How to Build a Preventive Maintenance Program That Actually Sticks

Key Takeaways

A preventive maintenance program is defined by what you leave off it. Coverage is what makes a schedule unfinishable, and unfinishable is what makes it optional.

The manufacturer's interval is a starting position. The frequency that holds is the one corrected by what the completed jobs keep finding.

Schedules survive by being edited. A task that never changes, on a machine that never surprises anyone, tells the crew what the schedule is worth.

Most preventive maintenance programs are built to cover everything, and the ambition is what kills them. Put every asset on a schedule and you generate more planned work in a month than the crew can finish. Once it can't be finished, somebody starts choosing which parts to ignore, and from then on it's a suggestion.

Someone organised keeps a spreadsheet of dates, it looked right in January, and by March it carries jobs nobody has done and nobody has cancelled. Capturing the work in one record comes first, and that's a separate argument.

The program that survives is the small one, and building it is mostly a matter of deciding what stays off. Size is the risk. McKinsey's transformation work puts the failure rate for large-scale organisational change at roughly 70%, and that covers programs of every kind rather than anything specific to maintenance.

Which assets actually belong on a preventive maintenance schedule?

The ones where a failure costs more than the checks, and where nothing tells you it's coming. That list is much shorter than your asset register. Most equipment shows you it's going wrong in time to act, or fails cheaply enough that preventing it was never worth the hours.

Three questions sort it, and you can answer all three from the floor.

Those questions need an input most sites don't keep. A head of engineering at a manufacturer running fifteen sites put the gap plainly: nobody above the maintenance team could see which assets drove the cost or what kept breaking.

For a real share of the equipment on any site, running it to failure is the correct engineering call. Saying so out loud keeps the rest of the program credible. A schedule that ticks boxes on things nobody minds losing spends its hours proving it isn't serious.

Expect the first list to be short. Ten or fifteen assets is a working program. Holistic Industries built theirs around the failures actually costing them and reported microbes and mould down 71% after putting preventive maintenance in place. That result landed in a quality measure rather than a maintenance one, which is the point of choosing by consequence.

How often should each preventive task run?

Start at the manufacturer's interval and treat it as a first guess. It was set for the machine in general, and it knows nothing about your duty cycle, your dust, or the operator who runs the thing hardest on nights. The interval that holds shows up in what the jobs keep finding, and you can't read that until a few cycles have run.

Every frequency in the first version is provisional. Write them down that way.

Over-maintaining is the failure mode nobody warns you about. A monthly inspection that finds nothing for a year costs you twice. Once in hours you owed somewhere else. Once in credibility, because the crew works out which tasks are decoration well before you do.

There's a harder version of this. Opening a healthy machine introduces a risk that wasn't there before: a gasket seated badly, a fastener over-torqued, contamination let into a housing sealed for two years. Maintenance is an intervention, and interventions carry their own failure rate. That's the argument for stretching an interval that keeps coming back clean, and it's the half of the case the category almost never makes.

What has to be in a preventive maintenance task for it to get done properly?

Everything the job needs, and nothing that exists for the report. The quickest way to kill a task is to ask a technician to do something on a screen they'd never do with a clipboard in their hand. Time logged against each line item. Fourteen boxes ticked for one grease round. Work shaped like that gets completed in the car park at the end of the shift, from memory.

What comes back is a completion with no observation inside it, which does more damage than a skipped job. A skipped job at least looks skipped. The managers we work with name the same failure: the system asked for something the job doesn't produce, so the job stopped producing anything.

A task that survives contact with the floor carries three things:

What to check, named precisely enough that two technicians would check the same thing.
What good looks like, so the answer is a reading or a comparison instead of a fresh judgement call every time.
What to do with a finding, so the problem spotted during the check lands somewhere the next person will see it.

Everything past those three is reporting, and reporting attached to a task gets paid for by the person holding the wrench.

What keeps a preventive maintenance program alive when the reactive work comes back?

Editing it. A program stops being real the moment it becomes a fixed object, because a schedule that never changes tells the crew a year of doing the work taught nobody anything. Set a review, quarterly is plenty, and go through the tasks rather than the people.

The question at that review is what each task found the last four times it ran.

Nothing, every time. Stretch the interval.
Something, every time. The interval is too long, or that machine needs a different kind of attention.
Something, occasionally. Leave it where it is.

Most tasks leave a schedule without anyone deciding to remove them. They die quietly. Skipping one costs nothing anybody can see, the machine is still running on Friday, and the hour goes to a job with a person waiting at the end of it. That's how most schedules empty out. An interval somebody has reviewed and defended holds up better, because it carries a reason a technician can hear.

Grow the program when the current list finishes reliably, and never on a calendar. Holistic Industries cut grow room turnover time by 50% once preventive work was scheduled against the equipment that needed it. The gain showed up in a changeover that used to be unpredictable, which is what a schedule that keeps running gives back. A routine job stops being an event.

The short list was the program all along. Everything you left off it was a decision, and those decisions are where the judgement lives. The compressor checked monthly because it takes the line with it. The pump left to run until it stops, because replacing it costs less than watching it.

A maintenance manager who can say why each of those sits where it does has a preventive maintenance program. A register with a frequency in every row is an intention, and it's what most teams have instead.