Managing Factory Maintenance

1. 1Write down your mission-statement answers for the questions nobody is around to answer at night: how much effort goes into quality, how much into safety. A tradesperson alone on a shift needs a reference point, and a poster nobody can recall (the $5,000 vehicle-manufacturer poster) is not one. Test it by asking each person on the crew to state the critical items without looking.
2. 2Go out onto the plant floor looking for impending problems instead of waiting for a call. Run the proactive activities the source names: inspection, cleaning, tightening, lubrication. Do them on the PM clock (for example, every 90 days) or on a usage clock (pieces, machine hours, cycles) for irregularly used assets.
3. 3Decide in advance that you will take equipment out of service for repair before it breaks. The reactive rule is 'if it ain't broke, don't fix it'; the proactive rule is 'don't let it break down, period.' Agree the authority to pull an asset with the operations customer before you need it.
4. 4Keep doing the proactive work without feedback or praise. Expect it to look irrational to outsiders and do not build a long justification for each task. Self-discipline is the toughest part, so make the task list, not the mood of the day, the trigger.
5. 5On every breakdown, stay with it until you understand the root cause and have repaired it, not just swapped the part. This first stage separates the real mechanic from the parts changer. Take the study time; the repeat repairs cost more than the study.
6. 6Do the second stage when it is within reach: re-engineer the system so that failure mode cannot recur. Escalate to engineering where the system is too complex for the mechanic. Where you can, spend on maintenance prevention, and pursue eliminating the need for maintenance as the most proactive stance.
7. 7Treat the deterioration you already have as an unfunded maintenance liability. Use PM and PdM to detect the 'tail' (the deterioration already under way) and estimate time to failure. Let PM discipline force a corrective action to repair, rebuild or replace. Put past-neglect costs in a separate budget line.
8. 8Hold the course for years. Deterioration moves slowly and the shift from reactive to preventive to predictive to world class is a multiyear effort. Look, record and document deterioration over time while planning a well-conceived intervention, and resist each new 'flavor of the month' program.

The source's sewer-line case shows the difference between firefighting and root cause. A crew snakes out an old cast or galvanized sewer line every month and is never confident it will stay open. A parts changer keeps scheduling the next snaking. A proactive mechanic studies the slow drainage and finds the root cause: hardened buildup inside the pipes. The analysis also estimates the cost of leaving the defective pipe in place, so the decision between continued snaking and replacement rests on numbers. The source adds that some problems should not be fixed, so the cost estimate decides, not reflex. The same logic applies to a crew moving from reactive to PM work. Many emergencies persist at first because of unfunded maintenance liabilities, so the proactive discipline is to keep the PM work going while funding the backlog separately.

• If a breakdown is repaired but you cannot state the underlying cause, then it is not closed; keep working or log it as an open root-cause item.
• If a repair has been done on the same problem repeatedly, then stop repairing and put time into studying the system, because the organization is already paying for the time it claims it lacks.
• If inspection or PdM detects deterioration on a critical asset, then schedule it out of service for repair before failure rather than waiting for it to break.
• If the root-cause fix needs re-engineering beyond the mechanic's ability, then escalate it to someone with the expertise rather than leave it as a recurring repair.
• If emergencies keep swamping a new PM program, then check for unfunded liabilities and raise a separate budget line for past neglect instead of abandoning PM.

• Parts changing: the failed part is replaced and the job closed. Tell: the same failure reappears and rework or call-backs rise.
• Reactive attraction ("excitement junkies"): crews prefer the variety of emergency work and drift away from PM. Tell: PM tasks are skipped when a breakdown call comes in, and people describe PM as boring.
• Proactivity that collapses without feedback: PM work stops when nobody is watching or praising it. Tell: PM completion falls on night shifts or when the supervisor is away.
• Flavor-of-the-month management: focus switches each year to empowerment, customer proximity, then quality. Tell: staff cannot say what the current priority is and treat each program as temporary.
• Premature decisions forced on the mechanic: the organization imagines there is no time for study. Tell: the same problem is repaired again and again while 'no time' is cited for analysis.

• ▸Asset / location: identify the asset and where it sits.
• ▸Trigger: PM clock (days, pieces, machine hours, cycles), inspection finding, or breakdown.
• ▸Impending problem observed: what you saw on the floor before failure (wear, leak, looseness, hot spot).
• ▸Action taken before failure: inspection, cleaning, tightening, lubrication, or planned removal from service.
• ▸Operator/customer informed of outage: who agreed and when.
• ▸Root cause (stage 1): the underlying cause you confirmed, not just the part replaced.
• ▸Repair verified as permanent? Yes/No, and the evidence.
• ▸Re-engineering needed (stage 2): design change to remove the failure mode, and who owns it.
• ▸Cost of leaving it in place: estimated cost if you do nothing, to decide fix or accept.

1. 1Size the gap first. List every PM you should do on an asset and compare it to the hours your maintenance people have. The source says that if you did all the PMs you should, you would find more maintenance than maintenance people. That gap is your case for using operators as the additional resource.
2. 2Pick the tasks operators can own. Start with the basic task-list verbs: inspect, clean, tighten, adjust, lubricate. Inspectors can be machine operators, so assign these tasks to the operator of the asset. Keep interruptive tasks, where maintenance must take custody of the asset, with maintenance.
3. 3Train and test before you hand anything over. The source says the operator is 'a great asset' once properly trained. For each task, write what to look for and what to do, show it, and confirm the operator can do it.
4. 4Form autonomous maintenance groups around the equipment. Operators in independent groups carry out routine maintenance and PMs and solve problems without management intervention. Maintenance is called for bigger problems that need more resources, technology or downtime. Write that escalation line down so both sides know it.
5. 5Extend ownership past routine tasks. Where it fits, let operators act as helpers on large repairs or as the safety watch person for confined-space entry. They then see the whole repair, not just the symptom.
6. 6Cross-train your mechanics. The aim is for one person to do more of the whole job so they can feel ownership and the pride of a job well done. Build the training and testing program before you rotate people across crafts. The source says multi-skilling depends heavily on it. Use outside trainers or tech schools if needed.
7. 7Open work-request access widely, but teach it. Letting many people generate requests gives them ownership and a sense of being part of the operation. Build a mini-course (PowerPoint, intranet or live) covering both which buttons to push and why each field matters. Teach people to report what they observe, not what they infer.
8. 8Ask operators and mechanics where the waste is. The source says they are best placed to know, because waste 'drips through their fingers'. Feed what they report into the continuous improvement cycle: commitment, measurement, information gathering, investigation, action.

A blending-system call shows why wide request access needs training. A work request described a broken main computer. The technician took the description at face value and was misled for several hours. The system was acting exactly as designed. The second product's pump was not activating because the source tank for the first product was empty, and the gauging system also looked out of range. The source's point is that good requests carry only the observable, not the inferred. Applied here, the requester should write what the display and pumps were doing, not a diagnosis of the computer. The ownership move is not to take request-writing away from operators. Keep access wide, because it makes people feel part of the operation, and run the mini-course so each request is something a technician can act on. The same logic covers PM handover. The trucker's pre-trip inspection is the model: operators do part of the checking, but only against a defined task, and it is required by law in trucking and aviation.

