Maintenance Optimization: Strategy, Schedule and Savings

Maintenance optimization is an ongoing activity, not a one-time project or a piece of software you install. It is the steady work of tuning your maintenance to how each asset actually behaves. At its core, it answers four essential questions pertaining to every asset:

• Which strategy: the maintenance method you apply to the asset and its failure modes, from preventive work through to run to failure.
• At what interval: how often the work happens, set from real failure data rather than a copied default.
• Scheduled when: the point in the production plan where the work causes the least disruption.
• With which parts and people: the spares staged and the technician hours available to actually finish the job.

Most teams put all their effort into the first question and neglect the other three, which is why so many programs stall once the analysis is done. It is worth being clear about what optimization is not. Buying new software does not optimize maintenance on its own, and neither does simply doing more preventive work. Often the right answer is doing less of it, but on the assets that matter.

Most teams already respect the cost of under-maintenance. Skip a needed inspection or stretch an interval too far, and the asset fails in service, usually at the worst possible time and the highest cost. That risk is well understood, so it rarely needs arguing.

What gets far less attention is over-maintenance, and it can quietly cost even more. Because servicing an asset looks responsible, few people question its frequency. But doing more work than an asset needs ties up labor, parts and uptime, and it can even cause failures. It leaks money and reliability in three ways:

• The maintenance task itself causes failures: Sometimes the task itself introduces the defect. A seal seated wrong on reassembly, dirt let into a clean system, or a good part swapped out early for a weaker one can push an asset into a failure it would never have reached on its own.
• The optimal interval was never chosen for this asset: Many preventive frequencies were copied from the equipment manual or the commissioning pack, written before anyone knew how hard the asset would really run. A pump might still be serviced every 90 days out of habit, even though the asset now runs a completely different load.
• Effort goes to the wrong places: Technician hours spent on tasks that never find anything wrong are hours taken away from the assets that are genuinely wearing out.

The goal, then, is not more maintenance or less. It is the right amount on each asset. A simple test tells you which way a program is heading: a real optimization effort usually ends with fewer preventive tasks than it started with, not more. If the task list only ever grows, nothing has been optimized. It has just been added to.

The right strategy is not a plant-wide choice. It is decided asset by asset, and often failure mode by failure mode, because the same machine can need different care in different places. One pump might get sensor-based monitoring on its bearing while a cheap backup seal is simply run until it fails. The table below is a starting point for that decision.

One quick note on the run to failure strategy. While a valid strategy, it should not cover for lack of planning. This needs to be selected as long as it is a deliberate choice with the spare already on the shelf and the consequences understood in advance.

The framework that produces these assignments across a whole fleet is reliability centered maintenance, which weighs each failure mode against its consequence. Adjacent methods like total productive maintenance push routine care onto operators through autonomous maintenance, but they support the strategy mix rather than replace the per-asset decision.

You cannot run a full failure analysis on every asset in a plant that has several thousand of them, and you do not need to. Ranking assets by criticality is what makes the work manageable, because it tells you where to spend your analysis effort first. A workable score needs three inputs at minimum.

• Consequence for production: what the failure costs in lost output, measured in tonnes or barrels per day, not a generic severity label.
• Consequence for safety and environment: whether a failure puts people, permits or the environment at risk.
• Detectability: whether the failure gives some warning, or arrives with none.

Here is the part most plants miss. Criticality was usually scored once at commissioning and never looked at again. An asset that was a minor player at start-up may now sit on your highest-margin product line, and its maintenance plan has never been revised to match. Re-ranking against how the plant runs today is often a bigger win than any single change of strategy.

A good strategy on paper still has to survive contact with the plant floor. Even a perfectly chosen plan saves nothing if the work cannot get done on time, and three things tend to get in the way.

1. Scheduling

A task falls due, but the job slips because the maintenance window was never lined up with the production plan. The number to watch is schedule compliance, the share of planned work orders finished in the period they were scheduled for. When it runs low, the strategy is fine and the calendar is not. As this repeats, work piles up, and a growing maintenance backlog is usually the first visible sign.

