-stream teams. The tell: events optimize one 'village' while the product's flow across stamping, painting, assembly and scheduling is unchanged.
• Letting events end without follow-up, or letting the teaching lapse. The tell: action items from a weeklong event are unfinished, and middle managers, feeling threatened, drift back to batches and inventory.

• ▸Value stream / product family: name the single stream under improvement.
• ▸Muda observed: list mistakes, unwanted output, unneeded steps, movement and transport, waiting.
• ▸Target in effort, time, space, cost, mistakes: state the 'perfect' end-state for each, not a competitor gap.
• ▸Kaikaku or kaizen: mark each action as radical redesign or incremental improvement, with timing.
• ▸Event dates and pass number: record the event length (aim for results within three days) and which repeat pass this is for the activity.
• ▸Team leader and facilitator: name the product-team leader and the promotion-office facilitator who owns follow-up.
• ▸Measured result: record the change in productivity, space, lead time and errors after the event (e.g., +56% labor productivity, -13% space).
• ▸Follow-up actions and owner: list unfinished items from the event and who closes them, with due dates.
• ▸Next pass and teaching needed: note what remains to attack next, and which principles or techniques need reteaching.