Move from emergency repairs to planned action by connecting maintenance, production, spare parts, and real machine performance.
“The most expensive machine failure is not the repair bill. It is the production, customer trust, and margin lost while the machine is unavailable.”
Every factory has maintenance work.
Machines need inspection. Bearings wear out. Belts loosen. Sensors fail. Tools require replacement. Lubrication must be completed. Software needs updates. Safety checks must be documented.
The real question is not whether maintenance will happen.
It is whether maintenance happens before the machine stops at the worst possible moment—or after it has already disrupted production.
Many manufacturers still operate in one of two ways.
The first is reactive maintenance: something breaks, the team reacts, production stops, and spare parts are ordered urgently.
The second is calendar-based preventive maintenance: a task is scheduled every month, quarter, or number of operating hours, whether the machine needs attention or not.
Both approaches have value. But both have limits.
Predictive maintenance adds a third option: use real machine, production, quality, and service data to identify risk before failure occurs.
Why Unplanned Downtime Is So Expensive
When a machine fails, the repair cost is often only one part of the problem.
Unplanned downtime can create:
- Delayed production orders.
- Missed customer delivery dates.
- Overtime and shift disruption.
- Expedited freight costs.
- Emergency spare-part purchases.
- Idle labour.
- Wasted material.
- Scrap and rework.
- Production rescheduling.
- Lost customer confidence.
- Margin loss.
- Safety risk.
A production line may stop for two hours, but the operational effect can last several days.
The maintenance team may repair the machine quickly. But planning still needs to update the schedule. Purchasing may need to find parts. Sales may need to inform customers. Warehouse teams may need to change shipment priorities.
A machine failure is never only a maintenance issue. It is a business event.
Reactive, Preventive, and Predictive Maintenance
Understanding the difference between the three models helps companies decide where to improve.
Reactive Maintenance
Reactive maintenance means fixing equipment after it fails.
This may be acceptable for low-cost, non-critical assets. For example, replacing a simple light, small fan, or low-value tool after failure may be cheaper than monitoring it continuously.
But reactive maintenance becomes expensive when the machine is a production bottleneck, safety-critical asset, or key resource with long repair time.
Preventive Maintenance
Preventive maintenance is planned on a calendar or usage basis.
For example:
- Change oil every six months.
- Inspect a machine every 1,000 operating hours.
- Replace a part every 12 months.
- Complete a safety check every quarter.
This reduces unexpected failures and creates discipline. But it can also lead to unnecessary maintenance if parts are replaced too early or machines are stopped when there is no real risk.
Predictive Maintenance
Predictive maintenance uses real data to estimate when maintenance is likely to be needed.
It can consider:
- Machine runtime.
- Cycle time.
- Temperature.
- Vibration.
- Energy use.
- Pressure.
- Error codes.
- Sensor readings.
- Maintenance history.
- Spare-part consumption.
- Quality results.
- Scrap rates.
- Operator notes.
- Production load.
- Asset age.
The goal is not to predict every failure perfectly.
The goal is to identify changing conditions early enough to plan a safer, cheaper response.
The Connection Between Maintenance and Production
Maintenance planning should not sit outside production planning.
A maintenance manager may know that a machine needs inspection. But if the production planner does not know, the factory may schedule an urgent customer order on that machine. If purchasing does not know, the required spare part may not be available. If sales does not know, a customer may receive a delivery promise that cannot be kept.
SIX ERP can connect maintenance with:
- Production orders.
- Work-centre schedules.
- Machine capacity.
- Labour availability.
- Spare-parts inventory.
- Purchasing.
- Quality records.
- Supplier lead times.
- Customer delivery commitments.
- Cost and margin data.
This allows the company to schedule maintenance at the best possible time.
For example, instead of stopping a machine during a high-priority production run, the system may identify a low-load period, confirm spare-part availability, reserve technician time, and prepare a maintenance task before the machine becomes critical.
The best maintenance plan protects both the machine and the customer promise.
The Signals That Matter
Predictive maintenance does not always require advanced sensors or expensive equipment.