2. Parts

The task is ready and the part is not. When spares are not staged and kitted before the scheduled date, the technician arrives, finds the gap, and the job is pushed. Getting parts ready ahead of the schedule is one of the fastest execution fixes there is.

3. Wrench time

This is usually the biggest hidden cost of all. Wrench time, the share of a shift a technician actually spends with tools on the equipment, tends to be lower than people expect. Across industry it runs about 25 to 35 percent, against a world-class mark near 55 percent. The rest is lost to travel, hunting for documentation, and above all waiting on permits and isolations, where a technician can lose the first hours of a shift before touching a tool.

These three gaps do more than delay a single job. What a technician records at the asset, or fails to record, becomes the data your next round of decisions is built on. Capture completion cleanly in the field, and the next optimization cycle starts from reality. Miss it, and you are tuning the plan on guesswork. In other words, good execution is not just the result of optimization. It quietly feeds the next round of it.

Optimization runs on data, so the first question is whether you have enough of it to trust the answers. Before any analysis, make sure five things are actually in place.

• An asset register with current criticality, not the criticality set at commissioning.
• Twelve to twenty-four months of work order history.
• Failure codes applied consistently enough to group by failure mode.
• Cost captured at the work order, for both labor and parts.
• The current PM task list, exported from your system of record.

If failure coding is missing or inconsistent, fixing the coding standard is your first project, not the strategy analysis. Analysis built on unreliable failure data produces confident, wrong answers.

With the data in place, the next discipline is scope. Resist the urge to do everything at once. One critical asset class at one site, taken from end to end, beats a site-wide effort that stalls at the analysis stage. End to end means analyzed, changed in the system of record, run for a full maintenance cycle, and measured against the baseline. Expect that first pass to take a few months, not a few weeks, because it only counts once a full cycle has actually been completed. From there, the detailed work of reviewing each preventive task, keeping the ones that earn their place and dropping the ones that do not, is the heart of preventive maintenance optimization.

Sooner or later you have to prove the program is working, usually to a plant manager or to finance. The catch is that most maintenance numbers can be nudged in the right direction without the underlying work getting any better. So no single metric can be trusted on its own. The table shows what real improvement looks like for four common signals, and how each one can mislead you.

The key is to read these in pairs, never alone. Track PM hours next to compliance, and cost next to reliability. On its own, each number can flatter the program, but together they show what is really happening. Planned maintenance percentage is a good place to start. Overall equipment effectiveness matters too, but treat it as an outcome you watch rather than a number you control directly.

Optimization gains erode quietly. After each incident, a new task gets added. The addition almost never comes with a removal, so over two or three years the task list grows back to its old size. This is PM re-inflation. None of it is anyone's fault: every task was added for a good reason at the time. The problem is that no one is watching the total.

The trouble is that re-inflation is hard to spot while it is happening. The clearest early warning is simple. Track total PM labor hours next to PM compliance. If hours keep climbing while compliance stays flat or improves, the task list is quietly re-inflating. Three controls keep it in check:

• A named owner for the maintenance strategy, so keeping the task list honest is somebody's actual job.
• A change-control gate, so any new task has to name the failure mode it addresses and either replace an existing task or justify the extra hours.
• A review triggered by real events, a failure, a process change, a shift in criticality, rather than a once-a-year calendar slot.

Run this way, as a managed capability under a framework like ISO 55000, optimization holds. Run as a one-time project, it slips straight back.

Maintenance cost is not a single number you can trim. It is four separate costs, each with a different owner and a different cause. That is why a blanket order from finance to cut maintenance spend by ten percent usually lands on the wrong one. Once you separate the four, each has a clear fix.

Where you start matters. The first three costs are largely inside your control and respond quickly to better planning and execution. Lost production is the biggest prize but the hardest to prove, so let the labor, contractor and inventory savings carry the business case. The practical moves for each are covered in how to reduce maintenance costs.

