It’s 2 AM. Your SAG mill gearbox just failed. The plant is down, losing $95,000 every hour. You call the warehouse: “Do we have a spare?”
That answer could mean the difference between a 4-hour repair and a 4-week nightmare waiting for an emergency shipment from overseas. The frustrating thing is, someone made a decision months or years ago about whether to stock that gearbox. They either got it right, or they didn’t. Tonight you find out which.
This scenario plays out daily across mining and processing operations worldwide, and most organizations don’t know they’ve got it wrong until that phone call comes. They’re either sitting on millions in dusty inventory that’ll never be used, or they’re one failure away from catastrophe. The sweet spot in the middle is surprisingly hard to find.
What makes it hard is uncertainty. We don’t know exactly when things will fail. If we did, inventory planning would be trivial—just have the part arrive the day before you need it. But failures are only vaguely predictable, lead times vary, and operating conditions change. Good reliability engineering isn’t about being perfect. It’s about getting things right enough that when the moment of truth comes, the odds are in your favor.
And that word—odds—matters more than people realize. Just because you’ve done a rigorous spares analysis doesn’t mean you’ll never experience a stockout. Probability doesn’t work that way. When a stockout does happen, the question isn’t “who screwed up?” but rather “does this still fall within the acceptable risk we signed up for?” A 95% service level means one year in twenty you might get caught out. That’s not failure—that’s the deal you made.
Part of the problem is that most organizations treat all spare parts the same way. A $15,000 mechanical seal gets managed with the same rules as a $280,000 gearbox or a $50 indicator light. But these parts behave completely differently. Consumables like filters have regular, predictable demand—you go through roughly the same number each month, and standard inventory methods work fine. Critical spares are a different animal. They sit on a shelf for years, then suddenly you need one urgently. The failure is hard to predict, and the consequences are severe. These demand a more mathematical, risk-based approach.
The good news is that randomness, properly understood, becomes surprisingly predictable. When you replace a failed component, the new one starts fresh—no memory of what came before, just its own independent lifespan ahead. Across a population of equipment, individual failures stay random, but the average failure rate stabilizes over time. Think of flipping a coin. You can’t predict any single flip, but over a thousand flips you’ll land close to 500 heads. Same principle. This is what allows us to answer practical questions like: how many spares do we need for a 95% chance of not running out?
Once you can estimate failure rates, different parts need different thinking. Cheap consumables are easy—stock enough to cover demand plus a safety margin, reorder when you’re running low. The interesting problem is at the other end of the spectrum: expensive, slow-moving spares that you might never need, but would be catastrophic to be without. Things like that SAG mill gearbox. The question isn’t how many to stock. It’s whether to stock one at all.
This one’s simpler than it looks. Compare two scenarios. Stock the spare and you pay holding costs every year—cost of capital, storage, insurance, obsolescence risk—typically 20-35% of the part’s value. But when failure hits, you’ve got the part ready. Don’t stock it, and you save those holding costs, but when failure hits you’re waiting weeks for delivery while the plant bleeds money.
The purchase price itself actually cancels out, which isn’t obvious at first. Whether you stock the spare or not, you’re going to buy it eventually when failure occurs. The only question is timing. So the real comparison is holding cost versus expected downtime.
Run the numbers on that gearbox. It costs $280,000, so holding runs about $70,000 a year. Based on operating hours and reliability data, say there’s a 13% chance of failure in any given year. Without a spare on hand, you’re looking at 4 weeks of lead time. At $95,000 per hour, that’s around $64 million in losses.
Expected annual cost of not stocking: 13% times $64 million. Roughly $8 million. Compare that to $70,000 in holding costs—a 117:1 ratio. Stock the gearbox. Sleep better.
Now flip it around. A $1,200 non-critical indicator that fails once every 5 years and causes minor inconvenience? The math says don’t bother. Order one when you need it.
Spare parts optimization lives at the intersection of reliability engineering, statistics, and economics. It’s not about eliminating risk—that would be impossibly expensive. It’s about making informed calls on which risks to carry and which to cover. Get it right and you free up capital, cut downtime, and stop firefighting.
And at 2 AM when that gearbox fails, you’ll have the spare ready (or not, it is a probability after all). That’s not just good engineering—it’s good business.
Leave a Reply