The stores-to-line handover: issue, backflush and the reconciliation that drifts

The stretch of floor nobody owns

A warehouse system tracks material to a location and hands it off. A production system tracks material against an order and picks it up. Where the handover happens is a decision, not a given. Is line-side stock still warehouse inventory, is it work in progress, or does it sit in a transit location both systems can see? Leaving it undecided is common, and the result is that counts never match because counters and record disagree about what should be there. Choose the model deliberately, name the locations, and make the transfer a recorded transaction rather than a forklift movement nobody enters. Warehouse-side execution itself belongs to the logistics stack; this boundary does not.

Three issue models, three different data consequences

Picking to a specific works order gives the tightest record and the heaviest transaction load; it suits low-volume, high-value assembly and irritates everyone in fast repetitive production. Replenishing a line-side location by pull signal moves material without reference to an order, so consumption must be inferred afterwards. Bulk stocking of low-value items is usually settled by periodic count rather than by event. Most plants run all three at once for different material classes, which is fine, provided the system is configured to expect it and each class has a written reconciliation method.

Backflush is convenient accounting with deferred truth

Backflushing deducts components according to the bill when an order is completed. It removes a great deal of transaction work, and it assumes the bill is accurate, nothing was substituted, scrap was reported, and nobody took a box from the line-side rack for another job. Every one of those assumptions leaks somewhere. The drift stays invisible until a count, by which point its causes are history. Backflush is defensible for low-value, high-usage components with a disciplined count cycle. It is a poor choice for expensive, traceable or frequently substituted material.

Receiving finished goods without dropping the record

Completion should create a receipt that carries identity forward: lot or serial, the order that produced it, and the quality status it holds. Two failures recur. The receipt lands in available stock before quality has released it, so a planner promises material that cannot legally ship. Or identity is dropped at the boundary and the warehouse holds a quantity of a part with no link to the run that made it, destroying traceability at the last possible moment. Both are configuration choices rather than software limitations, and both are cheap to fix before go-live and expensive after.

Count the line, not only the racks

Cycle counting programmes usually cover the warehouse and stop at the shop floor door, which is where the accuracy problem actually lives. Line-side bins, work in progress at each operation, material away at treatment or cleaning, and returns awaiting put-away all hold stock the record claims to know. Counting them exposes the real causes: unreported scrap, substitutions never entered, issues booked to the wrong order, and negatives created by backflushing more than was ever issued. Correct the transaction that caused the error rather than adjusting the balance, or the same variance reappears next cycle.

Frequently asked questions

Should stores or production own line-side stock?

Either works; what matters is that only one does and that everybody knows which. If stores owns it, replenishment and counting stay in one place but production cannot see its own buffer without an interface. If production owns it, consumption reporting is simpler and stores lose sight of material they are still expected to replenish. Decide by who will physically count it, because the counting team and the owning system should belong to the same organisation.

Why does our shop floor on-hand keep going negative?

Negative balances almost always mean consumption is being recorded against material never formally issued, or issued to a different location. Backflushing at completion is a frequent cause, since the system deducts what the bill specifies regardless of what physically moved. Substitutions, unrecorded returns and picking from the wrong bin supply the rest. Chasing balances with adjustments conceals the pattern; tracing a handful of negatives back to the transaction that created them does not.

Do we need a real-time link between stores and production?

Not everywhere. Material that gates a machine start, carries traceability or costs enough to matter deserves an immediate transaction. Consumables and low-value fasteners can be reconciled on a cycle without harm. Deciding by material class rather than building one uniform interface keeps the transaction load survivable and concentrates engineering effort where an out-of-date figure would genuinely cause a stoppage or a compliance problem.