What Closing These Leaks Looked Like at Indorama Ventures

Indorama Ventures, one of the world's largest producers of PET and chemicals, put exactly this into practice across its maintenance operations. The results show what happens when the four leaks are closed together rather than one at a time.

• A target of $50 million a year in maintenance savings, with $19 million in EBITDA savings already realized in 2025.
• Maintenance backlog cut by 58 percent.
• Outside contractor headcount reduced by 38 percent.

Those numbers came from changing how work was planned, scheduled and executed, not from cutting maintenance across the board.

A refining unit runs about 60 centrifugal pumps across two process trains, all on the same 90-day preventive interval inherited from commissioning. Six failures in a single year triggered a review. Here is how the optimization played out, step by step.

1. Re-rank by criticality: The pumps were scored against how the unit runs today. A handful sat on the site's highest-margin product line and had quietly become critical since start-up.
2. The split verdict: The single interval was wrong in both directions at once. It was too frequent for the low-duty pumps, which were being torn down for no reason, and too rare for the high-duty pumps, which were failing before the service came due.
3. Split the strategy: The critical high-duty pumps moved to condition-based monitoring (CBM). Two backed-up, low-consequence pumps moved to run to failure, with spares kept on the shelf.
4. Stratify the intervals: The remaining pumps were regrouped into three interval bands, each set from that group's real failure history rather than one shared number.
5. Fix the execution: None of this lands without execution, so parts were staged and kitted before each scheduled date, and permits and isolations were cleared ahead of the window.

The payoff was concrete. Over the following year the two trains freed up roughly 900 technician-hours, cut their backlog by several weeks of work, and brought unplanned pump failures down from six to one. A couple of the high-duty pumps actually ended up getting more attention than before, which is the whole point: the work was matched to each asset, not simply cut across the board.

The strategy decision is only half of maintenance optimization, and it is the easier half. Matching the right approach to each asset is a solvable analysis problem. Whether the plant can then execute that plan, and whether anyone still owns it after the project closes, is what actually decides the outcome.

The programs that hold their gains treat optimization as a continuous cycle of measuring, deciding and adjusting to how the assets really behave, and they make it somebody's standing job. In practice, that means closing the insight-to-action gap, the distance between spotting a problem and actually resolving it in the field. That gap is where most of the savings live, and it is the difference between a program that saves money once and one that keeps saving it.

Every problem in the scheduling and execution section above comes down to handoffs between systems and people. The plan lives in your ERP or EAM. The work happens in the field. The space in between, where jobs wait on parts, permits and paperwork, is where good plans quietly come apart. That is the gap Innovapptive is built to close.

It does not decide your maintenance strategy for you, and it does not replace your system of record. The connected worker platform sits on top of the systems you already run, turning the plan into work your technicians can actually complete and feeding clean data back so the next round of decisions starts from reality. In practical terms, that maps onto the gaps mentioned above:

• Planning and scheduling lines maintenance work up against the production plan, so jobs are scheduled when the plant can actually take them.
• Mobile maintenance puts work orders, permits and inspections on a phone or tablet, so technicians spend less of the shift on travel and paperwork and more of it on the asset.
• Inventory and warehouse management makes sure the right parts are staged and kitted before a job is due, instead of being found missing when the technician arrives.
• Operator rounds let operators log routine checks and early warning signs from the floor, catching small problems before they turn into failures.

As Indorama Ventures found at its Port Neches site, described earlier, gains on this scale came without ripping anything out. The platform activated their existing SAP and Maximo systems rather than replacing them, which is what made a change this large possible inside a live plant.

That breadth, covering both the moment a problem is spotted and the work that resolves it, is also why Frost & Sullivan named Innovapptive a leader in its 2025 Frost Radar for connected worker platforms, pointing to the depth of its execution and AI capabilities.