A spare part that isn't there when you need it costs downtime. A warehouse full of parts you never use costs money and space. Between those extremes lies a calculation you can actually make: when do you reorder, and how much buffer do you hold? This article covers the formulas, a worked example and the places where it goes wrong in practice. For the broader picture, see spare parts management.
Why not go on instinct
In many technical warehouses, experience decides when to reorder. That works as long as the same person is around and consumption is stable. It breaks the moment that person is on holiday, consumption shifts or a supplier stretches their lead time.
A calculated reorder point makes the decision transferable. It records your assumptions — this is what we consume, this is how long delivery takes, this is the risk we accept — so you can adjust them when reality changes. That is not the same as automating the decision; it is making it explicit.
The basic formula
The reorder point is the stock level at which you place a new order. It has to be large enough to bridge the lead time:
Reorder point = (average daily consumption × lead time in days) + safety stock
The first term is your expected consumption during the lead time. The second is the buffer for everything that disappoints: a spike in consumption, a supplier running late, or both at once.
Worked example
A bearing is consumed on average 24 times a year, roughly 0.1 units per working day across 240 working days. The supplier delivers in 10 working days.
Expected consumption during lead time: 0.1 × 10 = 1 unit. Without a buffer you would reorder at a stock level of 1. That fails half the time, because half the time you consume more than average.
Choose a safety stock of 2 units and the reorder point becomes 3. Order a quantity that tops up to 8 and you have a min/max of 3/8.
Safety stock: how much buffer is enough
Safety stock is the dial with which you buy or sell risk. More buffer means less chance of a stockout and more capital tied up.
The rule of thumb
For most parts a practical approach is enough: take consumption over half to one full lead time as your buffer. With long or unreliable lead times you move towards a full lead time; with fast, dependable suppliers towards half.
Calculating it properly
If you want to be sharper, work with variability. Safety stock then becomes the product of a service factor and the standard deviation of consumption during the lead time. A service factor of 1.65 corresponds to roughly 95 percent service level, 2.33 to roughly 99 percent.
The question is not which formula is most elegant, but whether you have the data to feed it. Without reliable consumption history, a statistical formula produces false precision. Start with the rule of thumb, gather a year of history and refine after that.
Reorder point or min/max
A reorder point says when you order. A min/max also says how much: top up to the maximum. In a technical warehouse min/max usually works better, because you don't have to decide on a quantity with every order.
You set the maximum based on ordering costs and shelf life. Parts you can cheaply reorder one at a time don't need a large maximum. Parts with high shipping costs or a minimum order quantity get a wider maximum, as long as shelf life allows it.
Lead time is not a constant
The biggest error in reorder point calculations is not in the formula but in the input: the lead time the supplier promises rather than the one they deliver.
So measure your actual lead times. If the stated lead time is 10 days but realised lead time varies between 8 and 25 days, your problem is not the average but the spread. In that case calculate with a lead time you meet in most cases, not with the mean.
This is also worth a conversation with purchasing. A supplier who knows you measure their performance often behaves differently.
What goes wrong in practice
Dividing annual consumption by 365. Your warehouse isn't open 365 days and machines don't run every day. Calculate with working days, or with run hours if consumption is tied to them.
One reorder point for the whole assortment. A three-euro seal and a four-thousand-euro pump rotor do not deserve the same consideration. Differentiate, for example through an ABC analysis.
Set and forget. Consumption changes as machines age or production shifts. Anyone who never revisits their min/max slowly builds a warehouse full of the wrong stock.
Calculating on physical stock. What counts is available stock: physical minus what is already reserved for open work orders. Without reservations a part can be promised twice.
Getting started
Don't start with the whole assortment. Take the parts you ran out of over the past year, plus the parts whose failure causes the most downtime. That is usually a few dozen. Determine consumption for those, measure actual lead time, set min/max and evaluate after six months.
Once that works, expand. A system that calculates replenishment needs automatically then takes over the arithmetic — see how that works on the page about warehouse and spare parts.
Frequently asked questions
What is the difference between reorder point and minimum stock?
In practice the terms are used interchangeably. Strictly, the reorder point is the level at which you order, and minimum stock is the floor you don't want to go below — often equal to safety stock. What matters is that your organisation uses one definition. The common terms are collected in the glossary.
Do I need a reorder point for every item?
No. For parts you rarely use and that are quickly available, ordering when you need them is cheaper than holding stock. A reorder point makes sense with regular consumption or long lead times.
How do I handle parts I use once every five years?
No consumption formula helps there. That is a criticality decision: what does downtime cost if the part isn't there, and how long does delivery take? For a critical machine with a lead time of months you keep one on the shelf, even if you rarely use it.
Can I use run hours instead of days?
Yes, and it is often more accurate. If a filter is replaced after 2,000 run hours and the machine runs 60 hours a week, consumption becomes predictable from run-hour registration rather than from a calendar.