To implement a preventive maintenance program, you build a complete asset register, rank every asset by criticality, write scheduled tasks against the highest-risk equipment, schedule them in a CMMS or work order system, and then measure compliance and downtime on a fixed review cycle. A small-to-mid plant can run that sequence in about six months, and most of the work is data collection rather than new equipment.
Plastics plants make a good test case because the asset list is long and the failure consequences are expensive. An injection molding machine going down stops the line, a dehumidifying dryer going down can scrap an entire lot of resin, and a chiller fault shows up as warped parts hours later. The nine steps below are the order I would use, with an owner and a deliverable attached to each one.
Table of Contents
- What You Need Before You Build a Preventive Maintenance Program
- How to Implement a Preventive Maintenance Program, Step by Step
- 1. Establish the Program’s Goals and Scope
- 2. Inventory Equipment and Build a Maintenance Baseline
- 3. Rank Equipment by Risk and Criticality
- 4. Create Preventive and Predictive Tasks
- 5. Set Work Instructions, Safety Controls, and Spare Parts
- 6. Pilot the Preventive Maintenance Program on Selected Equipment
- 7. Roll Out the Program Across the Plant
- 8. Track Reliability, Cost, and Compliance Results
- 9. Review and Continuously Improve
- The Mistakes That Break Most Preventive Maintenance Programs
- Frequently Asked Questions
- Who should own a preventive maintenance program?
- How often should preventive maintenance be performed?
- Is preventive maintenance required by OSHA?
- Do we need maintenance software to implement a program?
- How can a small maintenance team get started?
- When is predictive maintenance better than scheduled preventive maintenance?
- Conclusion: Start With Your Ten Most Critical Machines
What You Need Before You Build a Preventive Maintenance Program
Most failed rollouts are not technology failures. They start because six prerequisites were never assembled, and the team discovers the gaps halfway through month two.
Equipment records. Manufacturer manuals, as-built drawings, wiring diagrams, previous maintenance logs, and the nameplate data for each machine. OEM recommended service intervals are the starting point for your task list, and you should copy them rather than invent frequencies.
Staff involvement. A named program owner, a maintenance lead who can approve work, technicians who will actually perform the tasks, and a production representative who helps schedule downtime. If production management is not in the room, work orders get skipped at shift change.
Maintenance standards. The documentation format you will use for work instructions, checklists, and completion records. Decide this once, otherwise you end up with paper checklists, a spreadsheet, and a CMMS that all disagree.
Baseline data. What unplanned downtime cost you last year, how many work orders were completed, and how many reactive calls your technicians absorbed. Without a baseline you cannot prove the program worked, only that it existed.
Budget. Time for the work itself, which is the biggest line and the one most often ignored, plus money for meters, sensors, tooling, and critical spare parts. Labor hours lost during planned maintenance are part of the cost of the program, not an interruption to it.
Safety procedures. Your lockout/tagout program, machine-specific guarding rules, permit requirements, and contractor controls. Every PM task involving electrical, hydraulic, or pneumatic energy needs a defined isolation method before it starts.
Performance measures. A short list of metrics agreed in advance, typically planned-work compliance, mean time between failures, mean time to repair, and unplanned downtime hours. Pick these before rollout so nobody argues about the definition of success after the data is in.
How to Implement a Preventive Maintenance Program, Step by Step
The steps run in order because each one depends on the last. Skipping the criticality ranking is the most common shortcut, and it is the reason so many programs generate work that does not matter.
1. Establish the Program’s Goals and Scope
Start by writing down what the program is for and which equipment it covers. The goal is usually a mix of safety, quality, delivery, and cost: fewer recordable incidents, less scrap from process instability, more on-time shipments, and less unplanned downtime.
Scope it narrowly for the first phase. A 60-machine plant with 12 injection molding machines, 8 dryers, 3 granulators, 2 compressors, a chiller, and the temperature-control and ventilation systems is a reasonable first target. Leave the packaging equipment and forklifts for phase two.
Owner: the maintenance manager, with a plant sponsor at executive level.
Deliverable: a one-page charter listing scope, goals, named assets, and the review meeting date.
2. Inventory Equipment and Build a Maintenance Baseline
Build the asset register. Every asset gets an ID, a location, a make and model, a serial number, the OEM manual reference, the critical spares list, and a condition rating. Record runtime or cycle counts if the machine has an hour meter, because those numbers drive your usage-based triggers later.