• If PM workload exceeds maintenance capacity, then move basic inspect/clean/lubricate/tighten tasks to the operators of the asset, and only after training.
• If a task interrupts the machine or needs maintenance to take custody of the asset, then keep it with maintenance. Autonomous groups escalate problems that need more resources, technology or downtime.
• If a craft is understaffed while the department has enough people overall, then cross-train rather than hire for the narrow craft. This also cuts hand-offs between crafts.
• If you widen work-request access, then ship the mini-course with it. Wide access without field training produces vague requests and wild goose chases.
• If a skilled worker is retiring, then use that person's knowledge in the training of operators and mechanics instead of letting it leave with them. The source treats loss of old-timers as a core threat.

• Handing PM tasks to operators without training or testing. The tell: skipped checks, wrong lubrication, and maintenance redoing the work.
• Treating the operator as a button pusher or gauge watcher. The tell: operators report problems only after a breakdown, and small problems stay unaddressed.
• Cross-training with no training and testing program. The tell: people are rotated across crafts but still cannot finish a whole job, so pride and ownership do not appear.
• Routing every report through the supervisor, who decides whether to pass it along. The tell: few people file requests and operators stay out of the operation.
• Opening request access with no field guidance. The tell: requests carry inferred diagnoses like 'main computer broken', and technicians lose hours chasing them.

• ▸Asset / unit and operator group: name the machine and the operators who will own the routine care.
• ▸PM load vs capacity: list PM hours required against maintenance hours available, and note the gap.
• ▸Operator-owned tasks: list tasks from inspect, clean, tighten, adjust, lubricate, with one line of instruction each.
• ▸Training and test: record who trains, the date, and the pass check for each task.
• ▸Escalation line: state which problems go to maintenance (more resources, technology or downtime) and who is called.
• ▸Extended roles: note whether operators help on large repairs or act as safety watch, and for which jobs.
• ▸Mechanic cross-training: record the second craft skill being added per mechanic and the test that confirms it.
• ▸Work-request mini-course: record the date delivered and a check that requests state observable facts only.
• ▸Waste found by the team: record items reported by operators or mechanics, and the improvement action assigned.

1. 1Treat PM, TPM, RCM, and PMO as process aids, not as the process itself — first make the underlying process (work request → work order → planning → completion) whole and complete before you bolt on any initialed program.
2. 2Map the information flow explicitly: trace how a work request becomes a work order and how job completion feeds back. The supervisor and manager live or die by the quality and timeliness of this information, so pin down every hand-off point.
3. 3Separate maintenance work into its three distinct types and crew for them separately, because each type demands different skills and personalities; do not staff one team to do all three.
4. 4Plan larger jobs before dispatch — build the planned job package using the questions-before-planning list, since you save 3-5 hours of execution for every hour spent planning.
5. 5Derive PM task content from the failure behavior of the component (the performance-failure curve and MTBF distributions), not from calendar habit — the task must address the actual deterioration mode.
6. 6Balance PM: use the PM balance and total-cost-of-maintenance view so added PM keeps reducing total cost rather than crossing into over-maintenance.
7. 7Institutionalize improvement so it is a core part of everyone's day rather than dependent on one motivated person; a personal commitment 'will not have enough overall effect to make a difference.'

Take an air cylinder analyzed across the three strategies in the passages. Under 'bust'n'fix' each cylinder deteriorates and fails at its own point, and with many cylinders the failures spread into a bell-shaped distribution — the whole area under the curve is the whole population, and eventually they all break. PM shifts that distribution by intervening before failure, extending average MTBF. Permanent maintenance improvement (re-engineering the component or its application) shifts it further still and is why re-engineering and maintenance improvements account for 39% of continuous-improvement dollars in the passage's breakdown. The process-quality point is that you don't pick PM because it feels proactive; you pick the strategy that fits the component's failure distribution — 'strategies must fit the circumstances.' Building the PM task list, the planned package, and the completion feedback around that distribution is what makes the result repeatable instead of heroic.

• If the underlying request-to-completion process is broken, then fix the process first — do not deploy PM/TPM/RCM/PMO, because 'the underlying processes must be whole and complete for the aids to help us.'
• If a job is large, then plan it before dispatch, because you recover 3-5 hours of work for each planning hour.
• If improvement depends on one person noticing an opportunity, then treat it as insufficient and make improvement a daily core activity of the whole department.
• If a chosen strategy does not fit the component's failure circumstances, then change the strategy — 'strategies must fit the circumstances.'
• If the work spans three distinct types, then assign three teams matched to the required skills and personalities rather than a single crew.

• Mistaking initials for management: believing that having PM, TPM, RCM, or PMO means you are doing maintenance management. The tell — programs exist on paper but request/work-order/completion information is still late or missing.
• Relying on individual heroism: improvement happens only when one person champions it. The tell — gains evaporate when that person moves on, and no department-wide measurement exists.
• Adding PM without balance: piling on tasks past the point where they reduce total cost. The tell — total cost of maintenance rises while breakdowns don't fall (the PM-balance point crossed).
• Applying a strategy that doesn't fit the failure mode: forcing calendar PM on a component that fails randomly. The tell — MTBF doesn't improve despite the added tasks.
• Skipping planning to 'save time': dispatching large jobs unplanned. The tell — crews wait for parts, information, or access and burn the 3-5 hours planning would have saved.

• ▸Asset / component: name the item whose maintenance process you are assessing.
• ▸Failure distribution / MTBF basis: state what deterioration behavior the tasks are built on.
• ▸Work type (of the three): identify which type this work is and the crew assigned to it.
• ▸Information flow check: confirm request → work order → completion feedback is complete and timely.
• ▸Chosen strategy & fit reason: name bust'n'fix / PM / permanent improvement and why it fits the circumstances.
• ▸Planning done? (Y/N) + hours: record planning hours and the estimated execution hours saved.
• ▸PM balance position: note whether added PM still lowers total cost or has crossed the balance point.
• ▸Aid in use (PM/TPM/RCM/PMO): list any process aid and confirm the underlying process is whole first.
• ▸Improvement ownership: name who owns the improvement — department/plant, not a single individual.

1. 1Pick the assets first. Reliability work is directed at failure modes with the worst consequences (the RCM logic). Rank assets by what a breakdown costs in safety, downtime, damage and repair. Deliberately park low-consequence equipment as BNF (Bust 'N Fix, no PM) so effort goes where it pays.
2. 2Identify each asset's life cycle stage: start-up, wealth or breakdown. The PM strategy differs by stage, so do not apply one standard task list to everything.
3. 3Set the baseline from work-order history. For each critical asset, record breakdown events, the reason for repair, and the reason for write-up. Use these to calculate MTBF (reliability) and MTTR (cost of a typical failure). Count call backs and rework as failures of the original job, not as new, unrelated events.
4. 4Place the asset on the A/B/C curve. Curve A is life with little attention, set by design, build and operating conditions. Curve B is the longer life PM and predictive maintenance buy: cleaning, keeping everything tight, correct lubrication. Curve C is life that is longer even when left unattended. Write down which curve each asset sits on and what spend holds it there.
5. 5Build the PM and PdM task list to hit the failure mode. Choose the PM clock (days, pieces, machine hours or cycles) that matches how the asset is used. Add predictive inspection (vibration, oil analysis, infrared) where it can flag a failure before it occurs. Set starting frequencies, then adjust them with the failure data.
6. 6Use MTBF to schedule Planned Component Replacement (PCR). Replace components on a schedule based on MTBF, downtime costs and other factors, instead of waiting for the breakdown.
7. 7For every repeat failure, do root cause analysis and design a repair that prevents it (maintainability improvement). The aim is Curve C: more MTBF without pouring money in every month. First estimate the cost of the fix against the cost of the problem. Do not spend $50,000 to avoid a $100 problem.
8. 8Track MTBF and MTTR per asset on a fixed cycle. Protect the PM budget while you do. A cut to PM does not show up in the failure rate until the curve has decayed back to A, so a flat month after a cut is not evidence the cut was safe.

