The item record: what a thing is
An item is a definition, not a pile. It carries a name a human recognises, a unit of measure, a category, an optional SKU for lookup, a preferred vendor, and a reorder point. It does not carry 'how many we have' as an authored value — that number is produced by the locations underneath it.
The most common structural mistake is creating a separate item for every variation: one for charcoal, one for weathered wood, one for slate. That produces four hundred items, none of which roll up, and a reorder report nobody reads.
Variants: same item, different flavour
Colour, size, length and gauge are variants. They share the item's identity and unit of measure but hold their own stock and often their own vendor part number. Keeping them as variants means the item list stays the length of your catalogue rather than the length of your catalogue times its options.
The practical test: if you would describe two things to a supplier as 'the same product, just the other colour', they are variants of one item.
Locations: everywhere stock can legitimately sit
- A warehouse or shop — the primary holding place.
- A vehicle — a van or truck carrying working stock between jobs.
- A job site — material staged on site ahead of production.
- A secondary yard or storage unit.
Why the truck has to be a location
Businesses that only model the warehouse produce a predictable pattern: the warehouse number looks low, someone orders more, and the missing material turns out to have been in a van for two weeks. Stock in a vehicle is still owned, still cost money, and still needs to be counted.
Modelling vehicles as locations also makes the restock conversation concrete. 'Van 3 is short on fittings' is actionable in a way 'we're low on fittings' never is.
How deep to go
Bins, shelves and aisles are useful in a large warehouse and pure overhead in a two-bay shop. Start at the level where a person actually walks to a different place to get the thing. If the whole shop is one walk, the whole shop is one location.
Depth can be added later. Splitting one location into three is a straightforward operational change; merging thirty abandoned locations back into one after a failed rollout is not.