Then capture the maintenance baseline: what has failed in the last 24 months, how long each repair took, and what it cost in parts and labor. Most plants already know their worst three machines and cannot remember the rest, which is a sign the history lives in technicians’ heads rather than in a system.
For a plastics plant, capture the supporting equipment too. Chillers, compressed air, dust collection, material conveying, and the water treatment on the cooling tower all sit upstream of the molding machines and will take the line down when they stop.
Owner: a reliability engineer or senior technician, with production providing runtime data.
Deliverable: a searchable asset register and a one-page failure history summary per critical asset.
3. Rank Equipment by Risk and Criticality
Score every asset on how bad a failure would be, then use that ranking to decide how much maintenance attention each machine earns. A common method scores four consequences on a 1 to 5 scale and multiplies them by an availability factor.
| Consequence | Question to ask | Example consequence |
|---|---|---|
| Safety | Can failure injure a person? | Guard or interlock failure, hot surfaces, high-pressure lines |
| Quality | Can failure produce defective output? | Chilled water loss, unstable melt temperature, dryer fault |
| Production | Does the line stop without it? | Injection molding machine, main compressor, granulator |
| Environmental | Can failure cause a release or excess consumption? | Cooling tower drift, granulator dust, oil leak paths |
| Repair | How long and how costly is the fix? | Long lead time on the part, need for a rental or an outside tech |
Assets scoring in the top tier get frequent tasks, condition monitoring, and stocked spares. The bottom tier often does not need scheduled work at all. Running a low-consequence conveyor or a noncritical fan to failure is a legitimate strategy, not a failure of the program.
The 80/20 rule in maintenance usually shows up here. A small share of assets generates most of your failures and most of your downtime hours, and the criticality ranking is how you find that share before you start spending labor on everything equally.
Owner: the reliability engineer with production and EHS sign-off.
Deliverable: a ranked asset list in four tiers with a written strategy assigned to each tier.
4. Create Preventive and Predictive Tasks
Turn manufacturer guidance, your failure history, and the criticality ranking into tasks. Four trigger types cover most plants.
- Calendar-based, triggered by date. Filter changes, belt inspections, annual safety checks. Simple and easy to audit.
- Runtime-based, triggered by operating hours or cycles. Lubrication points on a molding machine usually work better on hours than on the calendar, because a two-shift machine accrues wear far faster than a one-shift one.
- Condition-based, triggered by a measured value against a threshold. Vibration on the screw drive gearbox, bearing temperature, filter differential pressure, chiller water temperature.
- Predictive, triggered by trend analysis rather than a single threshold. Useful on rotating equipment and hydraulic systems where degradation is visible in the data before failure.
Write each task with a title, trigger, estimated duration, required skill level, tools, and a completion criterion that can be judged pass or fail. “Inspect” is not a task. “Confirm barrel zone temperatures hold within 5 degrees F of setpoint over a 30-minute run” is a task.
Owner: the maintenance lead with the reliability engineer.
Deliverable: a task library keyed to asset ID, with a trigger type and interval for every entry.
5. Set Work Instructions, Safety Controls, and Spare Parts
Turn each task into a work instruction the technician can follow without asking a supervisor. That means the procedure, the required qualifications, the tools, the personal protective equipment, the lockout points, the acceptance criteria, and the record that has to be captured at the end.
Safety controls are not optional decoration on a PM work order. Isolate and verify the energy state under your lockout/tagout program before anyone opens a guard or touches a drive, treat the mold area and hot barrel as burn and injection hazards, and require a second person for anything that raises a machine off its supports. Trained personnel follow equipment-specific procedures and the manufacturer’s instructions, not a generic plant checklist that was written in a conference room.
Stock critical spares against the top tier assets only. Filters, seals, heating elements, thermocouples, contactors, and proportional valves are the usual candidates. A part you do not hold turns a two-hour repair into a two-week one, and that single delay can erase the savings from a quarter of good preventive work.
Owner: the maintenance lead with EHS for controls and the storekeeper for parts.
Deliverable: printed or tablet-ready work instructions with a critical spares list per asset.
6. Pilot the Preventive Maintenance Program on Selected Equipment
Run the program on a small group for 60 to 90 days before rolling out. Pick assets that are representative rather than convenient: one machine with a documented failure history, one asset with a simple monthly task, and one where the OEM interval is not yet known.
