Stock Reconciliation: Four Files, Three Variance Families

A stock figure is not one number; it is the point where four separate files agree. Month-end disputes are rarely a missing carton; they are two teams reading different files and calling both inventory.

Key facts

• Four files, four owners. The warehouse record, order export, despatch manifest and adjustment log come from different parties on different clocks.
• Movement is the tie-out key. A quantity can be matched only after every event that changed it is placed on one timeline.
• Accuracy is measured per order line. The denominator is the order line: mis-picks plus misses run under 0.01 percent of it.
• A variance is an ownership question before it is a number. Which file the loss sits in decides who reprices and who absorbs it.

Four files, and the clock each one runs on

Table: what each file legitimately proves, and the claim it cannot support.

The storefront export is a sales file, not a stock file: it records intent, and counts a cancelled order and an unshipped order in the same column.

Why the tie-out starts with movement, not quantity

Opening with the quantity gap tells you how much you lost; it does not tell you where. Start with the movement ledger — every event that changed a line between the two close dates — and set it beside the order export on one timeline. Four event families belong in it: receipts, despatches, returns to stock and adjustments. A gap with no unexplained event is a definition problem. An unexplained event is a finding to work.

Two rules decide whether that ledger is trustworthy. Stock instructable ahead of the 16:00 China-time cut-off leaves the same working day, so a later handover is a real event, not a clerical one. And once a pick list is built, a change becomes its own event with its own re-pick rather than slipping into a queue already started — which is why the change count, not the order count, explains most same-day gaps.

Three variance families, and where each belongs

Count against count. Physical quantity differs from the recorded quantity on a line with no open movement. That belongs to the site holding the goods, and is resolved by a recount then a re-slot check: a location that took on a second purpose hides a line in plain sight.

Timing. Both sides agree on the figure and disagree on the date: despatch recorded on one side, revenue on the other. That belongs to nobody; it ends when a convention — usually the manifest date — is agreed before the close.

Identity and scope. Two records for one product, or one record for a bundle: a storefront variant, a renamed SKU and a multi-pack are three stock facts. That belongs to whoever keeps the mapping table, almost always the seller. See Damaged, Missing or Written Off for how a loss is allocated once identity is settled.

What a cycle count must do to save the year-end count

A full physical count is an event, not a control. A control runs often enough to catch drift while a finding is still small. Two properties make a rolling count worth its cost.

It counts lines that moved, not lines that are convenient: an untouched line counted twice proves nothing. And it counts at the location, not from a report — a quantity read off a screen closes paperwork, never a mis-slot, which is the failure mode a count exists to catch. The inbound version is in The Pallet Arrives Short.

What the gap actually costs

A variance is rarely worth the value of the missing units. Its cost is that your replenishment decision was made on a number that no longer describes the rack. Sell the phantom and you have a cancellation, a refund and a marketplace metric to answer for; treat it as real and you hold capital in a line that cannot ship. Reorder arithmetic keys off the same figure, which is why a stale count compounds: Reorder Point on a Long Lane.

Two boundaries belong in that conversation. Per-line accuracy is the metric we publish, and one failing line is not one failing parcel: One Line, Not One Order. And a stock loss is settled against the declared value recorded for that consignment, not a retail price you might have realised.

The close pack, and what to demand of any provider

A close that takes a week is a close nobody trusts. At the first of the month four artefacts should arrive, each keyed to the same line identifiers: closing quantity by line and site; the movement ledger with dates and event types; the open variance list, each item tagged to an owner; and the period's adjustment entries, with the finding behind each. Anything that cannot be tied to a dated event is an opinion.

Before accepting a close, test whether a provider can produce those artefacts at all — they are what separates an operation from a spreadsheet: What a Fulfilment Partner Should Produce on Demand. Which clauses make them enforceable is a contract question, not a reporting one.

Send your order export and your last close and we will tie them to our movement ledger and name the family each gap belongs to: tony@sbt56.com, or the form at Contact. Standing answers are on the FAQ page.

Which file is the inventory of record when we disagree?

The one sitting closest to the physical goods, provided it is open to inspection. A warehouse record can be wrong, but it is wrong about what both sides are arguing over: the rack. A storefront export is never the inventory of record because it describes intent, not possession — an order can be paid, cancelled and refunded in that file while the carton never moved. Name the stock record as the agreement's baseline and reserve the right to challenge it with a dated count.

Can you count my stock while it is in your warehouse?

A count is available on instruction, and it is done at the location rather than from a report, because a screen read only closes paperwork. What we cannot offer is a count that replaces your own controls: identity questions — duplicate SKUs, renamed variants, bundles — are decided from your catalogue, not from a rack. Where the finding is a damaged or missing unit, ask for both outputs: the dated count and the adjustment entry it produced.

How long should a month-end close take?

Short enough that nobody works around it. A close that runs for weeks gets skipped in practice, and its findings keep compounding into the next replenishment. The real dependency is event discipline, not processing speed: when despatches, returns to slot and adjustments are dated events while the month is open, the close reconciles two ledgers instead of excavating one.

Operational guidance published 1 November 2026 (2026-11-01) by FulfillDock by SBT, with 24,000 m² at three sites — Shenzhen, Dongguan, Suzhou — under the company registration Shenzhen Shibotong International Logistics Co., Ltd. Reconciliation method and ownership logic only: no prices, no amounts and no rate card appear here. Accuracy and cut-off figures are published operating parameters, measured per order line.