The arithmetic, and its one honest assumption
Multiply how much you use per day by how many days the supplier takes, then add a buffer. If you use eight boxes a week and the supplier takes five days, you need roughly six boxes on the shelf when you order, plus whatever cushion your tolerance for a stockout requires.
The assumption is that usage is roughly steady. For truck stock and common consumables that assumption holds well enough. For material driven entirely by which specific jobs sold this month, it does not — and that is what job material planning is for.
Lead time is the number that bites
Usage drifts slowly. Lead times jump. A supplier that has delivered in two days for a year discovers a backorder, and every reorder point calibrated on two days is instantly wrong at exactly the moment stock is running down.
Two defences: buffer the items whose absence stops production, and revisit lead-time assumptions whenever a supplier surprises you. The second costs nothing and is almost never done.
Which items deserve a reorder point
- Items used on most jobs, regardless of scope.
- Items with long or unreliable lead times.
- Cheap items whose absence stops an expensive crew.
- Not: job-specific material bought per project, which belongs on a material plan.
Why rule-based beats predictive here
Predictive replenishment assumes demand emerges from a smooth statistical process. Service-business material demand emerges from which estimates got signed, which is lumpy, seasonal, weather-dependent and heavily influenced by a handful of large jobs.
A reorder point built from lead time and steady usage is transparent — anyone can see why an item is flagged and can argue with the number. A prediction nobody can interrogate gets overridden until it is ignored. Transparency wins in operations far more often than sophistication does.
The habit that makes it work
Pick a day. Open the below-point list. Decide each line: order, defer, or fix the point because it was set wrong. Twenty minutes weekly, done by the same person, prevents more emergency supply runs than any amount of automation, because the failure mode was never detection — it was that nobody looked.