The pilot is where you find out that the intervals are wrong, the work instructions are missing a step, the tasks collide with production shutdowns, the labor estimate is off by a factor of two, or the parts are not on the shelf. Finding those problems on 6 machines costs a month. Finding them on 200 costs a year.
Set a go/no-go gate at the end: at least 85 percent of scheduled tasks completed on time, no safety events during PM work, and a written answer for every task that got skipped.
Owner: the maintenance manager.
Deliverable: a 90-day pilot report with completion rate, downtime avoided, and a revised task list.
7. Roll Out the Program Across the Plant
Phase the remaining equipment in waves rather than all at once, and order the waves by criticality. Start with the top tier, move to tier two, and decide explicitly which assets are staying on run-to-failure.
Assign accountability by name. Every asset in tiers one and two has a named task owner in production and a named technician in maintenance. Train the people who do the work, not just the managers who approve it, and give them the checklist logic and the reporting channel before their first week of live tasks.
Schedule work in windows production can live with, usually a mix of planned weekend or shift-break stops for heavy work and operator checks during production for light work. Define what happens when a task goes overdue: who gets told, after how long, and what the escalation looks like. A silent escalation rule is not an escalation rule.
Owner: the maintenance manager with the plant manager signing off on downtime windows.
Deliverable: a rollout schedule by wave, a trained technician list, and a written overdue-work escalation rule.
8. Track Reliability, Cost, and Compliance Results
Measure a small number of metrics and interpret them together. Any one of them alone will mislead you.
| Metric | How it is calculated | What it tells you |
|---|---|---|
| Planned-work compliance | Tasks completed on time divided by tasks scheduled in the period | Whether the schedule is realistic and whether the team is executing it |
| Mean time between failures (MTBF) | Total operating time divided by number of failures | Whether reliability is actually improving |
| Mean time to repair (MTTR) | Total repair labor time divided by number of repairs | Whether parts, access, and skill are the constraint |
| Preventive-to-corrective ratio | Preventive labor hours divided by corrective labor hours | Whether work is shifting from breakdown to planned |
| Unplanned downtime | Hours of production lost to unplanned stoppages | The business outcome, in the unit the plant feels |
| Backlog ratio | Work order labor hours outstanding divided by available labor hours per week | Whether the team is losing ground on scheduled work |
| Repeat failures | Failures on an asset within 30 days of a repair, as a share of all failures | Whether root cause work is happening |
Watch the pairs, not the singles. If compliance rises while MTBF falls, the team is doing more work that does not matter. If MTBF rises while compliance falls, you got lucky on a few assets. If the backlog ratio climbs above one, the schedule exceeds the labor available and something has to give before the program collapses under its own paperwork.
For the business case, published industry studies put the maintenance-cost reduction from condition-based and predictive programs somewhere in the high teens to around 30 percent against purely reactive methods. Treat that as a directional benchmark rather than a promise for your plant, and build your own case from your own baseline data.
Owner: the reliability engineer, reviewed monthly with the plant manager.
Deliverable: a one-page dashboard reviewed on a fixed day each month.
9. Review and Continuously Improve
Hold a review on a fixed cycle, monthly for the first year and quarterly once the program is stable. Recurring failures get a root cause investigation rather than another PM task. A task that has been rescheduled four times in a quarter is telling you something, usually that the interval is wrong, the labor estimate is wrong, or the task does not prevent the failure it was written for.
Adjust intervals based on what the work orders show, revisit the criticality ranking as the plant changes, and retire tasks that produce no measurable benefit. Fewer, better tasks beat a crowded calendar. Software teams run into the same trap when they measure tickets closed instead of defects prevented, and maintenance is no different.
Report program performance to plant leadership quarterly in terms they act on: downtime hours, scrap attributed to equipment instability, and the reliability of the assets that were supposed to be fixed.
Owner: the maintenance manager.
Deliverable: a quarterly review with revised intervals, a revised criticality ranking, and a task list that is shorter or better justified than the one before.
The Mistakes That Break Most Preventive Maintenance Programs
Copying a calendar without a risk basis. Teams pull a 12-month service schedule from a template and apply it to everything, then wonder why the important machines are still breaking. Fix: rank by criticality first, then write intervals per asset tier. Run-to-failure is a valid choice for low-consequence assets.
Recording everything as preventive. If a work order is labeled preventive but the machine was already broken, the compliance rate and the PM ratio both lie. Fix: define preventive, corrective, and inspection work in your CMMS and audit a sample of records each month.