The source's cylinder example shows the mechanism. An asset with cylinders left alone follows Curve A: its MTBF is set by how it is used, how well it is engineered and built, and the conditions it runs in. Cleaning the cylinders, keeping everything tight and adding the right amount of lubrication shifts the life curve to the right (Curve B), giving a longer average life. That shift lasts only while PM labor and materials keep flowing. If a new manager cuts PM to look good, nothing happens to the failure rate until enough time passes for the curve to decay back to A. Profit rises in the short term, and the next manager inherits the breakdowns. The durable fix is Curve C: better seal kits, better cylinders and better protection, so the unattended life exceeds both A and B. To use this, record the cylinder's failures before PM, after PM, and after the redesign. Then check whether the MTBF gain survives without added monthly spend. The test is that the improvement must cost less than the failures it removes.

• If an asset's failure has severe safety, damage or downtime consequences, then put PM, PdM and redesign effort there first. If consequences are minor and PM does not pay, then leave it as BNF on purpose.
• If MTBF holds only while PM money keeps flowing (Curve B), then keep the PM funded and look for a redesign that moves the asset toward Curve C.
• If a job produces call backs or rework, then count it as a reliability failure and investigate the materials, skills or scope of the original job before adding PM.
• If a failure repeats, then do root cause analysis instead of repeating the repair. Like snaking a drain every month without removing the hardened buildup, repeating the repair treats the symptom.
• If a proposed improvement costs more than the failures it prevents, then do not do it. Test it against the cost of the problem first.

• Cutting PM to lift short-term profit. The tell: no change in breakdowns for months, then a rising failure rate once the curve decays back to A, often after the manager who cut it has moved on.
• Treating PM as the finish line. The tell: MTBF is stable only while labor and materials keep flowing, and no maintenance-improvement item appears on any asset's plan.
• Fixing the symptom repeatedly. The tell: the same asset appears in the breakdown history on a regular rhythm, with repairs but no root cause study.
• Spreading PM evenly across all equipment. The tell: heavy PM hours on low-consequence assets while high-consequence failure modes still produce breakdowns.
• Over-engineering the fix. The tell: a redesign whose cost far exceeds the cost of the failure it targets, the $50,000-to-avoid-a-$100-problem case.

• ▸Asset and life cycle stage: name the asset and mark start-up, wealth or breakdown.
• ▸Failure consequence: record safety, downtime, damage and repair cost of a typical breakdown, to decide PM, PdM or BNF.
• ▸Breakdown history: list the dates and reasons for repair from work orders, and flag call backs and rework.
• ▸MTBF and MTTR (baseline): enter both calculated from the history, with the period used.
• ▸Current curve position (A/B/C): state which curve the asset is on and the ongoing PM/PdM spend holding it there.
• ▸PM/PdM tasks and PM clock: list the tasks, the clock (days, pieces, hours, cycles) and the starting frequency.
• ▸PCR trigger: name the component, its MTBF-based replacement interval and the downtime cost it protects against.
• ▸Root cause and improvement: state the root cause of repeat failures, the proposed redesign and its cost against the cost of the problem.
• ▸Review result: after the next cycle, record the new MTBF and MTTR and whether the gain holds without added spend.

1. 1Define the ideal first. Per the source, OEE compares effective output with the ideal output possible from the same plant, area, or machine. For each machine or area, write down the nameplate speed and the planned running time. Without that denominator, every OEE number is arguable.
2. 2Split the gap into the three parts of the definition: availability (downtime), performance efficiency (running below nameplate speed, small jam-ups), and quality rate (defects). Record each as a percentage so the product is the OEE. Do not report one blended number.
3. 3Log every loss by cause using the source's own categories: breakdowns, model changes, material problems, small jam-ups, machines not run to nameplate speed. The source says the 20-50% that is not OEE is wasted resources, so each point needs a named cause.
4. 4Size the prize before you spend. The source says a typical factory sits at 50-80% OEE, and most plants could recover about half of the remaining gap (10-25% more output) without expansion capital. Express your gap as extra output at no new asset cost, then compare it with the maintenance budget.
5. 5Rank the loss categories by points of OEE lost, not by maintenance hours spent. The source warns that cutting maintenance cost is admirable, but the real money is in the OEE of the whole plant.
6. 6Assign each top loss an owner. Downtime and small jam-ups go to maintenance with operations. Model changes and material problems go to production and supply. Bring in operators as the TPM approach in the source describes (autonomous maintenance groups, the six production losses).
7. 7Take it to management as availability and output. The source says what bosses really want is 100% availability, safely and legally. Report OEE movement in those terms, and trend the three components monthly.
8. 8Protect the gain. Because the maintenance department is smaller and has less room for mistakes, check that recovered availability is not being paid for with deferred PM or lost expertise.

The source gives two anchors. First, the utility example: a Midwest generating station was designed for 85-95% availability. To support that, designers met with maintenance leadership and discussed every failure mode, planned for it, and designed it out where possible. The capital spares inventory was $24 million, since reduced by maintenance to $15 million. Availability was treated as a design target tied to maintenance input, not an afterthought. Second, the source's general OEE range: a typical factory runs 50-80% OEE, and most could add 10-25% output by recovering about half the waste. Illustration (the 60% figure is hypothetical, not from the source): a line scores 60% OEE, so 40 points are wasted. You log the 40 points to breakdowns, model changes, material problems, and small jam-ups. Recovering half of that waste, 20 points, takes the line to 80%, with no new machines. You then work the largest-loss categories first and report the result as added output, which is the language management uses.

• If OEE is 50-80% and the loss log is not categorized, then do not start a cost-cutting project. Categorize the waste first, because the source puts the real money in OEE, not in maintenance cost.
• If a loss is small jam-ups or running below nameplate speed, then treat it as a performance loss owned jointly with operators, even though no machine is down.
• If breakdowns dominate the loss log, then move effort to PM, predictive work, and design-out of failure modes, as in the utility example.
• If model changes or material problems dominate, then escalate to production and supply. Maintenance cannot recover those points alone.
• If management asks for a maintenance cost cut, then show the OEE points at stake against the dollars saved before agreeing.

