How to Forecast Maintenance Inventory Without Complications

Why forecasting spare parts is so difficult

Forecasting maintenance inventory is fundamentally different from forecasting production materials or retail demand.

“Most spare parts follow intermittent or lumpy demand patterns. A sensor might not be used for months and then suddenly three fail in a month,” says Jason Afara, a former maintenance manager in food manufacturing, now a Director at Fiix.

Failures, shutdowns, and overhauls are event-driven, not continuous consumption.

That’s why forecasting methods used in supply chain planning often struggle in maintenance environments. Instead of chasing perfect predictions, the most effective maintenance organization focus on three practical principles:

1. Define stocking policies based on risk and asset criticality
2. Use maintenance demand signals, not just past usage
3. Apply simple forecasting methods that fit each part type

Let’s walk through how this works.

Step 1: Start with criticality-based spare parts policy

One of the biggest mistakes in MRO inventory planning is trying to forecast every part the same way.

“Not all spare parts carry the same risk. A missing PLC module can stop an entire production line. A missing gasket might delay repairs but not halt operations,” says Jason.

He adds that if you treat both parts the same in your forecasting model, you’re wasting effort and creating risk. A better approach is to segment spare parts into criticality categories and define stocking policies for each.

A simple A–B–C–D framework works well.

A: Line-down or safety-critical parts

These are components that can shut down production or create safety risks. Examples include:

• PLC modules
• Critical bearings
• Major drive components
• Safety system parts

For these parts, the goal is near zero stockouts. That typically means:

• Higher safety stock
• Faster reorder triggers
• Close monitoring of supplier lead times

Inventory cost matters but downtime cost matters more.

B: Important but not catastrophic

These parts impact operations but won’t necessarily stop production. Examples might include:

• Secondary motors
• Standard sensors
• Maintenance-critical consumables

Here the goal is balancing cost and availability. Stock levels should support maintenance schedules without excessive inventory.

C: Regularly used parts

These are frequently consumed items such as:

• Filters
• Belts
• Lubricants
• Fasteners

These parts are ideal candidates for usage-based forecasting. Because demand is more consistent, it’s easier to predict.

D: Low-risk or non-critical parts

These parts carry minimal operational risk. For many of them, the best policy may be:

• Order on demand
• Keep minimal stock
• Avoid forecasting entirely

This step is important. If a part doesn’t justify stocking, forecasting it is a waste of effort.

Why these steps matter

By defining a policy first and forecasting second, maintenance teams avoid overcomplicating inventory planning. This approach aligns spare parts stocking with system availability risk, rather than simply past usage.

Step 2: Forecast using maintenance demand signals

Once parts are categorized, the next challenge is determining demand.

“Many organizations rely on a single data source: issue history from the storeroom. But this alone is rarely enough,” says Jason.

Because spare parts demand is tied to maintenance events, the most reliable forecasts come from combining multiple demand signals. Below are the signals that matter most.

Consumption history

Historical usage is still valuable. It shows:

• Actual part withdrawals
• Replacement patterns
• Seasonal maintenance demand

But on its own, it tells an incomplete story. Past usage reflects what happened, not necessarily what is planned. Consumption is not equal to demand; basing stock on consumption could be the incorrect metric because consumption could be substitutions or last second changes to the part.

Preventive maintenance schedules

Preventive maintenance programs generate predictable demand. If a maintenance task requires:

• 2 filters
• 1 belt
• 1 gasket

…then every scheduled PM creates a known parts requirement.

“Forecasting based on PM counts multiplied by bill-of-material quantities is often more accurate than historical usage alone,” says Jason.

Predictive maintenance alerts

Condition monitoring tools such as:

• Vibration analysis
• Thermal monitoring
• Oil analysis

…provide early warning signs of failures. These signals help forecast upcoming spare part demand before failures occur.

Asset runtime and production volume

Some components wear based on operating hours or production cycles.

