A 40 Seal Stops the Line While 2.4 Million Sits on the Shelf

These two items are governed by the same principle and it is not the one the ERP is using: service level should be set by consequence of unavailability, not applied uniformly.

The $40 seal

• Holding cost is trivial: three on hand is $120 of stock, perhaps $25–30/year at a 20–25% carrying rate.
• Consequence of a stockout is $60,000 of contribution, observed three times last year, so roughly $180,000/year of exposure.
• The ERP's 95% service level is a category error here. It was set by part price and demand volume, not by criticality. A part that stops a line at $60,000 per event warrants a service level of 99%+ and a minimum of 2–3 with a reorder point of 2.
• Also address the root cause: three stoppages on the same $40 seal means nobody escalated a repeat stockout. Add it to a critical-parts list with an escalation rule.

The 26-week gearbox

• Annualised cost of holding one spare, on a unit of, say, $120,000 at 20–25%, is $24,000–30,000.
• Consequence of failure without a spare is up to 26 weeks of lost production plus emergency sourcing premium. Unless there is a bypass, a second train, or a repair route with a short turnaround, one spare is justified by a very wide margin; the question is not whether but how.
• Confirm the 26-week figure first. A repair/rebuild route is frequently 4–8 weeks, and a guaranteed-turnaround repair agreement or a shared pool with another site may deliver the same protection for a fraction of the capital.
• Options in preference order: guaranteed-turnaround repair agreement or loaner; shared spare across sites; carrying one spare; redesign to a standard frame size that is more available.

Document both decisions in a criticality-based spares policy so the next item does not get argued case by case.

Sources, in order of safety:

1. Excess and obsolete. 34% of 11,000 SKUs had no movement in 24 months-about $816,000 at face value, though not all of it is disposable. Run a disposition: return to vendor for restocking fees, transfer to sister sites, sell through surplus channels, scrap the rest. A realistic recovery is $250,000–$350,000, and it is one-time.
2. Over-max quantities on moving stock. Parts held at four or five times annual usage can be brought to a sane max with no service risk. Expect $60,000–$120,000.
3. Duplicate and superseded item numbers. With accuracy at 61%, the item master almost certainly contains the same part under two numbers. Consolidation releases stock and simplifies everything downstream.

What must not be cut:

• Insurance spares for single-point, long-lead failures (the gearbox class), and anything with a lead time longer than the mean time to recover.
• Parts with demonstrated consumption-those are working stock, not excess.
• Anything on the critical-parts list without a documented alternative source.

The sequencing trap. Do not write off inventory before record accuracy is established. The 61% figure means the ledger is wrong; disposing against a wrong ledger destroys the case and may dispose of parts you actually hold. Count first, clean the item master, then dispose.

Say this to the plant manager: the $400,000 is achievable this year, but it is a cash release, not an expense saving, and roughly a third of the value on the shelf is not real. Set the expectation accordingly.

Days 1–30: establish the truth

• Freeze min/max changes for everything except items experiencing a stockout, so the baseline is stable.
• Cycle count the top 200 SKUs by annual usage value (A items) and record the error by cause: wrong location, wrong quantity, unit-of-measure error, unrecorded issue, duplicate item number. Cause coding is the whole point-accuracy does not improve until you know why it fails.
• Assign bin locations to the 2,000 unlocated items and put the critical-parts list behind controlled access with a signed withdrawal log.
• Appoint one named owner for spare-parts decisions; two planners dialling suppliers informally is how the min/max system quietly died.

Days 31–60: set policy, then buy

• Criticality review with maintenance and reliability on the top 100 single-point items: consequence of failure, lead time, availability of alternatives, repair versus replace.
• Set service levels by criticality, not by the ERP default. Stocking decisions for insurance spares follow the analysis in the previous question.
• Clean the item master: duplicates, descriptions, units of measure, supersessions.
• Begin the obsolete disposition.

Days 61–90: make it self-sustaining

• Cycle-count programme: A items monthly, B quarterly, C annually, performed by someone who is not the person who issued the part.
• Reorder point automation with a manual override log so informal adjustments become visible.
• Supplier agreements for the long-lead class.

Measurement. Record accuracy is measured by physical count against the system record, sampled by value class: target 95%+ on A items, 90% on B. Also track critical-item stockouts (target zero), fill rate, inventory turns, and obsolete value as a percentage of total. Do not report a blended accuracy figure across all 11,000 SKUs-it hides exactly the errors that stop the line.