Implement Lean in one bounded product family before applying changes across a plant. A useful first 90 days gives a small team time to define customer demand, observe the actual route, measure a baseline, test one countermeasure, and check whether the change holds. Do not start by buying a dashboard or launching a plant-wide tool campaign.
Small, evidence-led rollout rather than a site-wide decree. Mechanitec instructional diagram.
This guide provides a pilot charter, sample calculations, a roadmap, and a review board. All figures below are illustrative examples, not measured results. Adapt the process boundary and targets to your products, customer requirements, and approved work instructions.
Week 0: write the pilot charter
The plant manager names a sponsor and process owner. The planner, operators, quality, maintenance, and engineering contribute when their work is inside the selected route. Choose one family with a repeatable route and enough jobs to observe. State why the family matters, where the measurement starts and ends, and what must not get worse.
Copyable pilot charter for a small manufacturing team
Charter field
What to record
Customer need
Product, quantity, due-date, specification, and service condition
Product family and boundary
Included part numbers and first/last operation in scope
Problem statement
Observed gap, unit, sample, time window, and source
Owner and team
Sponsor, process owner, operator representatives, and support functions
Primary measure
Definition, unit, collection method, baseline, and review cadence
Guardrails
Safety, quality, traceability, customer, and regulatory conditions
Decision date
Date to adopt, revise, or stop; no automatic plant-wide rollout
Days 1–15: observe and map the current state
Walk the route with the people who perform and support the work. Follow a real job from release to shipment. Record operations, queue time, setup, batch size, inspection, rework, outside processing, information handoffs, and the reason work waits. Mark actual process data separately from estimates.
A value-stream map is useful when the team needs to see material and information flow together. It is not a replacement for a detailed routing, capacity study, safety review, or customer-specific control plan. Keep one map for one clearly defined family; do not average unlike orders into a fictional “typical” route.
Minimum current-state record for a selected product family
Step or event
Record
Unit or definition
Order release
When the job packet becomes complete and authorized
Date/time; revision and material readiness
Operation
Work content and observed run conditions
Cycle time per piece or batch; setup minutes
Queue
Waiting work and reason
Jobs or pieces; elapsed time; cause category
Quality gate
Acceptance and hold criteria
First-pass accepted pieces ÷ pieces started
Completion
When accepted work leaves the selected boundary
Quantity and date/time
Information flow
Who releases, schedules, revises, and receives the job
Trigger, owner, and delay
Days 16–25: define demand and the baseline
Use the same period and scope for demand and available time. In a simplified example, a shift has 450 planned production minutes after scheduled breaks and a customer requires 90 accepted pieces in that shift. Takt time is 450 minutes ÷ 90 pieces = 5 minutes per piece. This is the pace required to meet that demand; it is not a machine cycle-time target or evidence that the process can safely achieve it.
Define each measure before collecting it. For first-pass yield (FPY), count pieces accepted without rework at the selected step divided by pieces started at that step. For lead time, choose start and finish events and include waits consistently. For work in process (WIP), count pieces or jobs inside the exact same boundary. Record product mix and shift conditions alongside the measure.
Days 26–35: choose one constraint and a test
Compare observed demand with each operation’s available capacity and lost time. A large queue can identify where to investigate, but the queue may reflect missing material, engineering release, a batch decision, or inspection availability rather than machine capacity. Verify the reason with job-level observations.
Write a testable hypothesis: “If we prepare the approved fixture and verified tools before the scheduled changeover, then internal waiting should fall for family A, while first-piece acceptance and safe setup controls remain unchanged.” Identify the change owner, trial jobs, approval needs, primary measure, guardrails, and stop condition. Use a small sample to learn; do not present it as proof of lasting performance.
Days 36–60: test standard work and visual controls
Run the test under representative conditions. Operators record start/stop times and exceptions; the supervisor removes barriers; quality and engineering approve changes that touch controlled requirements. A visual board can show the job, current status, owner, due date, and help needed. It should not contain uncontrolled drawings, customer-sensitive information, or a status that overrides the quality system.
Review exceptions each shift. If a guardrail fails, stop or contain the affected work and follow the approved escalation path. If the method works only on one shift or one job, narrow the claim and continue observing rather than forcing standardization.
Days 61–90: review, standardize, or roll back
Compare baseline and pilot results using the same definitions, part family, and collection method. Show sample size and the range or distribution when available. Check adverse effects such as more rework, shortages, overtime, unsafe handling, or queues transferred to another operation.
If the result is useful and requirements remain satisfied, the authorized process owner updates controlled instructions and trains affected shifts. Set an audit and review date. If performance regresses, return to observation, identify changed conditions, and revise or roll back the countermeasure. A 90-day pilot is a learning cycle, not a guarantee of a specific savings percentage.
90-day roadmap and review board
Phase
Owner
Deliverable
Review question
Week 0
Sponsor and process owner
Approved pilot charter
Is the boundary controlled and the problem measurable?
