Spare parts: accounting for "downtime insurance" as a ledger

I. The economics of spare parts: money at both ends

The two most common failures in spare-parts management point in opposite directions. One is the "stockout culture": keep only cheap parts in stock, and when a critical part fails you wait for an import — a two-week shutdown whose loss far exceeds ten years of inventory capital. The other is the "hoarding culture": buy extra for fear of stockouts, and five years later three sets of obsolete models lie idle in the warehouse, tying up cash on the books while physically depreciating.

The common cause of both failures is managing spare parts with a single metric — "inventory value". The correct view is to write the total cost of spare parts as the sum of three terms:

This formula yields two counter-intuitive conclusions. Conclusion 1: the more expensive a part, the less it should be "over-stocked" — instead "lock in capacity" — dryer cylinders, shoe-press roll sleeves and large screen drums have very high holding costs; the right approach is a framework agreement with the supplier to lock the lead time, rather than buying one and storing it. Conclusion 2: the cheaper a part, the less it should be "finely managed" — building an economic order quantity model for a single bolt costs more in management than the bolt itself; the right approach is the two-bin system (one bin in use, one spare, refill when empty).

In terms of service level it is more intuitive: raising the "no-stockout probability" from 95% to 99% usually requires increasing safety stock by 50%–70% (because of the tail of the normal distribution), and whether that extra stockout loss is worthwhile must be calculated part by part — no one-size-fits-all.

There is another account often missed: the "time value" of spare parts. The same item, sitting in the supplier's warehouse is "available to buy anytime", while sitting in your warehouse is "available to use anytime"; the price difference between the two is the upper limit of premium you can bear. Conversely: if a supplier can deliver within 48 hours, then the inventory cost you pay for "48 hours earlier" should not exceed the expected loss of a two-day shutdown — in most cases this upper limit is surprisingly low, leading to the conclusion "don't stock it, sign a fast-delivery agreement".

II. ABC / XYZ: two dimensions defining a nine-box matrix

Spare-part classification cannot look at value alone. Value (ABC) answers "how much money", consumption pattern (XYZ) answers "is it predictable". Crossing the two dimensions, the management strategy of each cell in the nine-box matrix is completely different.

• Class A: the few parts accounting for 70%–80% of cumulative value (usually under 15% of the part types) — critical rolls, dryer attachments, main drive parts, large bearings. Management principle: a tailored policy per part, setting inventory and agreements individually.
• Class B: value 15%–20%, types 20%–30% — regular wear parts, pumps/valves, seals. Management principle: calculate safety stock by formula, review quarterly.
• Class C: small value share, many types — bolts, gaskets, clamps, indicator lights. Management principle: two-bin system or consignment, no management effort.
• Class X: stable, predictable consumption (e.g., fabrics/felts and filter elements replaced on a cycle) — suit stocking by consumption quota.
• Class Y: seasonal or trend fluctuation (e.g., concentrated consumption during shutdown season) — adjust by seasonal factor.
• Class Z: irregular, sudden consumption (e.g., accident-damaged parts) — can only be covered by safety stock or agreements.

III. Safety stock & reorder point: replacing "gut feel" with an algorithm

The essence of safety stock is "covering fluctuation with inventory". Fluctuation comes from two sources: uncertainty in consumption itself, and uncertainty in replenishment lead time. The superposition of the two uncertainties gives this classic formula:

Numerical example: a seal has mean monthly consumption 40 pcs, standard deviation 12 pcs, supplier lead time 2 months, required service level 95%. Then SS = 1.65 × 12 × √2 ≈ 28 pcs. The reorder point is the average consumption over the lead time plus safety stock:

Three usage points. Point 1: use actual delivery statistics for LT, not the contract lead time — actual delivery of imported parts is often 30%–50% longer than the contract, so safety stock calculated from the contract value is inevitably insufficient. Point 2: use a rolling window for σ (last 12 months), and exclude abnormal months of concentrated consumption during major overhauls, otherwise σ is inflated and inventory artificially high. Point 3: tier the Z value — use 99% for critical parts affecting downtime, 90%–95% for general parts, otherwise a blanket Z=2.33 pushes inventory to an unbearable level.

IV. Reverse-deriving the ledger from BOM and parts list

On a site building a new spare-parts system, what is missing most is not the formula but the "list". The good news: equipment delivery already comes with large amounts of BOM, parts lists, drawing catalogs, and electrical & hydraulic component tables — these documents are ready-made raw ledgers, just scattered across dozens of PDFs and Excel files. The task is to organize them into a searchable table.