• Reporting availability alone and calling it OEE. The tell: availability is high but output per hour is still below nameplate, with jams and slow running unlogged.
• Using a flattering ideal (a demonstrated rate instead of nameplate). The tell: OEE reads at the top of the 50-80% range but the plant still cannot meet demand.
• Chasing maintenance cost instead of plant output. The tell: the maintenance budget shrinks while breakdown hours and lost production rise.
• Counting only breakdowns and ignoring model changes, material problems, and small jam-ups. The tell: the loss log explains far fewer points than the 20-50% gap.
• Treating OEE as a maintenance-only metric. The tell: production does not attend loss reviews and no loss has a non-maintenance owner.

• ▸Machine/area and period: name the asset and the date range measured.
• ▸Ideal output: nameplate speed times planned running time, stated in units.
• ▸Availability %: running time over planned time, with the downtime hours behind it.
• ▸Performance %: actual speed against nameplate speed, with logged jam-ups and slow-running causes.
• ▸Quality %: good units over total units, with the defect causes.
• ▸OEE % (A x P x Q) and wasted points: the product, and 100 minus that figure.
• ▸Top three loss causes with points lost: choose from breakdowns, model changes, material problems, small jam-ups, below-nameplate speed.
• ▸Owner and action per cause: a named person and the next step, with a due date.
• ▸Recoverable target: roughly half the gap, expressed as extra output at no new asset cost.

1. 1Define the cost boundary before you count anything. Use the glossary's life cycle cost list: PM, repair labor, parts, supplies, outside services, rentals, downtime, energy, ownership, overhead. Total cost of maintenance is the sum of all of these, not the maintenance budget line alone. Adjust for the time value of money if the horizon is long.
2. 2Split the cost into two opposing streams for each asset or asset class. Stream A is the cost of PM activity: inspector labor, PM materials, and production time lost to interruptive tasks. Stream B is the cost of breakdowns: repair labor and parts, outside services, rentals, and downtime. The balanced PM level sits where A + B is lowest, so you need both streams side by side.
3. 3Price a typical failure with MTTR and MTBF. MTTR gives the repair duration and cost of a typical failure, and MTBF gives how often it occurs. Multiply cost per failure by failures per period to get Stream B. Add the consequences of a breakdown (Figure 33), since downtime usually outweighs the repair itself.
4. 4Place each asset in its life cycle stage: start-up, wealth, or breakdown. The right PM intensity differs by stage, so do not apply one PM level to the whole plant. Use the asset's critical wear point to decide where PM stops being useful and Planned Component Replacement (PCR) takes over.
5. 5Find where Stream A is wasted. Separate non-interruptive tasks, which run without stopping production, from interruptive ones, which require maintenance to take custody of the asset. Interruptive tasks that find nothing are the clearest sign of over-PM. Also look for failures caused by your own service person, which add Stream B cost from PM that was done badly.
6. 6Build the cost from the bottom up. Use the zero-based budget logic: look everywhere maintenance resources are used and build the total from those events, because maintenance is the sum of thousands of small events. This exposes hidden Stream A and Stream B cost that a departmental total hides.
7. 7Change PM frequency or content in one direction, then re-measure. Raise PM where breakdown cost dominates and cut it where PM finds little. Track breakdown frequency, MTTR, and PM hours. You are tracing the Figure 37 curve with your own data, so move one variable at a time.
8. 8Review against output, not only cost. Use OEE as the check: if total maintenance cost falls but OEE falls with it, you have moved to the under-PM side of the balance.

The source gives no dollar figures for the cost curve, so this example uses its OEE numbers to show the Stream B side. A typical factory has an OEE of 50% to 80%, so 20% to 50% of potential output is wasted on breakdowns, model changes, material problems, and small jam-ups. Take a plant at 60% OEE (an illustrative pick inside the stated range), which wastes 40%. The source says most plants could recover about half of the waste without spending on expansion, which is about 20 points of output here, or 10% to 25% across the stated range. That recovered output is the value of lower downtime, which sits on the breakdown side of the cost curve. Next, price one recurring failure: MTTR times the labor and parts rate, plus lost production, multiplied by its MTBF-based frequency. Compare that figure with the PM hours and PM-caused interruptions spent trying to prevent it. If added PM hours cost less than the breakdown cost they remove, you are on the under-PM side of the balance. If added hours stop removing failures, you have passed the minimum.

• If breakdown frequency or downtime is high and PM tasks are few, then add PM first, targeted at the failures with the highest MTTR cost.
• If PM tasks, especially interruptive ones, rarely find anything, then lengthen the interval or convert the task to non-interruptive or predictive inspection.
• If an asset is near its critical wear point, then switch from inspection PM to Planned Component Replacement, scheduled from MTBF.
• If total maintenance spend falls but OEE or downtime worsens, then treat the savings as false and restore PM before judging the cut a success.
• If you cannot attribute costs to individual events or assets, then build a zero-based view before changing PM levels, because you cannot find the minimum without the data.

• Counting only the maintenance budget. Tell: downtime, ownership, and overhead never appear in the cost comparison, so cutting PM looks like a pure saving.
• Treating more PM as always better. Tell: PM hours rise every year, interruptive tasks find little, and breakdowns do not fall.
• Using one PM level for all assets regardless of life cycle stage. Tell: new equipment is over-inspected while worn-out equipment still fails between inspections.
• Ignoring failures induced by the service person. Tell: breakdowns cluster shortly after PM or repair visits.
• Pricing a failure at repair cost only. Tell: MTTR cost is quoted in labor and parts, with no production loss or consequences from the breakdown.

• ▸Asset / class and life cycle stage: write the name and mark start-up, wealth, or breakdown.
• ▸Period and PM level tested: state the interval and the task list in force for this period.
• ▸Stream A, PM cost: inspector labor + PM materials + production time lost to interruptive tasks.
• ▸Stream B, breakdown cost: failures in the period × (repair labor + parts + supplies + outside services + rentals).
• ▸Downtime cost: failure hours × lost output value, using MTTR for typical failures.
• ▸Other ownership costs: energy, overhead, and any time-value adjustment.
• ▸MTBF and MTTR: record both, and note the trend against the prior period.
• ▸Total cost (A + B + downtime + other): compare with the prior PM level.
• ▸Decision: raise PM, hold, lower PM, or convert to PCR, with one-line reason and the next review date.

