Inventory accuracy in multi-warehouse distribution

The number you cannot trust

Ask a single-warehouse retailer how much of an SKU they have and they walk to the shelf and count. Ask a multi-warehouse distributor and you get a pause, a spreadsheet, and a hedge: "Around 400 cartons, give or take, depending on what is on the vans."

That hedge is the problem. In distribution, stock is never in one place. It is split across a main warehouse, regional depots, van-sales inventory moving through the market, goods in transit on a delivery run, and a returns bin nobody has reconciled in a month. Each location is a chance for the number to drift from reality.

Where accuracy leaks

The usual culprits:

1. Van stock that is never reconciled. A rep loads 100 cartons in the morning. By evening some are sold, some returned, some "borrowed" by another rep. If the van is not closed out daily, the warehouse number is fiction.
2. Transfers that exist only in someone's head. Stock moved from the main warehouse to a depot on a verbal instruction, recorded days later or not at all.
3. Returns limbo. Goods come back from the market — damaged, expired, refused — and sit in a corner, neither in saleable stock nor written off.
4. Receiving shortcuts. A delivery arrives, the storekeeper signs, and the system gets updated tomorrow. Between now and tomorrow, the number is wrong.

The principle: every movement is a transaction

The fix is conceptually simple and operationally demanding: nothing moves without a recorded transaction. Receiving, transferring, loading a van, selling, returning, writing off — each is an event with a location, a quantity, and a timestamp. The current stock at any location is just the sum of its transactions. There is no "adjustment" to paper over gaps, because gaps are not allowed to open.

This is the difference between a system that *stores* a stock number and one that *derives* it. A stored number rots. A derived number is only ever as wrong as the last unrecorded movement — and if every movement is recorded, it is never wrong.

Bin-level matters more than you think

For a serious operation, warehouse-level accuracy is not enough. You need to know not just "400 cartons in the Lahore depot" but which bin, which batch, and which expiry. Batch and expiry tracking is not optional in pharma and food — it is the difference between a clean recall and a regulatory disaster. And bin-level location is what lets a picker fulfil an order in two minutes instead of twenty.

Cycle counts beat annual stock-takes

The annual stock-take — shut the warehouse, count everything, reconcile the chaos — is a ritual that tells you how wrong you were, once a year, too late to act. Replace it with rolling cycle counts: a small slice of SKUs counted every day, high-value and fast-moving items counted more often. Discrepancies surface in days, not months, and you can trace them back to a specific movement while the trail is still warm.

How we built for it at DistroOps

We treated stock as an event log, not a stored balance. Every receive, issue, transfer, van load, sale, and return is a transaction against a specific location and, where it matters, a specific bin and batch. Van sales close out daily — load, sell, return, reconcile — so the warehouse number is honest by the next morning. Cycle-count workflows let a storekeeper count a slice on their phone and flag variances on the spot.

The payoff is the answer to that opening question becoming boring. "How much do we have?" stops being a hedge and becomes a number you act on without a second thought.

The test

Here is the test for your own operation: pick one fast-moving SKU, right now, and ask three people — the warehouse manager, the van rep, and the system — how much you have. If the three answers agree, your accuracy is real. If they do not, you have found exactly where to start.