Preventive
Done before the failure happens, according to a time- or usage-based plan. Replacing a filter every three months or greasing every 500 hours. It's the most widespread and the most skipped.
Guide
The types of maintenance, without the naming confusion
The classification varies depending on which manual you read and who's explaining it. This is the one used in practice, with what each term actually means when someone in the industry says it.
The criterion behind the classification
Every maintenance classification answers the same question: what triggers the intervention. If it's triggered by the calendar or usage, it's preventive. If it's triggered by a failure that has already happened, it's corrective. If it's triggered by a measurement indicating a failure is approaching, it's predictive or condition-based. And if it's triggered by a regulation, it's legally required maintenance, even though in practice it's carried out like preventive work. Understanding that criterion avoids most terminology arguments, which are usually about names rather than about different things.
The types, one by one
With what they actually mean when someone uses them in a meeting.
Corrective
Done once something has already failed. It can be urgent — service has to be restored now — or scheduled, when the failure doesn't stop things from running.
Predictive and condition-based
Triggered by a measurement: vibration, temperature, oil analysis, consumption. It anticipates the failure instead of waiting for it or preventing it on a calendar.
Legal or regulatory
Imposed by a regulation, with its own frequency and scope. Carried out like preventive work, but the date isn't negotiable and the proof has to be kept.
Routine (operator-driven)
The day-to-day running of an installation: readings, rounds, small adjustments. In buildings and plants with fixed staff, it's its own category and a high-volume one.
Modification or improvement
Doesn't repair or prevent: it changes the installation so it works better or adapts to a new use. Usually managed as a project or works job.
You don't pick one: they coexist
A real installation runs the first five at the same time, and the useful question isn't which one to use but in what proportion. A cheap, redundant, easily replaceable unit might not deserve any preventive maintenance at all: you let it fail and swap it, and that's a legitimate decision known as run-to-failure. A critical, expensive unit justifies intensive preventive work and, if the cost of downtime supports it, instrumentation to anticipate failure. Almost everything sits between those two extremes, and the only way to place each asset correctly is to know what it costs when it stops and how often it actually fails.
A warning about predictive maintenance
It's the most heavily sold and the least well implemented, because it has a prerequisite almost nobody mentions: detecting a pattern requires history, and having history requires having logged failures properly for a while already. Instrumenting equipment at a plant that closes out work orders with "checked and working" produces a very nice-looking dashboard fed with data nobody trusts. The order that actually works is: log corrective work with its cause, base the preventive plan on that history, and only then consider whether any critical asset justifies sensors.
Shall we look at it with your way of working?
Leave your details and we will get in touch to see whether we fit. No commitment, no lock-in period.
Not all corrective work is an emergency
The distinction that matters most in practice isn't between preventive and corrective, but within corrective work itself. A failure that stops production and needs attention right now is one thing, and a defect someone notices, logs and fixes when convenient is a very different one: a door that scrapes, a small leak, a burnt-out bulb. Both are corrective, because both fix something that already failed, but they're planned in opposite ways and cost very different amounts. Separating them also has a useful consequence in reporting: a company where ninety percent of corrective work is urgent has a problem; one where most of it is deferred and grouped into scheduled visits has maintenance that works, even if both have the same number of corrective jobs.
From theory to a field on the work order
This whole classification only matters if it ends up as a piece of data on every work order, and that's where it most often fails. If the type is filled in by guesswork, or if half the orders go out with no type because the field is optional, the reports built on top of that mean nothing, and the worst part is nobody notices: the chart still gets drawn either way. Two rules avoid most of the problem. First, the type should be set by default whenever possible — an order generated by a preventive plan is born preventive, and nobody should have to be asked. Second, keep the categories few: four or five that everyone understands the same way are worth infinitely more than twelve nuanced ones that each person interprets their own way.
Frequently asked questions
How many types make sense to use in practice?
Fewer than the theory suggests. Four or five categories that everyone understands the same way give reliable reports; a dozen nuances that each person interprets differently give reports that look precise but aren't. You can always refine it later, once the history proves the distinction is actually needed.
How many types of maintenance are there?
It depends on the classification, but in practice six are used: preventive, corrective, predictive or condition-based, legal or regulatory, routine, and modification. They all answer the same question: what triggers the intervention.
What's the difference between predictive and preventive?
Preventive is triggered by a calendar or accumulated usage; predictive is triggered by a measurement indicating a failure is approaching. Predictive avoids unnecessary interventions, but it requires instrumentation and prior history.
Is legally required maintenance a separate type?
It's carried out like preventive maintenance — a plan, a frequency and a checklist — but with two differences: the frequency is set by a regulation rather than technical judgment, and documentary proof has to be kept.
Is having a lot of corrective work bad?
Not necessarily. Some equipment is cheap and redundant enough that letting it fail and replacing it is the right call. What is a problem is predictable corrective work: failures a reasonable plan would have prevented.
Where do I start if I have nothing set up?
By logging the corrective work that's already happening, against its equipment and with its cause. With a few months of that history, the preventive plan gets designed around what actually fails at your site, not around the manufacturer's catalog.
What's your mix?
In a short demo we look at your assets and see what proportion you're actually running, which is almost never what people assume.