Many useful signals already exist inside the business.
Operational Signals
These can include:
- Longer cycle times.
- Increased setup time.
- Reduced output.
- More frequent stoppages.
- Higher changeover time.
- Lower machine utilisation.
- Repeated error messages.
- Higher maintenance-call frequency.
Quality Signals
These can include:
- Increased scrap.
- More rework.
- More failed quality checks.
- Measurement drift.
- Higher reject rate from one machine.
- Increased customer complaints.
- Repeated product defects.
Maintenance Signals
These can include:
- Frequent replacement of the same spare part.
- Shorter time between repairs.
- Recurring technician notes.
- Maintenance tasks completed late.
- Repeat failure after repair.
- Increasing service cost per machine.
Inventory and Purchasing Signals
These can include:
- Low stock of critical spare parts.
- Long supplier lead times.
- Rising spare-part price.
- Supplier quality issues.
- Delayed spare-part deliveries.
- Emergency purchase history.
The first step is not installing more sensors. The first step is connecting the data you already have.
A Practical Example: The Machine That Is Not Yet Broken
Imagine a factory with a critical cutting machine.
The machine is still operating. It has not stopped. But several small changes are visible:
- Cycle time has increased by 8%.
- Vibration readings are gradually rising.
- Scrap has increased on one product family.
- The same bearing has been replaced twice in the last year.
- The required replacement bearing has a supplier lead time of 12 days.
- The machine is scheduled for high-priority production next week.
A reactive approach waits for the breakdown.
A preventive approach may wait for the next scheduled maintenance date.
A predictive approach identifies the combined risk.
SIX ERP can help prepare a recommendation:
“Machine C-14 shows a rising maintenance risk based on cycle time, vibration, scrap rate, and prior bearing replacements. Schedule a three-hour inspection on Thursday. Required bearing is available in stock. Performing maintenance before next week’s production run reduces the risk of an unplanned stop affecting two high-priority customer orders.”
The maintenance manager still decides. But the decision is supported by evidence.
Predictive maintenance turns weak signals into planned action.
Spare Parts Must Be Part of the Plan
Predictive maintenance fails if the required parts are not available.
A maintenance recommendation is only useful when the business can answer:
- Do we have the spare part in stock?
- Is it in the correct warehouse?
- Is it reserved for another machine?
- Is it an approved replacement?
- Is there a substitute?
- What is the supplier lead time?
- What is the cost?
- Does the part have a shelf-life or storage requirement?
- Can it be transferred from another site?
- Should we order a safety stock?
SIX ERP can manage spare parts as controlled inventory items with:
- Item codes.
- Machine compatibility.
- Supplier details.
- Warehouse locations.
- Reorder points.
- Batch or serial tracking where required.
- Minimum stock.
- Purchase history.
- Cost.
- Approved substitutes.
- Maintenance-kit relationships.
A machine cannot be repaired with a dashboard. It needs the right part at the right time.
Maintenance Work Orders Create Accountability
Predictive maintenance should not remain a recommendation in a report.
It should become a controlled maintenance work order.
A maintenance work order can include:
- Machine or asset.
- Maintenance task.
- Priority.
- Reason for the task.
- Risk level.
- Required spare parts.
- Required tools.
- Safety instructions.
- Assigned technician.
- Planned date and duration.
- Production impact.
- Completion status.
- Actual labour time.
- Parts consumed.
- Notes and photos.
- Follow-up tasks.
- Cost.
This creates accountability and provides a historical record for future analysis.
If the same issue returns, the company can see what was repaired, which part was used, who completed the work, and whether the action solved the problem.
Maintenance history is not paperwork. It is machine intelligence.
AI and Predictive Maintenance
AI can improve predictive maintenance by finding patterns across large amounts of data.
For example, an AI model may identify that a certain machine is likely to need maintenance when several conditions appear together:
- Energy consumption rises.
- Cycle time increases.
- Scrap rises above a threshold.
- Vibration changes.
- A specific material is used.