Minimum field set for the ledger (recommended no more than 20 columns; too many fields and no one maintains it): part number (plant code), name, specification/model, applicable machine type & station, material, supplier & contact, OEM part number, substitute part number, unit, unit price, lead time, MOQ, current stock, safety stock, reorder point, ABC class, consumption record, bin location, remarks (drawing number).

V. Localization substitution: five-step verification — skip one step and it is a gamble

Imported parts are expensive and long-lead, so localization substitution is something almost every paper mill does. But the failure modes of substitution are very concentrated: dimensions match, it runs after installation, then problems appear three months later — the cause being invisible differences in material, heat treatment, dynamic-balancing grade, and seal structure. So substitution is not "buy one and try", but a verification process.

• Step 1 · Reverse engineering: complete mapping (including tolerances and geometric tolerances), material analysis (spectroscopy + hardness), inference of key processes (heat treatment, surface treatment, welding method). This step determines all subsequent judgments and cannot be skipped.
• Step 2 · Benchmarking: write the original part's key indicators into an acceptance table — material grade & mechanical properties, hardness gradient, dynamic-balancing grade, seal-face roughness, fatigue-life requirement. Substitution without an acceptance table can only be accepted by "looks about the same".
• Step 3 · Bench or no-load test: parts that can run on a test rig run first (pumps, valves, rolls), measuring flow-head, leakage, temperature rise, vibration; parts that cannot run get hydrostatic, flaw-detection, and dynamic-balancing individual tests.
• Step 4 · Field trial: first install on a non-critical station or during low-load periods, set a clear observation period and criteria (vibration, temperature rise, dryness, paper defects), and keep the original part as a fallback.
• Step 5 · Life tracking and scale-up: track to the first replacement cycle and compare with the original part's life; only scale up to critical stations once 80% or more of the life is reached, and at the same time write the part into the spare-parts ledger's "qualified substitute" field.

VI. Long-lead parts: cannot be solved by inventory, only by agreement

A paper machine has a batch of parts whose lead time is longer than repairing a whole machine: roll covering/grinding 8–16 weeks, shoe-press roll sleeve 12–20 weeks, dryer cylinder 4–9 months, large screen drum 8–12 weeks, imported vacuum pump 3–6 months. The inventory strategy for these parts has only one rule: rely on capacity locking, not inventory.

A framework agreement must spell out six things: price and adjustment mechanism, delivery-time commitment and penalty clauses, capacity reservation for emergency insert-orders, quality acceptance criteria and non-conformance handling, after-sales and technical-support response time, and a statement of interchangeability between spare and original. Miss any one, and when truly needed the agreement becomes "re-negotiate".

Consignment (VMI) is another practical tool: the supplier places goods in your warehouse and you settle only for what you use. It shifts inventory capital pressure to the supplier at the cost of a slightly higher unit price. Parts suited to consignment: high unit price but predictable consumption (seals, bearings, electrical spares), or where the supplier is willing to use your warehouse as a regional transit hub.

VII. Obsolescence, inventory count, and four metrics

The part of the spare-parts system that rots most easily is not stockout but obsolescence — parts unused for two years occupy bin locations and capital, and also mask the true inventory structure. Curing obsolescence relies on mechanism, not campaigns: stamp a "last issue date" at warehousing; beyond 12 months with no issue, automatically enter the pending-disposal list; review that list quarterly, with four outlets — reassign to another station, return to supplier, cannibalize for parts, or scrap.

Inventory-count discipline: Class A quarterly, Class B semi-annual, Class C annual; count discrepancies must trace process causes (issue not recorded, return not recorded, loan not logged), not stop at "adjusting the books". When book-vs-physical mismatch exceeds 3%, any inventory model is running idle — because the input data is wrong. Another often-ignored discipline: every issue must be tied to a work-order number. Issues without a work order can neither back-calculate single-machine spare consumption nor correlate consumption anomalies with equipment condition, and the ledger degenerates into a mere journal.

VIII. The linkage between spare parts and maintenance strategy

The spare-part grade is not determined by price, but by "failure consequence × predictability", which shares the same coordinate system as the maintenance strategy (BM/PM/CBM). Once the two are aligned, the four questions for a part answer themselves: whether to stock, how many, who replenishes, and what to do when it fails.

At its core this linkage is one sentence: spare parts are the physical implementation of the maintenance strategy. The strategy says "this part is by condition monitoring", so spares cannot just hoard physical stock; the strategy says "this part is by periodic replacement", so spares must be stocked to quota. When the two are disconnected, either money is wasted or downtime is waited out in vain; when aligned, one table answers "why stock three of this part and none of that one".