Predictive Maintenance Gives You 30 Days. Now what?
In Part 1 of our predictive maintenance series, we looked at how manufacturers have moved from historical failure models and fixed maintenance schedules towards condition monitoring and AI-assisted predictions. The technology has changed considerably, but the objective remains familiar: identify a developing problem early enough to act before it becomes a failure.
So what happens when predictive maintenance does exactly that?
Imagine a bearing on a critical production asset.
The bearing hasn't failed. Production is still running. But the condition data tells a different story. Vibration has been increasing, and the available evidence suggests that the bearing is deteriorating. Maintenance expects that an intervention will probably be required within approximately 30 days.
This is exactly what predictive maintenance is supposed to achieve.
Instead of discovering the problem when the machine stops, the maintenance team has advance warning. Thirty days to prepare, plan the intervention and avoid an unplanned shutdown.
Thirty days sounds like plenty of time.
Until someone has to determine exactly which bearing needs replacing. Find the correct material record. Check whether it is actually in stock. Find out whether another site has one. Verify that the part is still available from the manufacturer. And, if nobody in the organisation has it, source and deliver it before the intervention.
Suddenly, 30 days doesn't sound quite so long.
In Part 2, we follow that prediction beyond the machine. Because predictive maintenance can give you time. Your spare parts processes determine how much of that advantage you keep.
First: What Exactly Needs Replacing?
A predictive maintenance system can identify signs of degradation. Maintenance and reliability teams can assess those signals and determine whether an intervention is required.
But Procurement cannot order "increasing vibration".
At some point, information about the condition of an asset has to become a precise material requirement.
Ideally, that connection already exists. Equipment and maintenance Bills of Materials, or BOMs, are designed to connect technical assets with their components. SAP, for example, describes an equipment BOM as a structure that can be used to assign spare parts to equipment for maintenance purposes. Its maintenance functionality also allows material components from a BOM to be assigned directly to maintenance tasks and orders.
In practice, however, that connection is only as reliable as the data behind it.
The material description may be incomplete. The manufacturer number may be missing. The same bearing may exist under several material numbers. A component may have been superseded without the information being updated consistently. Or the equipment may have been operating for decades while its spare parts data has evolved across different systems, sites and naming conventions.
This creates an uncomfortable situation: the organisation has invested in technology that can detect degradation earlier, but then loses valuable time trying to determine which spare part is actually required.
The faster you detect a potential failure, the more important it becomes that the data connecting the asset to the spare is ready to use.
This is where clean spare parts master data becomes part of predictive maintenance readiness. With SPARROW.Clean, manufacturers can harmonise and enrich material data, identify duplicate records and improve the information needed to find and manage the correct spare.
The prediction came from the machine. Acting on it depends on the data around the machine.
Do We Already Have the Part?
Suppose we have now identified the correct bearing.
The next question seems straightforward: Is it in stock?
For an enterprise manufacturer, however, that question is too narrow.
The better question is: Do we have it somewhere?
Imagine the plant where the bearing is deteriorating has no replacement in its warehouse. A conventional local stock check shows zero.
Another plant, however, has two.
From the perspective of the first warehouse, there is a shortage. From the perspective of the organisation, there may not be one at all.
This isn't simply a theoretical possibility. Spare parts inventory research has long examined lateral transshipments and inventory pooling as alternatives to holding every required component independently at every location. Research by Hartanto Wong and colleagues, for example, has modelled repairable spare parts systems in which stock can be transferred between locations when a local stockout occurs.
For manufacturers operating dozens or hundreds of sites, that changes the meaning of availability.
A spare part sitting unused in another warehouse is potentially part of your maintenance response, but only if you know it exists, can identify it as the same or a suitable part, and can move it within the available timeframe.
A local spare parts shortage is not necessarily an organisational shortage.
Our customer Royal Avebe provides a good case study of why this distinction matters. The company operates three core production sites in the Netherlands located within 50 km of each other. Yet each site historically had its own stockroom organisation, inventory policies and data structure, with no centralised planning or visibility across sites. Despite their proximity and overlapping operations, MRO inventories were effectively managed separately.
Working with SPARROW, Royal Avebe began bringing spare parts data and inventory management across the three sites together. SPARROW.Stock was introduced to digitise inventory and improve visibility, while the broader project laid the foundation for shared inventory planning across the sites.
The example illustrates a broader point: enterprise-wide inventory transparency can turn stock held at one site into a potential maintenance resource for another. Instead of every plant solving the same availability problem independently, manufacturers can first determine whether the required part already exists somewhere in their network.
Predictive maintenance may tell you that a component will probably be needed soon. Network visibility can reveal that the solution is already sitting on one of your own shelves.
Can You Get It Before the Machine Needs It?
Now suppose nobody has the bearing.
The question changes again.
Not simply: Can we buy it?
But: Can we get it in time?
If Maintenance has approximately 30 days to prepare and the part can be delivered within five days, the organisation has room to plan.
If current delivery takes 25 days, that margin becomes considerably smaller.
If it takes 60 days, the predictive maintenance system may have worked exactly as intended and the plant could still face downtime.
Lead time is therefore not just a Procurement parameter. In this context, it becomes part of the maintenance risk.