Leaving the work order data in the technician’s head. Hours, parts, and failure notes written on a clipboard in a truck never reach a dashboard. Fix: mobile entry at the point of work, with the checklist and the labor capture on the same device.
Skipping operator checks. Production staff see the machine every day and catch leaks, noise changes, and loose guards hours before a technician would. Fix: short daily or per-shift operator checklists tied to the assets they already own, with a route to report a defect.
Understocking critical spares. The one part you do not hold is the part the top tier asset needs. Fix: build the spares list from the criticality ranking and measure time-to-repair for the assets where the repair is on the long-lead side.
Measuring only maintenance cost. A cheap maintenance department that lets the plant stop is not performing well. Fix: pair cost metrics with downtime, MTBF, and quality outcomes, and review them together.
Ignoring change management. Scheduled work pulls technicians away from the urgent calls they are judged on, and the schedule quietly gets ignored. Moving a maintenance organization from unscheduled breakdowns to planned and monitored work changes how the whole department operates day to day, and that shift is genuinely difficult. Fix: involve technicians in writing the tasks, publish the compliance number where the team can see it, and give people credit for work they completed on time rather than only for fires they put out.
One more practical tip. Start the program on assets with a documented failure history, not the ones that are easiest to reach. An easy pilot teaches you that your system works; a hard pilot teaches you what to fix before you scale.
Frequently Asked Questions
Who should own a preventive maintenance program?
Ownership usually sits with the maintenance manager or reliability engineer, because the program is fundamentally a work-order and data discipline. A plant manager sponsors it and approves the planned downtime windows, production supervisors own the operator checks, and technicians own task completion. The failure mode is assigning it to a single person with no production support, since scheduled work that conflicts with production gets skipped within a month.
How often should preventive maintenance be performed?
Frequency depends on the asset, its criticality, and the manufacturer’s recommended interval, so there is no single correct schedule for a plant. Calendar-based checks suit filters, belts, and safety inspections. Runtime-based triggers suit lubrications and wear parts on high-hour machines. Condition-based and predictive triggers suit rotating equipment, hydraulics, and anything where degradation shows up in a measurement before it fails. Start with OEM guidance and adjust from your own failure history.
Is preventive maintenance required by OSHA?
OSHA does not require a preventive maintenance program as a written program, but it does require the underlying safety work through several standards. Lockout and tagout under 29 CFR 1910.147 covers energy control during servicing, electrical standard 1910.303 covers equipment inspection, 1910.212 covers machine guarding, and 1910.147(f) covers periodic inspection of energy control devices. State plans can be stricter, and NFPA 70B adds electrical preventive maintenance guidance. Check your own jurisdiction.
Do we need maintenance software to implement a program?
No, and software will not rescue a program that has no asset list or no task definitions. A spreadsheet works for a small site with fewer than roughly 50 assets and a stable equipment base, provided someone owns it and it generates scheduled work orders. Once you need meter-based triggers, automatic work order generation, labor and parts history, mobile entry in the field, and overdue escalation across multiple sites, a CMMS is usually cheaper than maintaining a spreadsheet by hand.
How can a small maintenance team get started?
Pick the five to ten assets whose failure would stop the line or create a safety or quality event, and build the program on those first. Capture their manual, service history, and runtime hours, rank them, write two or three tasks each from OEM guidance, and schedule them. Run that for 90 days, measure completion and downtime, then widen. A small team that masters ten critical machines beats one that has a calendar for four hundred assets and completes 40 percent of the work.
When is predictive maintenance better than scheduled preventive maintenance?
Predictive maintenance is better on assets where failure is gradual and measurable, such as rotating machinery, gearboxes, pumps, compressors, and hydraulic systems. It lets you act on a real trend instead of a fixed calendar, which avoids replacing components that are still in good condition. It also costs more up front, since condition monitoring needs sensors, data collection, and analysis capability. On low-consequence assets it rarely pays for itself, so schedule those instead.
Conclusion: Start With Your Ten Most Critical Machines
How you implement a preventive maintenance program comes down to a sequence: define the scope, build the asset register, rank by criticality, write the tasks and work instructions, pilot on a small group, roll out in waves, measure with paired metrics, and keep reviewing. The order matters more than the speed.
Do one thing this week. List the machines whose failure would stop your line, rank them, and draft the first three tasks for the top one from the manufacturer’s manual. Everything else in the program follows from that list.