More Inventory, More Escalations: Fixing the Real Problem in Your Service Parts Network | June 2026

Inventory up. Escalations up. Both.

Spare parts inventory is at an all-time high. So are customer escalations. Both keep rising at the same time, and no one in the room can explain why.

The numbers are not small. Manufacturers tie up around 10% of annual sales in spare parts. Customer satisfaction with after-sales support runs 10 to 15% below expectation. In many networks, half of all service calls are delayed because the right part is not where it needs to be [1].

So executives reach for the usual three tools, and each one makes the picture worse.

• Cut the inventory. Demand for repairs does not arrive on a schedule. Cutting the buffer to please the CFO simply moves the failure into the customer’s downtime and into next quarter’s escalation report. Escalations get worse.
• Add more inventory. It feels safer. It is not. Roughly a quarter of the parts already on the shelf go obsolete every year. Adding volume to a poorly shaped network buys depreciation, not availability. CFO gets worse.
• Tighten the SLA. Promising a faster response without redesigning what sits behind it converts a contract clause into a liability. This means more credits paid, more renewals exposed to a sharper competitor.

The reason all three fail is the same. The network is being managed against a single aggregate number that hides the only thing the customer actually feels. The average is defended by overstocking cheap, fast-moving parts, while the expensive, low-volume, mission-critical parts that decide whether a customer is up or down quietly run thin. On paper the network is optimized. In the customer’s facility it is failing. That is how inventory and escalations rise together [2].

This is not a budget problem. It is a design problem. You are spending against the wrong shape of network. The companies that win the aftermarket treat their service promises as products that they design, price, produce, and deliver [1].

The network is a chain, not a hierarchy

In earlier articles we named availability as a board-level KPI. Later, we named the service geography problem. This article goes one layer deeper, into the network shape behind both.

Most service organizations describe their network in tiers: Tier 1 central, Tier 2 regional, Tier 3 local. The vocabulary is borrowed from commercial distribution and it misleads. Tiers describe a hierarchy of warehouses, each with its own budget and its own performance number. A service network is not a hierarchy. It is a chain, and the customer outcome is decided at the handoffs between links, not inside the warehouses [3].

Five points cross that chain. Each one exists for a different reason.

• The global pool absorbs the risk of expensive, rarely needed parts that no single region could justify holding alone.
• The regional hub balances cost against reach across a territory.
• The forward stocking location (FSL) sits near customer concentrations to buy time against the SLA clock.
• The technician’s van kills last-mile delay by putting the part in the hand that will install it.
• The customer-floor stock (consignment) removes the clock entirely for the assets that cannot wait.

These are not five sizes of the same warehouse. They are five different jobs, and the network only performs when each is doing the one it was designed for [4]. The chain breaks at the handoffs, not inside the warehouses. An empty central hub destroys the local site’s promise no matter how well that local site is managed.

Once you see the network as a chain, the next question comes naturally. Which parts belong on which link?

Segment before you stock

A ten-dollar fan can take down a quarter-million-dollar server rack in a data center in minutes. Most service networks stock the spare power supplies instead. That is the segmentation problem in one sentence.

Most planners answer the “which parts where” question by SKU, sorted by unit cost. The right answer sorts by type of SKU, on two axes.

Criticality. Not unit cost. Not engineering complexity. The only question that matters is what happens to the customer’s asset when the part is missing [4].

Demand pattern. Some parts move on a schedule and behave like ordinary inventory. Others arrive only when something breaks and exist as insurance against failures no one can predict. The two cannot be planned with the same logic [5].

Cross those two axes and the network sorts itself.

The piece most segmentation models leave out sits inside each cell: who decides. Without a named owner per quadrant, the matrix collapses back into a single planner’s spreadsheet, and within a year the network drifts back to the initial mess.

A board that gets this right protects what is now the most profitable revenue stream in the business. BCG estimates that aftermarket service represents a third or more of total income for industry leaders [6]. The segmentation matrix is how that income gets defended.

From SLA tier to stocking policy

The contract says 4-hours-delivery. The customer’s experience says it depends on the postcode. But the contract is not the operating model. The clause is one sentence. The stocking policy is what turns that sentence into a working rule the network can execute the same way everywhere.

Most B2B service contracts collapse into three real response tiers, regardless of how each customer’s paperwork labels them.

• 4-hour response. Mission-critical, usually on-site or near-site.
• Next business day. The industrial default.
• Next week. Low-criticality, planned maintenance.

Each one earns a different rule on five levers.

The most expensive mistake is the default. One blanket stocking target across all SKUs and all customers. It feels easy. You over-stock the parts that did not need it and you under-stock the parts that did. McKinsey’s 2024 survey of 250 senior aftermarket procurement executives makes the consequence clear: 70% reported no improvement in service from their providers over a full decade, with poor spare parts availability cited as a top-three customer complaint [7].

Design for the worst day, not the average

A service network can hit every monthly target and lose the contract anyway. The customer does not remember the months it worked. They remember the morning it did not.

The dashboard reports the average. The renewal turns on the incidents. A network that performs to spec on average and fails the spec by a wide margin even once a quarter has a problem the executive team can usually feel before the data shows it. The average is comfort. The worst day is consequence [8].

The worst incidents almost always come from the same place. Slow-moving, high-criticality parts. Rarely needed, but when needed they stop the customer. Standard forecasting tools rate their demand signal as weak, so they are also the first parts cut when working capital pressure arrives. That is how a routine budget review becomes the cause of next quarter’s escalation.

The fix begins with measurement. Stop measuring only the average of the parts-wait clock. Add the worst 5% and the worst 1% of all service events to the service reports. Most service organizations cannot tell those two numbers today, which is why the worst incidents keep surprising them. If you do not measure the rare bad days, you cannot design for them, and you will pay for them.

