A balance that is already right when you go to count it
Nobody keys inventory into Backbone MRP. Stock moves because a receiver was locked, a shipment was locked, or an operator saved a production ticket — the same acts that were going to happen anyway. A physical count is then what it should be: finding the difference between what the floor says and what the documents say, and posting the difference.
-
One write path, and no way around it
Every stock change in the product goes through a single routine that adjusts the balance and writes the ledger row in one transaction. There is no second door through which a balance can move without a ledger entry explaining it — no import, no admin screen, no adjustment field on a part.
That constraint is the whole reason the ledger is worth reading. A history with one unexplained movement in it is a history nobody trusts, and once it is not trusted the floor starts keeping its real numbers somewhere else.
-
A balance per location, with a ledger behind it
Inventory is not a number on a part. It is one stock row per part, per company location, and a part carried at three plants is three rows, each independently correct. Locations nest under a parent, so a company running several factories describes that structure once and stock, receiving, shipping and production lines all hang off it.
Each ledger row records the quantity, its units, when it happened, who did it, a note, and — the important part — a pointer to the document that caused it: a shipment line, a receiver line, a run result, a material-used line, or a count tag.
-
Ask for the balance as of a date, and it works backward
Balances as of a past date are derived by working backward from today through the ledger, rather than forward from an opening figure. That is a deliberate choice for the situation every new customer is actually in: a company arriving with stock already on the floor and no history behind it has no trustworthy opening balance to work forward from.
Counts do not overwrite history either. A count adjustment is itself a ledger row citing its own tag and noted with the count number, so the correction and the reason for it are both permanent.
-
Two people can count the same part without double-counting it
Several people can count the same part in the same location without stepping on each other. A naive system compares each slip against the full on-hand figure and posts the variance twice; Backbone MRP recomputes the baseline as tags are edited and distributes it across the sibling tags covering that part and location, weighted by how much each one counted.
And because two people can hit Sync at the same moment, the sync takes an exclusive claim on the count before it validates anything — two simultaneous syncs cannot double-post. Variance is shown against what the documents say and valued at a unit price, so you see the money and not just the pieces.
Inventory accuracy is not really an inventory problem; it is a data-entry problem wearing a different hat. Balances go wrong because somebody had to remember a second step after doing the first one, and count day goes wrong because the count is trying to correct a figure nobody can explain. Deriving stock from documents that were going to be created anyway, forcing every movement through one auditable path, and treating a count as reconciliation rather than reconstruction is what makes the number right on an ordinary Tuesday — not just after a count.
Worth knowing: stock is tracked per part, per company location. There is no bin or shelf-level stock, and no lot, batch, serial or expiry tracking — a count tag has a free-text spot description for where a thing was found, but the balance itself lives at the location. The Inventory Counts subscription governs how many counts you may create, not whether your inventory is tracked.