Spare-parts inventory planning: balancing stock-outs against overstock

The real cost of a missing part

When a robot goes down in a European factory, every hour of downtime is measured in lost output. But the cost of a missing spare part is not just the price of the part—it is the cost of the entire stoppage, plus the expedited shipping, plus the technician’s idle time, plus the customer’s lost trust. For a service network being set up to support Chinese robotics manufacturers in Europe, the spare-parts inventory is not a back-office concern; it is the frontline of customer satisfaction. Yet getting it wrong in the other direction—overstocking—ties up capital and warehouse space, and risks obsolescence as robot models evolve. This article lays out a practical framework for balancing these two failure modes, using failure-rate-driven stocking, ABC classification, and a clear-eyed view of what a stock-out really costs under the EU’s Right to Repair rules.

Why failure-rate-driven stocking beats guesswork

The most common mistake in spare-parts planning is to stock based on intuition or sales history. But the demand for spare parts is not random—it is driven by the failure rates of components in the installed base. A part that fails on average once every 10,000 operating hours will need a different stock level than one that fails every 1,000 hours. The first step is to collect field data: for each robot model, track the mean time between failures (MTBF) for each replaceable unit. This data may come from the manufacturer’s own testing, from early field returns, or from the service network’s own repair logs. Over time, the network can build a failure-rate table for every part, broken down by robot model and operating environment.

Once you have failure rates, you can calculate the expected number of failures per period for the installed base. For example, if you have 100 robots of a model, each running 6,000 hours per year, and a part has an MTBF of 20,000 hours, the expected failures per year are 100 × 6,000 / 20,000 = 30. That gives you a baseline annual demand. But you also need to account for variability—failures do not occur at a steady rate. A simple Poisson distribution can model the probability of a given number of failures in a month, and you can set a service level (e.g., 95% or 99%) that determines how many parts to keep on hand to avoid a stock-out. This is the essence of failure-rate-driven stocking: it ties inventory to the physics of the equipment, not to guesswork.

ABC classification: focus on the parts that matter

Not all parts are created equal. A typical robot may have hundreds of spare parts, but a small fraction of them account for the majority of the inventory value and the majority of the downtime risk. ABC classification is a standard tool to prioritize. Class A parts are high-value, high-criticality items—such as servo drives, controllers, or gearboxes—that are expensive and whose failure stops the robot. Class B parts are medium-value, medium-criticality, like sensors or cables. Class C parts are low-value, low-criticality, like filters or fuses.

The classification should be based on two dimensions: the cost of the part and the impact of its failure. A part that costs €5,000 and stops the line is clearly A. A part that costs €50 and can be replaced in minutes without stopping production is C. The classification drives the inventory policy: A parts get higher service levels and safety stock, because the cost of a stock-out is high. C parts can be stocked in smaller quantities or even ordered on demand, because the risk is low. This is not just about money—it is about where to focus the network’s attention and warehouse space.

The cost of a stock-out under Right to Repair

The EU’s Right to Repair legislation, which is being implemented in various forms across member states, changes the calculus of stock-outs. Under these rules, manufacturers are required to make spare parts available for a certain period after a product is placed on the market—often 7 to 10 years for certain products. For robots, this means the service network must be able to supply parts for the entire lifecycle of the installed base, which can be a decade or more. A stock-out is not just a missed sale; it is a potential violation of the legal obligation to provide spare parts. That can lead to fines, legal action, and reputational damage.

But the cost of a stock-out is not only legal. Consider a scenario: a robot in a German automotive plant fails on a Tuesday. The part is not in stock. The network orders it from the manufacturer in China, which takes 5 days to ship. The customer’s line is down for 5 days. At a cost of €10,000 per hour of downtime, that is €1.2 million in lost production. The customer may demand compensation, or simply switch to a competitor’s service. The cost of the part itself is negligible compared to the cost of the downtime. This is why the cost of a stock-out must be calculated not as the price of the part, but as the cost of downtime plus the cost of expedited logistics plus the cost of customer churn.

Balancing stock-outs and overstock: a practical framework

So how do you balance the two? The key is to set target service levels for each part class, based on the cost of a stock-out. For A parts, you might target a 99% service level, meaning you are willing to accept a stock-out only once in 100 orders. For B parts, 95%. For C parts, 90% or even lower. These targets translate into safety stock levels, which you can calculate using the demand distribution and the lead time from the manufacturer. The lead time is critical: if the part is stocked locally, lead time is hours; if it must come from China, it is days. The longer the lead time, the more safety stock you need.

But overstocking is also a cost. Inventory ties up capital, requires warehouse space, and risks obsolescence. A part that sits on the shelf for years may become obsolete when the robot model is updated. The cost of overstock is not just the purchase price—it is the opportunity cost of that capital, plus the cost of disposal if the part becomes obsolete. The optimal inventory level is where the marginal cost of holding one more part equals the marginal cost of a stock-out. This is a classic newsvendor problem, and it can be solved with a simple formula: order up to the level where the probability of demand exceeding stock is equal to the ratio of the holding cost to the sum of the holding cost and the stock-out cost.

Comparison of inventory strategies

To make the trade-off concrete, the table below compares three common inventory strategies: minimal stock (just-in-time), balanced (service-level-driven), and overstock (safety-first). Each has its own cost profile and risk profile.

The balanced strategy is usually the right default. It uses failure-rate data and ABC classification to set service levels, and it calculates safety stock based on the actual cost of a stock-out. Overstocking is only justified for a few truly critical parts where the cost of downtime is astronomical and the lead time is long. Minimal stock is risky for anything but the most trivial parts.

Implementation steps for a new service network

For a service network being set up in Europe, the first step is to gather data. Start with the installed base: how many robots of each model are in the field, and what are their operating hours? Then collect failure data from the manufacturer and from early field returns. If the network is new, it may not have its own failure data yet; in that case, use the manufacturer’s MTBF estimates and adjust as real data comes in. Next, classify parts using ABC analysis. Then set target service levels for each class, based on the cost of downtime for the typical customer. Finally, calculate safety stock for each part, using the lead time from the manufacturer and the demand distribution.

One important consideration is the location of the inventory. The IndexBox source notes that having a local spare parts inventory can be a differentiator for service networks, as it reduces turnaround time. The IDC source similarly emphasizes the importance of parts hubs and turnaround time. For Europe, a central warehouse in, say, Germany or the Netherlands can serve the whole continent, but for high-criticality parts, it may be worth placing smaller stocks at regional hubs closer to major customers. The trade-off is between the cost of multiple warehouses and the benefit of faster response times.

Another consideration is the legal environment. Right to Repair rules vary by country, and the network must verify the specific requirements in each market. Some countries may require parts to be available for a certain number of years, while others may have different rules. The network should build a compliance calendar to track these obligations.

Conclusion

Balancing stock-outs against overstock is not a one-time exercise. It requires continuous monitoring of failure rates, demand patterns, and lead times. As the installed base grows and robot models evolve, the inventory must be adjusted. The goal is not to eliminate stock-outs entirely—that would be too expensive—but to reduce them to a level where the cost of prevention equals the cost of the stock-out. By using failure-rate-driven stocking, ABC classification, and a clear understanding of the cost of a stock-out under Right to Repair, a service network can build an inventory that is both cost-effective and responsive. The result is a network that can keep robots running, customers happy, and the business viable.

Sources

• IndexBox — machinery services — https://www.indexbox.io/ (accessed 2026-01-03)
• IDC — Robotics market — https://www.idc.com/ (accessed 2026-01-03)