The structural answer is not to bloat every regional hub. It is a small set of insurance parts pooled centrally, sized against the risk of an extreme outage, with a same-day expedite lane already in place [8]. The industry has started to ship this discipline as product [9].

Three networks that look fine on paper

The dashboards in each of the three companies below were green. The renewals were not.

Data center. Empty forward stocking. A service provider opened 32 forward stocking locations (FSLs) near customer concentrations. Each holds only a fraction of the catalogue. When a ticket arrives, the local site has the right part one time in three. The other two times, the part arrives by same-day expedite. Within a year the expedite bill is larger than the inventory savings. The reflex of adding stock at every site makes the working capital problem worse without solving the dispatch problem. The footprint was sized by geography, not by the segmentation matrix.

Medtech. Insurance parts at every regional hub. A medical equipment manufacturer holds a complete kit of expensive insurance parts at every regional hub, on the theory that any severe outage should be covered locally. Three years in, working capital is up by millions and the customer experience has not improved. The company could have held a single set centrally, with an expedite lane in place with a Next-Flight-Out agent ready, and covered the same risk for a fraction of the cost.

Hi-tech OEM. A hub that did not survive the lanes. A hi-tech OEM opens a single European hub in the Netherlands and quotes a same-day SLA across the region. The hub is well-stocked and well-run. The SLA holds for the Benelux and parts of northern France. It fails everywhere else. Italy, most of Germany, and the rest of Europe sit outside the reach of a same-day promise from Amsterdam once the morning cut-off has passed. The reflex (open a second hub) only moves the problem somewhere else.

Three stories, three verticals, three failure modes. Over-distributed. Over-replicated. Designed against a map instead of against the lanes. All three networks looked rational on paper. All three failed in practice. Operational excellence cannot rescue a poorly designed network [6].

Who is in the room when the network is designed

Most service organizations have three voices in the room when the parts network is being designed. Operations defends the SLA. Commercial defends the contract. Finance defends the working capital. None of them sees the network at the granular level. None of them sees which lanes consistently break, which parts consistently miss, which sites consistently expedite.

The missing voice is the partner that runs the operational layer. A fourth-party logistics provider manages the forward stocking footprint, executes the replenishment cadence, and handles the expedite lanes. That is necessary, not strategic.

The strategic value sits one layer up. A 4PL or neutral lead logistics provider is the only point in the network where the exceptions accumulate. Every late lane, every missed part, every avoidable expedite passes through it. Those signals do not naturally roll up to an internal dashboard. They live in the daily exchanges between the operations team and the field. The data discipline that makes those exchanges visible is what turns the segmentation matrix from a one-time exercise into a continuously re-tuned operating model. The matrix gets re-scored against last quarter’s exceptions. The policy table gets re-examined when SLA breaches start to cluster. The footprint gets re-validated when lane-level data shows a structural mismatch. This is a discipline, not a project.

Service organizations that treat their 4PL as a warehouse operator never run this loop. The ones that treat the 4PL as part of the design team always do [6].

Two moves for next quarter

We see it day to day – parts network is a design problem. Most attempts to fix one fail because they try to fix it all at once. The two moves below are deliberately under-ambitious.

Move one. Build the segmentation matrix for your top three SLAs and your top one hundred SKUs. Not the full network. Not every tier. Not the whole catalogue. The smallest version of the matrix that is still a real useful resource. The matrix is the by-product. The point is the argument it will force inside the room. Operations, commercial, parts engineering, and finance, in one room, deciding who owns each quadrant. Force the decision once. The network stops breaking.

Move two. Make the worst day visible in the monthly service review. Add two numbers to the standing service pack: the worst 5% and the worst 1% of the parts-wait clock. Most service organizations cannot produce these two numbers today, which is why their boards keep being surprised by renewals they thought were safe.

The redesign of a service parts network is not done in 90 days. The visibility that makes the redesign possible is. The companies that protect their service margin over the next decade will be the ones that started this quarter, not the ones that waited for the next budget cycle [6].

Use the Service Parts Network Canvas to turn insight into a working network design-

Download The Service Parts Network Canvas>>

Author: Eyal Yossef, VP Supply Chain Solutions at Unilog

References

[1] Cohen, M. A., Agrawal, N., & Agrawal, V. (2006). Winning in the aftermarket. Harvard Business Review, 84(5), 129-138.
[2] Topan, E., Bayındır, Z. P., & Tan, T. (2017). Heuristics for multi-item two-echelon spare parts inventory control subject to aggregate and individual service measures. European Journal of Operational Research, 256(1), 126-138.
[3] Sherbrooke, C. C. (1968). METRIC: A multi-echelon technique for recoverable item control. Operations Research, 16(1), 122-141.
[4] Cohen, M. A., Zheng, Y.-S., & Agrawal, V. (1997). Service parts logistics: A benchmark analysis. IIE Transactions, 29(8), 627-639.
[5] Caserta, M., & D’Angelo, L. (2025). Intermittent demand forecasting for spare parts with little historical information. Journal of the Operational Research Society, 76(2), 294-309.
[6] Boston Consulting Group. (2025). Aftermarket services drive growth and higher margins for industrial manufacturers. BCG.
[7] McKinsey & Company. (2024). How aftermarket service providers can meet new customer expectations. McKinsey & Company.
[8] Lamghari-Idrissi, D., Basten, R. J. I., & van Houtum, G. J. (2020). Spare parts inventory control under a fixed-term contract with a long-down constraint. International Journal of Production Economics, 219, 123-137. https://doi.org/10.1016/j.ijpe.2019.05.023
[9] Syncron. (2025). Intelligent inventory buffering. Syncron.