Days 1–15
Team and operators
Current-state map and observation record
Does the map match actual work and exceptions?
Days 16–25
Planner and process owner
Demand, takt, WIP, lead-time, and FPY definitions
Are units and time windows consistent?
Days 26–35
Cross-functional team
Verified cause and controlled experiment
Can one small change test the cause safely?
Days 36–60
Operators and supervisor
Trial log and exception review
Does the change work across representative jobs?
Days 61–90
Process owner and sponsor
Adopt, adjust, or stop decision
Did the primary measure improve without harming guardrails?
When to pause or change scope
Pause if the team cannot define a safe, quality-compliant condition or if an uncontrolled drawing or process revision appears.
Change the boundary if the selected family has incompatible routes or too little repeat work to support a comparison.
Do not use takt as a command to exceed safe capacity or remove required breaks, inspections, or maintenance.
Do not expand a pilot until the process owner can show the result, limits, training needs, and follow-up plan.
Before observations begin, the sponsor and process owner write a one-page charter. It should state the customer or downstream need, the exact process boundary, product family, start and end dates, current condition, primary measure, safety and quality guardrails, team members, required approvals, and decision date. Name one person accountable for maintaining the data and one person authorized to stop the test. A charter does not authorize changing a drawing, control plan, guarding, or staffing rule.
Copyable Lean pilot charter fields
Field
What to write
Review question
Customer need
Product/service, due date or response, acceptance condition.
Which downstream customer experiences the gap?
Boundary and family
First and last events, included work centers, part family.
Can the team use one route and data definition?
Problem statement
Observed gap, place, time period, sample, and impact.
Does the evidence separate fact from suspected cause?
Primary measure
Formula, unit, data source, owner, and collection interval.
Can another person reproduce the calculation?
Guardrails
Quality, safety, customer, delivery, staffing, and maintenance conditions.
Which condition stops the test?
Change and approvals
One test, affected shifts, approver, training, and rollback.
Who must authorize a controlled change?
Decision date
Review date, sponsor, process owner, and evidence required.
Who adopts, adjusts, or stops the method?
Keep the team small enough to make decisions but broad enough to understand the route. Include operators from affected shifts, a supervisor, and relevant support roles such as quality, maintenance, engineering, materials, or safety. Ask each role to identify constraints and approvals before scheduling a workshop. If the sponsor cannot supply the owner, time, or approval path, pause the project and resolve that governance gap rather than producing an action list nobody can implement.
Days 1–15: observe the current condition
Walk from the customer-facing end back through each operation to order release. Follow one representative job and a contrasting job when routings differ. Record material movement, information release, queue locations, batch sizes, setup events, inspection gates, holds, rework, and outside processing. Ask operators to show what happens when the normal path fails. Confirm the map with the people who perform and schedule the work before treating it as a baseline.
Use a common observation sheet. For each job, record identifier, product family, revision, release and completion events, queue entry and exit, operation start and stop, quantity, disposition, and exception reason. Do not collect unnecessary customer-sensitive details on a public board. Keep controlled drawings and personal information in approved systems. If a record is missing, mark it missing and assign a data correction instead of inventing a timestamp.
At the end of this phase, identify one or two candidate constraints based on actual delay and flow evidence. A large queue is a clue, not proof: late material, incomplete engineering release, batch policy, inspection availability, or a schedule change can create it. Check the mechanism with the process owner and verify whether the condition repeats across representative work.
Days 16–25: lock measurement definitions
Define the primary measure and every denominator. If the team tracks lead time, specify the start and finish events and whether weekends or nonproduction days count. If it tracks first-pass yield, specify which pieces started and what counts as accepted without rework at the selected step. If it tracks WIP, define whether the count includes blocked, held, rework, and offsite jobs; report these categories so the total cannot hide risk.
For the guide’s simplified demand example, 450 planned production minutes divided by 90 required accepted pieces gives a takt time of five minutes per piece. This pace describes demand under the stated assumptions. It is not a target cycle time for every operation. Confirm available time, planned exclusions, demand period, product mix, and quality disposition with the people who own each source. If the data are not trustworthy, repair collection before comparing performance.
Take a baseline long enough to include normal shifts, product mix, and common exceptions. Show the sample size and spread, not only a mean. If orders differ greatly, segment them into families with comparable routes. Write down shutdowns, staffing changes, overtime, supplier disruptions, and other unusual events that affect interpretation. A measurement change must be documented and shown as a break in the series.
Days 26–35: test one verified cause
State the hypothesis in a form that can be checked: “If we stage the approved fixture and verify the current revision before a family-A changeover, then setup waiting should decline while first-piece acceptance and safe setup controls remain unchanged.” Identify the action owner, approved sample, start and end, measure, guardrail, stop condition, and rollback. If the condition is caused by missing material rather than setup, do not test a fixture-staging change simply because it is easy.