- The machine has completed a certain number of operating hours.
- A maintenance task was delayed.
AI can also help compare similar machines across different sites.
It may identify that one machine is consuming more spare parts than similar equipment. Or that one production line has a higher reject rate after a particular maintenance interval.
But AI should not make uncontrolled decisions.
It should provide:
- A clear alert.
- The data behind the alert.
- A confidence level.
- Suggested actions.
- Expected production impact.
- Required approval.
AI should help maintenance teams see earlier. It should not replace engineering judgment.
Predictive Maintenance and Quality Go Together
Machine health and product quality are closely connected.
A machine may continue to run even when its performance is slowly deteriorating. The first visible sign may not be a breakdown. It may be poor-quality output.
For example:
- A worn tool may create incorrect dimensions.
- A temperature change may affect coating quality.
- A worn bearing may create vibration and product defects.
- A sensor problem may lead to incorrect filling or dosing.
- A pressure issue may affect sealing or bonding.
- A calibration problem may cause repeated quality failures.
When maintenance, production, and quality data are connected, the business can see these patterns earlier.
A quality trend can trigger a maintenance inspection. A maintenance event can trigger increased quality checks. A supplier-material change can be compared with machine performance and product results.
The best time to repair a machine may be before the customer sees the defect.
Measuring Maintenance Performance
Predictive maintenance should be measured through business outcomes, not only the number of maintenance tasks completed.
Useful measures include:
- Unplanned downtime.
- Planned maintenance completion rate.
- Mean Time Between Failures.
- Mean Time To Repair.
- Overall Equipment Effectiveness.
- Maintenance cost per machine.
- Emergency spare-part purchases.
- Spare-part stockout rate.
- Repeat failure rate.
- Scrap and rework linked to equipment.
- Schedule adherence.
- Production output after maintenance.
- Customer-delivery impact.
These measures show whether maintenance is creating real value.
A maintenance team should not be judged only by how quickly it repairs failures. It should be judged by how effectively it prevents them.
How SIX ERP Supports Predictive Maintenance
SIX ERP can connect the information needed to move from reactive maintenance toward a more predictive model.
The platform can support:
- Asset and machine records.
- Maintenance schedules.
- Maintenance work orders.
- Technician assignments.
- Service history.
- Spare-parts inventory.
- Reorder planning.
- Purchasing and supplier information.
- Production schedules.
- Work-centre capacity.
- Machine downtime records.
- Quality-inspection results.
- Scrap and rework tracking.
- Cost analysis.
- Mobile service and maintenance processes.
- Approval workflows.
- Dashboards and alerts.
- Integration with machine, sensor, IoT, or MES data where available.
This gives maintenance teams a connected operational view rather than a separate maintenance spreadsheet.
How to Start
Start with one critical asset.
Choose a machine that:
- Creates a production bottleneck.
- Has frequent downtime.
- Produces high-value goods.
- Has expensive or long-lead spare parts.
- Creates quality risk.
- Affects important customer delivery dates.
Then collect the available data:
- Maintenance history.
- Downtime records.
- Spare-part use.
- Production output.
- Scrap and quality results.
- Operator notes.
- Supplier lead times.
- Machine runtime.
Create a simple risk score. Review it weekly. Compare predicted risk with actual maintenance findings.
As the process improves, add more machines, better data, and—where useful—sensor integration and AI analysis.
The goal is not to create a futuristic factory overnight. The goal is to prevent one costly failure before it happens.
The Future Is Planned, Not Reactive
Factories will always need maintenance teams.
Machines will always wear, fail, and require attention.
But the strongest manufacturers will not wait for the breakdown.
They will connect machine data, maintenance history, spare parts, production schedules, quality records, and customer commitments. They will see risks earlier. They will plan maintenance better. They will protect capacity and delivery performance.
Predictive maintenance is not about guessing the future. It is about using real operational data to make fewer expensive surprises.
SIX ERP helps manufacturers connect maintenance, production, inventory, purchasing, quality, and service processes into one practical operational system.