Tracking:

• Machine runtime
• Production throughput
• Start/stop cycles

…can reveal when parts will likely need replacement.

Mandatory overhaul intervals

Major equipment overhauls often require large quantities of spare parts. Planning for these events months in advance ensures parts are available without last-minute ordering.

Work order backlog

Planned maintenance jobs that haven’t yet been executed still represent future demand. If parts are tied to work orders, maintenance teams gain visibility into upcoming inventory needs.

Supplier lead times

Even if demand is known, replenishment risk still matters. Key questions include:

• How long does it take to get to the part?
• Are suppliers reliable?
• Are there single-source dependencies?

Long lead times increase the risk of stockouts, especially for critical parts.

Step 3: Use simple forecasting methods that fit each part type

Once demand signals are identified, the next step is choosing forecasting methods.

“This is where organizations get kind of stuck; stats’ models exist for demand forecasting, but they aren’t always the best or necessary,” says Jason.

In most environments, he adds that simple, fit-for-purpose approaches work best.

For regularly used parts (C items)

Parts with steady demand can use simple statistical techniques such as:

• Moving averages
• Exponential smoothing

These methods smooth out fluctuations and provide reasonable forecasts with minimal complexity.

For intermitted or critical spare parts (A and B items)

Demand for these parts is often unpredictable. Instead of forecasting for exact quantities, it’s often used to track:

• Time between failures
• Size of demand when it occurs

These insights help estimate the likelihood of future demand and determine appropriate safety of stock levels. Some advanced approaches, such as Croston-style forecasting (Syntetos et al) are designed specifically for intermittent demand.

But even without complex models, understanding demand intervals can significantly improve planning.

For planned replacement parts

Some parts follow strict maintenance intervals. Examples include:

• Filters replaced every 3 months
• Belts replaced every 1,000 operating hours
• Components replaced during scheduled overhauls

For these parts, forecasting is straightforward:

This method is often reliable because demand is driven by scheduled work.

Step 4: Set reorder points based on lead time and service level

Forecasts alone don’t determine when to order spare parts. That’s where reorder points come in. A commonly used formula is:

This formula balances availability and cost.

Expected demand during lead time

This represents the number of parts likely to be used while waiting for a supplier’s delivery. For example, if a bearing supplier has a 30-day lead time and expected demand is 2 bearings per month, then the case reorder point would be 2 units.

Safety stock

Safety stock protects against uncertainty. It covers situations such as:

• Unexpected failures
• Supplier delays
• Forecast errors

The amount of safety stock depends heavily on part criticality. Critical A items usually require higher safety stock levels. Lower- risk parts may require little or none. A key concept in determining the right level is the protection period, which represents the total span of time during which demand can occur, but cannon be met with new replenishment.

In other words, the protection period is the total time your inventory must cover demand before replenishment can protect you. During this window, demand continues, yet your team cannot respond in real time with fresh supply. Your safety stock must therefore be sized to absorb this exposure. The formula to calculate your protection period is:

“Setting reorder points is ultimately about choosing a service level. Higher service levels cut the risk of stockouts but increase inventory investment, lower service levels reduce cost but increase the risk of downtime…the right balance depends on the operational impact of each spare part,” says Jason.

Step 5: Avoid the data traps that break MRO planning

Even the best forecasting approach will fail if the underlying data is unreliable. Several common data issues frequently undermine spare parts planning.

Step 6: Hold a monthly MRO planning review

Even the best forecasting systems benefit from human insight. Many successful maintenance organizations hold short monthly MRO review meetings involving:

• Maintenance planners
• Storeroom managers
• Reliability engineers

These meetings often take 15-30 minutes but provide valuable alignment. Typical discussion topics can include covering upcoming shutdowns, critical spares at risk of stockout, supplier lead time changes and more. The conversation helps catch issues that no spreadsheet or forecasting model can predict.

Step 7: Track KPIs to know if the plan is working

Forecasting improvements should translate into measurable results.

