Corrective, preventive, condition-based, predictive: these four terms come up in every discussion about maintenance strategy, usually with a fair amount of confusion. People talk about “doing predictive” when they are simply reading a meter, or call a fixed calendar “condition-based.” For a maintenance manager, this clarity is not a vocabulary debate: it decides where to invest, which equipment to monitor, and how to spend a budget that is always limited.
This guide defines each type of maintenance clearly, lays out its benefits, limits, and cost, then helps you choose based on equipment criticality and start in a realistic way.
Corrective maintenance: repair after failure
Corrective maintenance means acting once a failure has occurred. The machine runs until it breaks, then the team restores it. There is emergency corrective work, unplanned, that stops production, and deferred corrective work, where a tolerable fault is noted and the repair is scheduled.
Its advantage: no unnecessary intervention and immediate adoption, with no heavy organization. For low-criticality, redundant, or end-of-life equipment, it is often the rational choice.
Its limit: unplanned downtime disrupts production, emergency repairs cost more, take longer, and are sometimes done without the right parts in stock. A failure left to happen can trigger others. Corrective work remains essential — no strategy eliminates it entirely — but relying on it for your critical machines is the most expensive path.
Preventive maintenance: anticipate by time or usage
Systematic preventive maintenance triggers work on a fixed plan, regardless of the machine’s actual condition. Two triggers:
- Calendar-based: every month, every quarter, every year.
- Meter / usage-based: every 500 running hours, every 10,000 cycles, every 20,000 km.
Its benefits are concrete: fewer unplanned stops, longer equipment life, technician work you can plan ahead with parts and tools ready. It is the foundation of any serious strategy and, for most sites, the biggest immediate gain.
Its limit is over-maintenance: following a rigid calendar leads to stripping and replacing parts that are still healthy, wasting time and parts and sometimes introducing defects into a machine that was running fine. Calendar-based preventive work also ignores real usage: a lightly used machine gets as much service as one pushed to its limit. The meter-based trigger partly corrects this by tying frequency to actual usage.
Condition-based maintenance: act on readings and thresholds
Condition-based maintenance relies on the equipment’s actual state, measured through readings: temperature, pressure, vibration level, thickness, oil analysis. You set a threshold; as long as you stay below it, you touch nothing; as soon as it is crossed, you schedule the work.
Its advantage: you intervene only when justified, neither too early nor too late. You avoid both over-maintenance and unplanned failure. It suits equipment where a wear parameter can be measured easily.
Its limits: you have to know which parameter to monitor, how often to record it, and where to set the threshold — which requires understanding how the machine degrades. Regular manual readings cost time. But condition-based maintenance does not necessarily require expensive sensors: an operator who notes a temperature or a vibration level at a measurement point each week is already doing condition-based maintenance.
Predictive maintenance: analyze data to forecast
Predictive maintenance goes further: instead of comparing a single reading to a threshold, it uses history and continuous data — often from connected sensors — to model degradation and estimate when failure will occur. The goal is to schedule work just before the breakdown, at the best possible moment.
Its theoretical upside is the highest: maximum downtime reduction, service kept to the minimum, decisions based on trends. But its cost and prerequisites are just as high. You need to instrument the equipment (sensors, data acquisition), collect and store clean data over time, then analysis tools capable of producing reliable models. On heterogeneous equipment or with little history, those models are fragile. Predictive maintenance is justified mainly on highly critical, costly-to-stop equipment, where the investment pays off.
Let us be clear: it is a demanding approach, not a checkbox. Many tools promise “AI predictive maintenance” that, in practice, comes down to threshold rules — that is, condition-based maintenance.
How to choose based on criticality
None of these strategies is “the best.” The right mix is decided equipment by equipment, based on its criticality — the impact of a stoppage on production, safety, and quality.
- Low-criticality, redundant, or replaceable equipment: corrective often suffices.
- Standard equipment: calendar- and meter-based preventive work, the backbone of your plan.
- Critical equipment with measurable wear: add condition-based maintenance on readings and thresholds.
- Highly critical, costly-to-stop, well-instrumented equipment: predictive maintenance may be justified.
Where to start, concretely
The classic trap is wanting to jump straight to predictive. In reality, the most profitable step is taken well before that. Setting up a preventive maintenance plan on calendar and meter triggers, with work orders generated automatically and a traced history, is already a huge step forward for the many sites still running maintenance on a spreadsheet or from memory.
Condition-based maintenance comes next, naturally: configure readings and meters on your sensitive equipment and trigger service when a threshold is crossed. This is achievable without heavy hardware investment.
Predictive maintenance, finally, is a project in its own right: sensors, data, models. Only take it on for the rare equipment that justifies it, once preventive and condition-based work are under control.
In all honesty: the CMMS covers calendar- and meter-based preventive maintenance as well as the threshold readings that structure condition-based maintenance. The software does not do sensor- or AI-based predictive maintenance, and we do not claim it does — but for the vast majority of teams, that is exactly where the fastest and safest gains are found.