1. 1Define output units first. Pick the unit your plant sells or produces (cars assembled, cases of soda bottled, barrels of oil refined) and record maintenance effort and parts consumed per unit. This is the unit-cost baseline that competitiveness is judged against.
2. 2Set last year's numbers as the benchmark to beat. Track completed work orders versus incoming work orders, PM hours to total maintenance hours, percentage and quantity of emergency work, callbacks, downtime with reason codes, production quantity/quality, and maintenance cost per product shipped. Each trends toward lower cost, higher availability, or higher quality.
3. 3Find the waste with the people closest to it. Ask machine operators and maintenance workers where waste is dripping. Sort the findings against the Canadian maintenance provider's study: 39% re-engineering and maintenance improvements, 26% PM improvement and correct application of PM, 27% more extensive predictive maintenance, 7% storeroom improvements.
4. 4Price each decision on life-of-unit cost, not purchase price or immediate repair cost. Compare pothole repair with resurfacing (resurfacing pays above one pothole per 1000 sq ft and lasts about 6 years). Compare direct-drive with belt-driven exhaust fans (cheaper to buy, quieter, less maintenance in new installs).
5. 5Give maintenance the financial and production data it needs to decide: downtime costs, overhead costs, fixed-asset accounting methods, equipment retirement budgets, interdepartmental priorities. Without these, a correct maintenance decision can still work against the organization.
6. 6Tie availability to a stated target. Compare actual uptime with the design specification (the Midwest generating station was specified at 85-95% availability) and use the gap to educate top management on what maintenance can and cannot deliver.
7. 7Fund training and R&D for maintenance as a standing line. Benchmarks from the passages are 96 hours per person per year at an automobile plant and 5% of direct hours (104 hours) at a Midwest high-tech manufacturer. Re-test each year's input-per-unit and reset the benchmark.
8. 8Treat the competitor threat as the review frame. Each year ask whether you can make the same output with less input, or more output with the same input. If not, expect outsourcing pressure or loss of market share.

A Maryland plastics extruder had a maintenance worker use a new vibration analysis tool and find a problem in the gear driving the main extruder screw. Repairing it immediately cost $500; after breakdown the bill would jump to $5000, so he claimed a $4500 saving and was proud of it. He took the plan to the shift manager, and the passage warns that a good maintenance decision can still work against the organization when maintenance lacks production and financial data, such as the cost of downtime at the moment of repair. The lesson is that competitiveness is measured at the level of unit cost and availability, not repair cost alone. Before scheduling the repair, price the stoppage against production and schedule it where it costs the plant least. The $4500 figure covers only repair cost and ignores that timing.

• If a repair or purchase decision is under review, then compare total life-of-unit cost (purchase, maintenance input, durability, noise) rather than first cost; resurfacing beats potholes above one pothole per 1000 sq ft.
• If a new installation is being specified, then choose the lower-maintenance option (direct-drive fans); if it is an existing unit, then do not replace unless extensive other work is needed.
• If maintenance lacks downtime costs, overhead costs, or production priorities, then obtain them before approving the maintenance decision; do not rely on repair cost savings alone.
• If input per unit of output has not dropped versus last year's benchmark, then treat the department as stagnating, not stable, and open an improvement project.
• If a team feels satisfied because it understands its failure modes, then treat that as a warning sign and force a new improvement target.

• Counting repair savings only (the $4500 claim) and ignoring downtime and production cost. Tell: maintenance savings reports with no production or availability figures beside them.
• Buying technology to fix a basic practice problem. Tell: a new computer or scanner is proposed while PM, planning, and work-order basics are weak, and the passage notes most good maintenance practices are basic and low tech.
• Letting maintenance stay out of the information loop. Tell: the department cannot state downtime cost, overhead, or retirement budgets for its own equipment.
• Complacency after mastering known failure modes. Tell: benchmarks are unchanged year over year and the team says it can handle anything.
• Cutting training because it takes time away from the floor. Tell: no annual training-hours target exists, while benchmark firms invest 96 to 104 hours per person.

• ▸Output unit: the unit you produce and count (cars, cases, barrels).
• ▸Maintenance input per unit this year: labor hours plus parts divided by output units.
• ▸Last year's input per unit (benchmark to beat): record it and compute the change.
• ▸Availability/uptime: actual versus design specification, with the top downtime reasons.
• ▸Quality indicators: callbacks, emergency work percentage, production quality.
• ▸Planned-versus-emergency split: PM hours to total maintenance hours.
• ▸Top three waste sources named by operators and maintenance workers: map each to re-engineering, PM, predictive, or storeroom.
• ▸Training hours per person versus target (96 or 5% of direct hours): note the gap.
• ▸Next-year target: same output with less input, or more output with the same input; state the number.

1. 1Teach top management the reality that can be taught. The exact minute a gear cracks is unknowable, but machines do break down if they are not serviced. Make education of top management an explicit mission of the maintenance department, and brief executives on failure modes the way the finance VP briefs them on risk.
2. 2Write the case as a long horizon. Deterioration moves at a slow, steady pace and long-decay assets respond to a constant approach. Reframe the ask as a multiyear climb from reactive to preventive to predictive to World Class, and say plainly that it is a multiyear effort.
3. 3Request a five- or even ten-year maintenance and rehabilitation budget, not a one-year number. Use it to show the corporation future needs and to protect resources from being taken away with every bump in quarterly results.
4. 4Set patience expectations in writing before work starts. Tell managers used to monthly results that substantial system and procedure improvements require time and investment, and that initial results often take a year or more. Agree on the review dates that fit that span.
5. 5Get downtime and resources committed, not only money. The utility example works because designers and maintenance leadership met in detail, discussed every failure mode, and planned for or designed out each one. Ask for that same early access and the capital spares that protect against excessive downtime costs.
6. 6Anchor the effort in a mission statement staff can recite. Check it with a quick test: ask supervisors whether the company has a written mission and what it says. A statement nobody can remember guides nothing.
7. 7Block flavor-of-the-month swaps. When a new initiative arrives, ask what happens to the maintenance agenda and budget. A moving agenda for maintenance goals works against the department.
8. 8Treat the core as stable and the tactics as flexible. Follow Covey's point that an unchangeable core lets the organization react to the marketplace, and name the way you treat people and assets as one of those core values.

A power utility builds a generating station in the Midwest with a design specification of 85-95% availability. To support it, the designers hold long, detailed meetings with the maintenance leadership. Every failure mode is discussed, looked at, planned for, and where possible designed out. The capital spares inventory is $24 million, which the maintenance department has just reduced to $15 million. Management funds a large spares position because it understands that profit in power generation depends on maintenance input to the design and operation of the plant. Maintenance is consulted at design time, trusted with a large inventory, and credited when it trims that inventory. That is top-management understanding, funding, and constancy of purpose in practice. Use it as the model: seek the design-stage seat, bring failure-mode detail, and show the dollar link between spares and downtime.

• If management asks for a short-term cut, then show the long-decay consequence and offer a ten-year rehabilitation budget view rather than arguing one year's number.
• If executives measure only monthly results, then agree up front that initial improvements often take a year or more, and set milestones on that timeline.
• If a new management initiative arrives (empowerment, customer closeness, quality), then fold maintenance goals into it without dropping the core agenda, because a flavor-of-the-month pattern is what destroys effectiveness.
• If staff cannot state the mission statement, then treat it as fake and rewrite it in concrete terms covering quality, safety, downtime, and improvement.
• If a failure's exact timing is unknowable, then still teach management the knowable part: unserviced machines do break down.

• Pitching maintenance as a cost line only. The tell: executives respond with cost-cutting targets and not with questions about availability or downtime.
• Accepting a one-year budget horizon. The tell: resources are taken away with every bump in the quarterly results and the PM or predictive effort stalls.
• Chasing each new management fad. The tell: last year's initiative is gone, nobody can say why, and maintenance goals keep moving.
• Expecting results in a quarter. The tell: managers lose patience before the year-plus payoff and cancel the improvement.
• Posting a showpiece mission statement. The tell: an expensive poster hangs in the conference room but half your supervisors say there is no statement and none can recall it.