Use a small sample to learn whether the mechanism is plausible. Ask operators what part of the change is difficult, and observe rather than relying on a survey alone. Include a check for adverse effects: defect, near miss, extra inspection, queue transfer, overtime, shortage, or repeated workaround. A test that changes several variables at once may be difficult to interpret; separate them when practical.
Before work starts, confirm whether the proposed change affects an approved work instruction, quality plan, machine safeguard, customer requirement, supplier agreement, or regulatory duty. Obtain the responsible approval and train affected people. If approval is denied or evidence is incomplete, return to observation or select a different test that stays within current controls.
Days 36–60: run, review, and learn
Use a daily or shift-level review that records the measure, exception, owner, and next check. Keep the review close to the work and short enough to continue. The supervisor removes barriers and returns decisions; operators record actual conditions; support roles resolve technical and controlled issues. Escalate immediate hazards or product containment through the site’s urgent process instead of waiting for a project meeting.
For every exception, record what happened, whether the test rule was followed, and what changed. Do not exclude a failed trial from the data because it is inconvenient. If the test is unsafe, disrupts a required control, or threatens an order, stop it and follow the rollback plan. If the process works only on one shift or one product, narrow the conclusion and continue learning under an approved scope.
At day 60, compare like periods and product scope. Show raw counts and denominators for the primary metric, plus guardrails and unintended effects. Check whether work moved to another operation or cost center. If the sample is too small or the baseline changed, state that the result is inconclusive and set a further measurement window instead of claiming success.
Days 61–90: standardize, adjust, or stop
If the evidence supports the change and all requirements remain satisfied, the authorized process owner updates the controlled instruction, board, traveler, or system. Train all affected shifts and confirm understanding with observation or demonstration. Assign a backup owner, revision date, audit method, and escalation route. A standard should describe the approved current method and make abnormal conditions visible; it should not lock in a poor method or prevent workers from reporting a problem.
If results regress, inspect changed conditions and the actual process. Reopen the cause when needed. If the method has no sustained value, stop it and remove obsolete instructions or labels through document control. Keep the reason, evidence, and any useful learning in the project record so the next team does not repeat the same experiment without context.
Is the process boundary coherent and owner assigned?
Day 25
Demand, takt, WIP, lead-time, FPY definitions and sample.
Can the team reproduce the measures?
Day 35
Verified cause, charter, approval, training, stop and rollback rules.
Is the test safe, bounded, and informative?
Day 60
Trial data, exceptions, guardrails, comparison with baseline.
Continue, adjust, or stop the experiment?
Day 90
Approved standard, trained shifts, owner, monitoring plan, or rollback record.
Adopt the method, extend the learning, or close the project?
Pause when the project’s conditions change
Pause if an uncontrolled revision appears, safety or quality controls are uncertain, required data cannot be reconciled, or an order is at risk. Change scope if the family has incompatible routes or too little repeat work for comparison. Do not use takt to justify unsafe pace, remove required maintenance, or bypass inspection. Expand only when the process owner can show the result, its limits, training needs, and follow-up evidence.
The 90 days are a planning structure, not a performance guarantee. Some processes need more observation, approval, or seasonal coverage. The responsible owner may extend a phase when evidence is incomplete and document why. A useful project ends with a verified operating decision, not merely with the calendar reaching day 90.
Check readiness before the first workshop
Confirm that the sponsor can release the team for observation, the process owner can access reliable records, and the selected family has enough work to observe. Make a list of required drawings, schedules, inspection criteria, maintenance records, and supplier information. Agree which role can authorize changes and how the team will protect customer and employee information. If one of these conditions is missing, use the first phase to resolve readiness rather than pretending the project can proceed on a fixed date.
At every gate, keep a short decision record with evidence reviewed, unresolved risks, action owners, and date. This prevents a handoff from losing assumptions and gives a new team member a clear starting point. A sponsor can approve added time or resources when needed; the process owner remains accountable for the technical method and ongoing control.
Before closing, ask the receiving owner to explain the new standard, demonstrate the abnormal response, locate the current revision, identify the measure and review date, and name the role to contact when the method fails. If any answer is unclear, complete the handoff before closing the project.
Maintain the control after the 90-day decision
After adoption, schedule an early check and a later sustainability review with the receiving process owner. Confirm that the measure is still collected, exceptions reach the named role, and affected shifts use the current revision. Check at least one representative job at the point of work and compare it with the approved standard. If the process drifts or a new constraint appears, reopen observation and document the next action. A completed project file is not evidence that the operating method remains effective.
Review lessons before expanding to another family
Before a second team copies the pilot, identify which inputs were specific to the first product: route, demand pattern, supplier time, equipment, inspection, and staffing. Reuse the charter format and measurement definitions only where they still fit. The new owner confirms local assumptions, obtains approvals, and records a separate baseline. A control that worked in one family is a candidate for testing elsewhere, not an approved universal standard.