Failures are unpredictable, suppliers change, and production environments evolve. But maintenance teams don’t need perfect predictions.

“Maintenance teams need fewer surprises,” says Jason. According to him, by combining:

• Criticality-based stocking policies
• Maintenance-driven demand signals
• Simple forecasting methods
• Clear reorder point strategies

Jason adds some additional insight into his experience and how it can be taxing, adding that “ideally you are improving your PM program, which in fact improves your inventory. You should start slow, think about 1 production line, or 1 type of asset, and work your way through it all.”

He also highlights that AI could potentially help with data analysis or searching for other vendors that may also have that stock.

“Stock could also been obsoleted by the OEM, so you might need to search eBay or be creative in ensuring you have those critical spares,” says Jason.

Maintenance managers can dramatically reduce emergencies without turning inventory into a complex analytics project. When forecasting works well, the benefits ripple across the entire operation. That looks like:

• Less downtime
• Fewer expedited shipments
• Better maintenance planning
• Lower inventory stress

Which means fewer midnight phone calls, and a lot fewer conversations with finance about storeroom value.

Step 8: Use advanced forecasting models only where they help

By now, it should be clear that not every spare part should be forecasted the same way.

Since Croston’s 1972 breakthrough, which split intermittent demand into quantity and interval, a lineage of evolving models has emerged. Today, we can move beyond simple historical trends by integrating service maintenance data and installed-base signals, to ensure parts are available when equipment actually needs them.

In MRO, the right method depends less on whether a part is a bearing, motor, or relay, and more on how its demand behaves. Is it driven by scheduled maintenance? Does it fail unpredictably? Is demand steady, intermittent or lumpy, or fading out with an aging asset base?

That matters a lot because different parts need different planning logic/models. The table below provides a high-level view of models grouped by the problem they solve.

For maintenance teams, the takeaway is simple:

• If the part moves because maintenance is scheduled, use simple models that follow that cadence.
• If the part moves irregularly with long gaps between demand, use intermittent-demand methods.
• If the part is fading out with the asset population, use methods that can adapt downward.
• If future demand can be explained by PM schedules, work orders, runtime, or asset condition, ML may help.
• If the part is rare but critical, focus on stockout risk and service continuity, not just forecast accuracy.

In MRO, the best forecasting method is the one that fits the part’s real demand pattern and helps reduce downtime risk.

Sources and influences

Criticality and risk-based spare parts policies

Van Houtum, G. J., & Kranenburg, A. A. (2009). Spare parts inventory control under system availability constraints. Springer.

Moubray, J. (1997). Reliability-centered maintenance (RCM II). Industrial Press.

Gulati, R. (2013). Maintenance and reliability best practices (2nd ed.). Industrial Press.

Maintenance-driven demand forecasting

Palmer, D. (2019). Maintenance planning and scheduling handbook (4th ed.). McGraw-Hill Education.

Intermittent demand forecasting methods

Syntetos, A. A., Boylan, J. E., & Croston, J. D. (2005). On the categorization of demand patterns. Journal of the Operational Research Society, 56(5), 495–503.

Pince, Ç., Dekker, R., & Zuidwijk, R. (2021). Intermittent spare parts demand forecasting: A review. European Journal of Operational Research, 291(2), 1–16. (Note: verify page range if needed)

Inventory policy and reorder points

Zipkin, P. H. (2000). Foundations of inventory management. McGraw-Hill.

Silver, E. A., Pyke, D. F., & Peterson, R. (1998). Inventory management and production planning and scheduling (3rd ed.). Wiley.

Storeroom practices and MRO planning

DeWald, D. M. (2011). Maintenance storerooms and MRO made simple: A practical guide to effective inventory management. Industrial Press.

Demand Forecasting

Altay, N., & Litteral, L. A. (2011). Service Parts Management: Demand Forecasting and Inventory Control. Springer Science & Business Media.

Phillip Slater, Spare Parts Inventory Management: A Complete Guide to Sparesology (Industrial Press, 2016).