• ▸Asset or system and its decay cycle: describe how slowly it deteriorates and what it supports.
• ▸Current maintenance level: reactive, preventive, predictive, or World Class.
• ▸Target level and multiyear timeline: state the climb and the years it takes.
• ▸Five-to-ten-year budget: list maintenance and rehabilitation needs by year.
• ▸Downtime and resources requested: windows, spares, and staff you need committed.
• ▸Known failure modes and planned response: list what is planned for or designed out.
• ▸Expected first results date: state that results often take a year or more.
• ▸Mission statement test result: record how many supervisors can recall it.
• ▸Initiative conflicts: list competing programs and how maintenance goals stay intact.

1. 1Build the failure list from your own history. Pull work orders by 'reason for write-up' (PM, breakdown, vandalism, capital improvement) and flag breakdowns and rework. Rework signals a problem with materials, skills, or scope. Callbacks and repeat breakdowns show where you are still in bust-and-fix.
2. 2Rank failure modes by consequence, not by noise. Use the three tests in the definition: most dangerous, most expensive, most common. Safety outranks everything else, as in the source's example that a safety problem is fixed before an energy-improvement job. This is the RCM move: PM tasks and redesign go to the failure modes with the worst consequences.
3. 3Ask 'why' on each top-ranked mode before choosing a task. The source's sewer line shows the trap: you can snake it monthly and never be confident it stays open. The root cause is hardened buildup in the pipe. Study the problem, then estimate the cost of leaving the defective pipe in place. The source also notes that some problems should not be fixed.
4. 4Place each asset in its life cycle. The source treats PM as needing a different strategy for each of three equipment life cycles, and it describes a critical wear point and infantile mortality (early failures). Put new equipment under inspection and testing. Percentages of new steam traps clog or fail in the first year, so early inspection catches them. Put mature plants toward PMO, which the source says suits them.
5. 5Match the technique to the failure mode. Use PM (inspection, cleaning, tightening, lubrication) where the failure responds to routine care. Use predictive inspection where technology can see the failure coming, such as infrared surveys for hot spots in electrical distribution or vibration analysis. Use redesign or design-out where the mode is chronic. Use TPM autonomous maintenance and its six production losses where operators can carry the routine care.
6. 6Set the PM clock and frequency. Use calendar days (for example, every 90 days) for assets in regular use. For irregularly used assets, use pieces, machine hours, or cycles. The clock starts the task list, and frequency follows from the clock.
7. 7Write down the trigger to take equipment out of service before it breaks. The proactive stance depends on the will to do this. 'Don't let it break down, period' replaces 'if it ain't broke, don't fix it'.
8. 8Review the strategy in a standing meeting. The utility example discussed every failure mode, planned for it, and designed it out where possible. Record what changed, and keep the long horizon. Resources for the strategy should not be pulled with every bump in quarterly results.

The source describes a utility that holds detailed meetings with maintenance leadership. Every failure mode is discussed, looked at, planned for, and, where possible, designed out. The utility treats profit in power generation as depending on maintenance input to design and operation. In this case the capital spares inventory was $24 million, and maintenance has just reduced it to $15 million. That outcome follows from strategy fitted to consequence: modes are understood and planned for, so spares are bought for known risks rather than as general insurance. A smaller example from the source's glossary shows the same logic at pipe scale. A crew snakes an old cast or galvanized sewer line every month, which is reactive work repeated forever. The root cause analysis finds hardened buildup inside the pipe. The study estimates the cost of leaving the pipe in place against the cost of replacing it, and the decision follows the numbers. In both cases the action is chosen after the failure mode and its consequence are understood.

• If a failure mode threatens safety, then rank it ahead of cost or energy jobs and assign PM, predictive inspection, or redesign before anything else.
• If the same symptom is repaired repeatedly (the monthly snaked sewer line), then run a root cause analysis and compare the fix cost with the cost of leaving the defect in place. Accept that some problems should not be fixed.
• If the asset is new, then inspect and test it early. Infantile mortality exists, and a share of new steam traps fail in year one.
• If the plant is mature and a PM program already exists, then use PMO to optimize it. If the failure modes are unknown or consequences are severe, use RCM to uncover them.
• If the asset runs irregularly, then drive the PM clock with cycles, machine hours, or pieces. If it runs regularly, use calendar days.

• Doing PM everywhere at the same weight. The tell: the task lists are uniform across assets, and the worst-consequence failure modes were never ranked.
• Treating repetitive repair as proactive work (snaking the pipe monthly). The tell: the same work order recurs with no root cause analysis or cost-of-leaving estimate.
• Buying predictive technology without a target failure mode. The tell: infrared or vibration data exists, but no failure mode names it as its detection method, and no action trigger is written.
• Letting strategy drift with management fashion. The tell: a 'flavor of the month' turnover, and resources cut with quarterly results. The source says maintenance deterioration is long-term and needs a long horizon.
• Waiting for breakdown on critical assets because 'if it ain't broke, don't fix it'. The tell: no one has authority to pull critical equipment from service before it fails, and breakdown is the main reason for write-up.

• ▸Asset and failure mode: name the asset and the specific way it fails.
• ▸Consequence class: mark safety, cost, or frequency (most dangerous, most expensive, most common), and note the worst consequence.
• ▸Reactive history: count breakdown work orders and rework or callbacks from reason-for-write-up data.
• ▸Root cause: record the underlying cause and the estimated cost of leaving it in place.
• ▸Life-cycle stage: mark new, in-service, or mature, and note the critical wear point if known.
• ▸Strategy chosen: pick PM, predictive, RCM or PMO review, TPM autonomous care, or redesign, with one line of reasoning.
• ▸PM clock and frequency: state the unit (days, hours, cycles, pieces) and the interval, or the predictive technique and its inspection interval.
• ▸Pull-from-service trigger: write the finding that sends the asset out for repair before breakdown, and who decides.
• ▸Review date and result: set the next failure-mode review and record what was changed or designed out.

1. 1Separate the two jobs on paper. Planning answers 'what does this job need?': resources, estimate, job package. Scheduling answers 'when does it run, and who agrees?': production coordination, weekly schedule, backlog. A job is planned only when every resource is identified and written into the job package. Only then is it schedulable.
2. 2Run a pre-planning question pass on each job (the book's 'Questions Before Planning'). Then build the package with a task-and-estimated-time breakdown and a material list, completed before work starts. The source puts the payoff at 3-5 hours saved per hour of planning on larger jobs, so spend planning time on large jobs first.
3. 3Set estimates from standards. Establish standard times or accurate job estimates for every repetitive job, PM and routine task. Divide PM task lists into interruptible and non-interruptible tasks so the scheduler knows which can be slotted around production.
4. 4Move a job into the 'ready backlog' only when it is planned, has its materials, and has authorization to proceed. Hold the ready backlog at 2-4 weeks of the crew's available hours. Under 2 weeks, the schedule starves. Over 4, the plans go stale.
5. 5Calculate available hours the way the questionnaire does: crew hours left after PM and breakdowns. Share that number with operations, and fill next week's slots with backlog-relief work from the ready backlog.
6. 6Build the weekly schedule against the master production schedule. Write each craftsperson's work a week ahead (target 75% or more of craftspeople). Have operations management and other interested parties review it, update it and sign it weekly.
7. 7Prioritize by criticality, not by who calls loudest. Use a rational scheme (the book lists RIME-prioritized work orders) so the most critical units are worked first. Treat unplanned, unscheduled work as 'non-scheduled work' and track it separately.
8. 8Review the week after it ends, asking why the schedule was or wasn't met, not who failed. Track whether jobs finish on time and on budget and whether mechanics arrive on schedule (the questionnaire's target is 90% of the time). Feed the misses back into estimates and job packages.

