Stocktakes without stopping the warehouse

The annual stocktake has a shape everyone recognises. A Saturday in January. Two people per aisle, goods-in stopped, goods-out stopped, a customer told their pallet cannot move until Monday. By Sunday evening you have a number and everybody is relieved. By February the number is wrong again, and nobody can say when it stopped being right.

That last part matters more than the lost Saturday. A count that produces a figure without producing an explanation tells you what you have and nothing about how you got there — which means there is nothing in it you can act on.

The freeze is the smallest cost

Stopping the warehouse for a day is expensive and visible, which is why it gets all the attention. The costs that are harder to see are worse.

• Time pressure produces its own errors. A team working to finish by Sunday recounts less, estimates part-cases, and nudges the second count towards the first. Some of the variance you find on Monday was created on Saturday.
• The counters are strangers to the goods. Office staff drafted in for the day cannot tell two similar articles apart, and have no reason to be suspicious of a shelf that looks tidy.
• One number a year cannot be attributed. A discrepancy found in January covers twelve months, three seasons and probably some staff turnover. It cannot be tied to a process, so it gets called shrinkage — a word that ends an investigation rather than starting one.

In fairness, the wall-to-wall count does one thing well: a single consistent snapshot taken at one moment, which is what a valuation at a balance date wants. Its weakness is as an operational instrument.

A count is only meaningful against a ledger

Suppose you count 96 where the system says 100. What have you learned? That you are four short. Not when they went, not from where, not through which process, and not whether they ever arrived. You post an adjustment of four and carry on. That adjustment is not a correction — it is a new opening balance, adopted because you have nothing better.

Now suppose every movement is recorded: goods received against a purchase order, reserved, picked, transferred, consumed on a work order, released back to a customer — each with a timestamp, a user and a document behind it. The same count now says something different: the ledger and the shelf disagree by four, since the last time this location was counted, across a known and listable set of movements in that window.

That is the whole argument, and it is worth stating plainly: a stocktake measures the gap between the ledger and the shelf. Without a ledger there is no gap to measure — only a fresh number replacing a stale one once a year, with the interval between them unexamined.

It is also the test to apply to the system you already have. If a correction can be typed straight over the on-hand figure, your movement history has a hole in it exactly where corrections happen. An adjustment should be a movement like any other: reason, quantity, who, when.

Counting a slice at a time

Cycle counting inverts the annual count. Instead of everything once, you count a small part of the warehouse continuously — a shift’s worth of locations each week, all year — so every article comes round on a cadence. A few things separate a routine that works from one that quietly stops:

• Frequency follows importance, not the alphabet. Fast-moving lines, high value, anything handled in loose pieces, anything with a history of variance — count often. Slow, bulky, whole-pallet stock — rarely. A flat cadence spends the same effort on both, and is the usual reason the routine comes to feel pointless.
• Count locations, not articles. A counter walks to an address and counts what is there. Counting by article sends the same person to four places and misses stock sitting somewhere it should not be — itself a common cause of variance.
• Count blind. If the expected figure is on the sheet, you will collect confirmations rather than counts. Show it after entry, not before.
• Count when the location is quiet. Picks queued against it, or a reservation being served, give you a variance that is really a timing artefact.
• Record who counted, and when. The window between two counts is what turns a difference into a short list of movements to look at.

The honest trade-off: cycle counting replaces one large annual effort with a permanent small one — easier on the operation, harder on the discipline. A heroic Saturday happens because it is in the calendar; two hours every Tuesday happens only if somebody owns it and someone notices when it is skipped. A programme that decays after four months leaves you with neither an accurate ledger nor a habit of counting.

Variances are signals

Once counts are dated and movements recorded, the pattern of variance starts carrying information.

• Small differences in both directions on one article usually mean a unit problem — someone counting cases where the system holds pieces. It nets out over a year, which is exactly why an annual count never finds it.
• One-directional drift on a single article points at a process: picking from the neighbouring location, an unrecorded consumption, goods issued against the wrong line.
• Variance concentrated at one address is usually labelling rather than theft — two locations that look alike, or stock that belongs one bay over.
• Variance that appears only after a particular process — returns, part-consumption on a work order, a transfer between warehouses — tells you which procedure to go and watch. You never get that from an annual count, because by January the process has run a thousand times.

The realistic goal is not zero variance. It is variance that is small, stable and explainable, plus a threshold above which somebody is actually required to look; a tolerance nobody enforces is the same as no tolerance. And if you hold goods belonging to your customers, a discrepancy is not an adjustment you post quietly to your own books — it is a conversation with the owner, and the paperwork you issued when the goods came in is what that conversation runs on. We wrote about that separately in storing customer goods.

What cycle counting does not remove

It does not remove the year-end obligation, and it is worth being precise rather than optimistic here.

Under the Swedish Lag (1955:257) om inventering av varulager för inkomstbeskattningen, a business required to keep books must inventory each item of stock held for sale or consumption, draw up a list stating the value each item has been recorded at, and sign that list with a declaration on honour and conscience that nothing has been left out. The consequence of not following it is specific: the reported stock value is not accepted for income taxation.

Whether a well-documented continuous routine can support that balance-date list — what coverage it needs, what evidence, what sign-off — is a question for your accountant and auditor, and one to ask before you redesign the routine rather than after. What cycle counting reliably changes is the character of the year-end: from a discovery, where the size of the adjustment is a genuine surprise, to a confirmation of a figure you have maintained all year.

Get the ledger right before any of it. Counting against a figure that is only sometimes updated produces variances that measure your recording habits rather than your warehouse, and the team learns to distrust the exercise. Then give every location an address, pick a cadence you can sustain through a peak week, count blind, and put every adjustment through the ledger as a movement with a reason on it.

In navichain, stock is a ledger rather than a number: on-hand per article and per warehouse with every movement on an audit trail, reserved goods held so the same pallet is never promised twice, and floor plans, zones and locations so a pallet has an address and not a guess. You count in what you actually count in — pallets, kilos, litres or pieces — warehouse goods and workshop parts share one set of shelf figures, and customer goods held in custody get a sealed, barcoded goods receipt note. It is all on every plan, from 995 kr a month; the platform page shows what is inside, and transport and warehouse in one system covers why the two belong together.