Research into reliability-centred spare parts inventory makes this connection explicit. Spare parts stocking decisions need to balance the cost of holding inventory against the risk and cost of waiting for a part when equipment requires it, with lead time among the information required to make that decision.
There is another complication: the lead time stored in the ERP may not be the lead time you would get today.
Lead times change as supplier capacity, sourcing routes, availability and market conditions change. Yet many organisations still work with lead-time values entered into their ERP and updated only periodically.
That creates a dangerous mismatch.
The predictive maintenance system may be working with continuously changing information about the machine while the spare parts decision relies on a static assumption about how quickly the replacement can be obtained.
The relevant question is therefore not simply: What lead time is stored in SAP?
It is: How long would it actually take us to get this part now?
Where Did Your 30 Days Go?
This is where the promise of predictive maintenance meets operational reality.
You started with approximately 30 days of warning.
Some of that time may be spent confirming which component needs replacing. More may disappear resolving incomplete or conflicting material information. Then the organisation searches its warehouses. Procurement contacts the supplier and discovers that the delivery time in the ERP no longer reflects current availability.
The predictive maintenance system hasn't failed.
It did its job.
But every delay between prediction and action reduces the advantage it created.
As SPARROW CEO Meir Veisberg puts it:
“A 30-day failure warning only has value if 30 days are enough to act. If it takes a week to identify the correct spare, another week to understand whether you already have it somewhere in the organisation, and the supplier then quotes a six-week lead time, you still have a downtime problem. Predictive maintenance needs to be connected to spare parts readiness.”
This is also why the P-F interval we introduced in Part 1 matters beyond condition monitoring. The interval between detecting potential failure and reaching functional failure is not only time for Maintenance to diagnose the problem. It is also time to prepare the intervention.
From a spare parts perspective, that means the warning needs to be long enough to identify, locate or source the required component.
And sometimes it simply isn't.
What If 30 Days Were Never Going to Be Enough?
Suppose our bearing is critical.
There is no suitable replacement anywhere in the organisation. Current supplier lead time is 60 days. The condition data gives Maintenance approximately 30 days of actionable warning.
Even perfect material data and an efficient Procurement team cannot make those numbers fit.
The obvious conclusion might be that the bearing should have been held in stock all along.
But that creates another problem.
Manufacturers cannot stock every spare part that might one day be needed. Many MRO components are expensive and slow-moving, and some may never be used before becoming obsolete. Research into spare parts inventory therefore treats stocking as a trade-off between availability and the cost of holding inventory rather than simply recommending that more parts be kept on the shelf.
The real question is whether the risk associated with this component justifies stocking this component.
And that calculation should not rely on information that stopped changing when someone entered it into SAP.
This is where SPARROW.Plan takes the same dynamic principle we have discussed in predictive maintenance and applies it to spare parts planning.
SPARROW.Plan uses reliability-based methods and operational data to determine which parts should be stocked and in what quantities. Crucially, this includes current lead-time information rather than relying solely on static lead times stored in the ERP.
Imagine that a component did not justify stocking when it could reliably be sourced within 15 days. Its current lead time is now 60 days, while the realistic warning provided by condition monitoring may only give Maintenance around 30 days to prepare.
The component hasn't changed.
The risk has.
And the stocking recommendation should be able to change with it.
There is good academic precedent for connecting maintenance information with more dynamic inventory decisions. Researchers Sha Zhu, Willem van Jaarsveld and Rommert Dekker studied the use of planned condition-based maintenance as advance demand information for spare parts. Their research explicitly proposes more forward-looking inventory control based on information about upcoming maintenance requirements.
We will return to this idea later in the series when we explore what predictive maintenance could mean for predictive spare parts planning.
Predictive Maintenance Needs Spare Parts Readiness
Predictive maintenance is often discussed as a technology challenge.
Can we detect degradation earlier? Can we distinguish a genuine anomaly from noise? Can machine learning improve Remaining Useful Life estimates?
All of those questions matter.
But once a developing failure has been identified, another chain of decisions begins.
Which component needs replacing? Can we reliably identify the spare? Is it available locally? Does another site have it? Can it be moved in time? If nobody has it, what is the actual current sourcing lead time? And if sourcing will take longer than the warning period, should that part have been stocked in the first place?
The machine doesn't care whether those questions belong to Reliability, Maintenance, Master Data, the warehouse or Procurement.
From its perspective, there is only one outcome that matters: whether the intervention can happen before degradation becomes failure.
Predictive maintenance is becoming increasingly dynamic. The information around spare parts needs to keep pace.
Knowing Earlier Isn't the Same as Being Ready Earlier
Our bearing gave us approximately 30 days.
Whether those 30 days prevent downtime depends on everything that happens after the prediction.
Better failure prediction creates valuable time. Clean spare parts data, inventory transparency, cross-site visibility, current sourcing information and the right stocking strategy determine whether the organisation can use it.
Knowing earlier is valuable. Being ready earlier is what prevents downtime.
Next in the series: How Does a Machine Know It's About to Fail?
In Part 3, we go back to the shop floor to explore the sensors, signals and condition-monitoring technologies that can reveal degradation before functional failure occurs.