At a Maryland plastics extruder, a maintenance worker used a new vibration analysis tool and found a problem in the gear driving the main extruder screw. Fixing it now would cost $500. After a breakdown, the bill would be $5,000, so he claimed a $4,500 saving. He and the maintenance manager agreed with the shift manager to repair it on 2nd shift. The repair took unit 5 off line for a shift and a half. At Wednesday's production/maintenance scheduling meeting the manager met a chilly silence: production had missed a just-in-time delivery to a new customer, putting almost $40,000,000 of new business at risk. The planning was sound: the problem was found early, the repair was identified and the cost was estimated. The scheduling was the failure. The decision was never run through production, the master production schedule was never checked, and the weekly schedule was never signed off by operations. Applying this construct, the repair would have been planned with a job package and estimate. It would then have been slotted into a window agreed with production, with the delivery commitment visible, and the $4,500 'saving' would have been tested against what the downtime cost.

• If a job lacks any identified resource (parts, tools, authorization, time estimate), then it is not planned. Keep it out of the ready backlog and out of next week's schedule.
• If the ready backlog is under 2 weeks of available crew hours, then planning is the constraint, so add planning effort. If it is over 4 weeks, then stop adding to it and re-validate the oldest packages.
• If a maintenance decision takes a unit off line, then clear it with production through the weekly scheduling meeting or the master schedule before committing, even when the repair is clearly cost-justified.
• If a PM or routine task can be done without interrupting production, then tag it interruptible/non-interruptible and slot it into available hours around the production schedule.
• If the weekly schedule is missed, then review the cause (estimate, parts, priority, emergency load) in the next meeting and fix the process, not the person.

• Planning without scheduling: detailed job packages exist but nobody agrees the time with production. The tell is that repairs are surprises to the production floor, as at the extruder, and jobs get cancelled or delayed on the day.
• Scheduling without planning: slots are filled with jobs whose materials and estimates are missing. The tell is crews waiting on parts mid-job and jobs overrunning, with no estimate to measure against.
• Scheduling the whole crew's hours: PM and breakdowns are ignored, so the schedule overcommits. The tell is a weekly schedule that is routinely blown by emergencies. Only the hours left after PM and breakdowns are available.
• The post-week review becomes blame: the question is who failed, not why compliance missed. The tell is that people stop reporting accurate times and the same misses repeat.
• Unfunded maintenance liabilities are ignored: a reactive crew facing deferred deterioration cannot hold a weekly schedule. The tell is that emergencies still swamp the schedule after months of effort.

• ▸Job / asset and work order number: identify the work and the unit it affects.
• ▸Resources identified (parts, tools, skills, authorization): list each, and mark 'not planned' if any is missing.
• ▸Task list and estimated hours: break the job into tasks with times and a material list, from standards where they exist.
• ▸Interruptible or non-interruptible: state whether production can run through it or the unit must come down.
• ▸Priority (criticality rating): record the rational priority basis, such as RIME, not the requester's urgency.
• ▸Crew available hours next week: crew hours minus planned PM minus expected breakdown hours.
• ▸Production window agreed with: record the production contact, the date, and the master production schedule item checked.
• ▸Sign-off (operations + maintenance): record names and dates on the weekly schedule.
• ▸Post-week result (on time? on budget? why not?): note the cause of any miss, not the person.

1. 1Define the record chain first. Per the glossary, a work request is a formal, time/date-stamped request from a user, and it forms the basis of the work order. The work order is written authorization to proceed with a repair or other activity to preserve an asset. Chapter 6 treats the flow as request, work order, job completion. Write down the minimum fields each stage must carry, so history can be traced from the request to the closed job.
2. 2Fix the unit of record. The glossary says a 'unit' can be a machine, a system, or a component. Decide the level at which work is charged, and build the asset master file (Figure 27) to that level. A work order that names 'the line' instead of a unit cannot feed any later per-asset analysis, such as the component MTBF the book uses to compare maintenance strategies.
3. 3Audit the header before the body. The book gives separate audit tools for the work request header (Figure 15), the header information (Figure 24) and the body of the work order (Figure 25). Pull a sample of recent work orders. Score each header field as present, correct and consistent, and score the body separately. Keep header scores apart from body scores so you can see whether the fault is at intake or at completion.
4. 4Standardize the failure story. The book includes a 'Reason for Repair' figure (22) and an 'Example Repair Lesson' (23). Use a fixed reason-for-repair list rather than free text, and require a short repair lesson on completed jobs. This makes events from different technicians comparable and useful as the 'historic tail to breakdown events' (Figure 30).
5. 5Make the work type identifiable. The book separates standing work orders (Figure 19), string PM work orders (Figure 18) and ordinary repair work orders (Figure 17). It also tracks sources of work (Figure 8) and backlog status types (Figure 7). Require every order to carry its type, source and status so PM, user requests and breakdowns can be separated in any report.
6. 6Train to the body of the work order. The book includes a curriculum for work order (body) training (Figure 26). Teach technicians what to write in the body, why history depends on it, and what a good entry looks like. Use real audited orders as the teaching examples.
7. 7Close the loop on completion. Chapter 6 says supervisors and managers 'live or die by the quality and timeliness' of information. Set a rule that a job is not closed until the body fields are complete. Review the audit scores with the crew on a regular cycle and feed repair history into the technical library, which the glossary says holds repair history alongside manuals and drawings.
8. 8Check the CMMS against the paper process. The book warns that maintenance is one of the toughest departments to computerize, and that aids only help when the underlying processes are 'whole and complete'. Fix the fields and the flow first, then configure the CMMS to enforce them. Do not expect the software to repair a broken process.

The passages give no numeric case, so this one is constructed from the book's own definitions and is illustrative only. The book's improvement discussion tracks the average life (MTBF) of a component such as an air cylinder across maintenance strategies. Suppose you want that figure for a cylinder on one packaging machine. You pull the work orders. Three are written against 'Line 4' rather than the unit, one has a blank body, and two have free-text reasons ('leaking', 'bad'). You score the header audit: the unit field is wrong on three of six orders. The body audit: one order is empty. The result is that only two of the six orders can be used for MTBF. You then add a required unit field tied to the asset master file, a reason-for-repair pick list, and a closure rule that blocks a job with an empty body. After the next quarter you re-audit, and every cylinder failure can be placed on the right unit with a consistent reason. At that point the MTBF comparison between 'bust-and-fix' and PM becomes defensible.

• If a work order cannot be traced to a specific unit in the asset master file, then do not use it for history or benchmarking until it is corrected.
• If the header audit scores poorly, then fix intake (the request form and dispatcher checks); if the body audit scores poorly, then fix technician training and the closure rule. Do not apply one remedy to both.
• If reasons for repair are mostly free text, then replace them with a controlled reason list before you build any trend report.
• If you are considering a new CMMS or new modules, then confirm that the request-to-completion process and field definitions are complete first. The book says the aids only work when the underlying processes are whole.
• If a job type (standing, string PM, repair, user request) cannot be filtered in reports, then add a work-type field before running backlog or PM-compliance analysis.

• Bulk-closing or auto-closing orders to clear the backlog. The tell: closed orders have empty bodies and identical timestamps.
• Writing work against the line, area or department rather than the unit. The tell: asset-level history is thin while the work order count is high.
• Letting free-text failure descriptions stand in for coded reasons. The tell: the same failure appears under five different wordings, and trends are impossible to run.
• Treating the CMMS purchase as the data fix. The tell: the software is live but audit scores for headers and bodies have not changed from the paper era.
• Auditing only the header, or only completed jobs. The tell: dates and unit fields look clean, but repair lessons and body descriptions are missing, so history cannot answer why a failure happened.

• ▸Work order number and type: record the number and whether it is a repair, standing, string PM or user request.
• ▸Request time/date stamp present (Y/N): confirm the originating work request carries a stamp.
• ▸Unit/asset identified to master file (Y/N): confirm the unit matches an entry in the asset master file, not just a line or area.
• ▸Source of work: record where the work came from (for example, user request, PM, survey).
• ▸Reason for repair coded (Y/N): confirm a value from the standard list was used, not free text.
• ▸Body completeness: note what was done and found, and whether it is legible and specific enough to reuse.
• ▸Repair lesson captured (Y/N): note whether a lesson or cause note is recorded for the event.
• ▸Status and closure check: record status and whether closure rules were met before the order was closed.
• ▸Auditor score and corrective action: give a header score and a body score separately, and name the fix (intake, training or CMMS field).

1. 1Set the budget envelope first. Take total direct maintenance hours for the year and convert the 1-5% target into a hours-per-person figure and a total pool. Pick a point in the range deliberately. Stay near the low end if equipment and skills are already matched. Move toward the high end if the plant runs newer technology with a workforce trained on older technology. The source frames the gap as running a modern factory with 70's, 80's and 90's skills.
2. 2Build the competency list from the work, not from course catalogs. For each job class, write down what the equipment and processes require today: mechanical, electrical, electronic and microprocessor-based controls, planning, root cause analysis. This is the role of the Competencies figure and the Job Requirement Assessment (Form 1) in the book's training chapter.
3. 3Assess each person against the list. Use the Trainee Needs Assessment (Form 2) and record the gap per person per competency. Take stock of the training assets you already own, such as the microprocessor training set inventory, so you do not buy duplicates.
4. 4Sort gaps into the three training types in the definition: continual craft (deepen the trade), multi-craft (cross-training across trades), and management (supervisors and leads). Rank within each type by the equipment that hurts most when it fails, and by the work type: the source says different work types need different skills and personalities.
5. 5Choose the delivery method per gap from the Common Methods of Training to Consider list: vendors, prepackaged materials, and in-house options. Match method to gap. Equipment-vendor training suits new technology, and structured prepackaged materials suit repeatable basics.
6. 6Use real work as cross-training. Assign a mechanic to a root cause analysis. It merges reactive skills (the failure) with proactive skills (equipment history, operating conditions, engineering) and is described as the ultimate cross-training experience. Put people on standing or ad hoc teams (safety, environmental, planning, improvement) so they practice leadership safely.
7. 7Write a Training Agenda (Form 3) per person: what, which method, when, and who covers their work during the training. Schedule it into the plan and shutdown calendar, not into leftover time.
8. 8Track the hours spent against the 1-5% envelope and re-run the needs assessment on a fixed cycle. Treat training as a standing function. The source says maintenance leaders are in the training and development business whether they want to be or not, so regularize it.

The book's refinery mothball case shows what an undistributed maintenance knowledge base costs. When oil prices fell, a company decided to mothball a refinery. Accounting estimated savings of $10 to $15 million per year and a restart cost of $75 to $100 million. To maximize the savings, they laid off the entire maintenance staff and left only security on site. After 8 years, oil prices recovered and the company wanted the refinery back online for large contracts. Because of a lack of basic maintenance knowledge and of effective communication, the refinery was almost a complete loss and cost almost $700 million to bring back into service. Maintenance knowledge would have shown that $2 to $3 million per year preserved the plant, or $16 to $24 million over the 8 years. The same gap exists inside a running plant, just smaller: finance and other departments decide without maintenance knowledge, and maintenance people lack current technical skills. Spreading that knowledge through cross-department teams and structured training is the investment the 1-5% hours target pays for.

• If the equipment technology is newer than the skills on the crew, then set training toward the top of the 1-5% range and begin with the gap rated most damaging in the Job Requirement and Trainee Needs assessments.
• If a skill gap spans trades (for example a failure that needs mechanical, electrical, and engineering knowledge), then cross-train through a root cause analysis or a team assignment rather than a classroom course.
• If a person is already leading teams, planning jobs, or supervising, then fund management training (communication, delegation, time management) ahead of more craft hours.
• If training cannot be scheduled without pulling people off urgent work, then plan it into shutdown or planned-work windows and name coverage in the Training Agenda. Unplanned training gets cancelled first.
• If you are choosing between vendor and prepackaged materials, then pick vendor training for specific new equipment and prepackaged materials for broad, repeatable content, and check either against the Competencies list before buying.

• Buying courses before assessing needs. The tell: training hours are spent, but the Trainee Needs Assessment gaps are unchanged at the next review.
• Treating training as a perk and not as skill matching. The tell: the course list follows what is available or popular, not the Competencies list tied to current equipment.
• Cross-training only inside one trade. The tell: electricians and mechanics each know their craft but nobody can work a failure that spans both, and root cause investigations stall at department lines.
• Starving the budget when pressure rises, as the refinery layoff did. The tell: training is the first line cut and the hours percent drifts below 1% while equipment gets more complex.
• Skipping management training. The tell: strong craft people promoted or put on teams without communication, delegation, or time-management skills, so team and planning work runs poorly.

• ▸Name / job class: who this plan covers and their current trade.
• ▸Required competencies (from Job Requirement Assessment): list what current equipment and processes demand of this job.
• ▸Current level vs. required, per competency: rate the gap, using Trainee Needs Assessment data.
• ▸Training type: continual craft, multi-craft, or management; one per gap.
• ▸Priority and equipment at risk: which asset or process suffers most if the gap stays open.
• ▸Method: vendor, prepackaged, in-house, or on-the-job (such as a root cause analysis or team assignment).
• ▸Planned hours and window: hours to be spent and the date or shutdown period they fall in, with who covers the work.
• ▸Share of annual direct hours: planned hours divided by this person's direct hours, checked against the 1-5% target.
• ▸Review date and result: when you re-assess the gap and what changed.