The number nobody defends
Ask a finance director which figure on their balance sheet they would least like to be questioned on, and in a stock-heavy business the answer is stock. Not because they think it is wrong exactly, but because they know how it was arrived at: a count done by tired people on a Sunday, valued using a method that was chosen years ago, fed by documents from forty suppliers in forty formats.
What makes stock unusual is that it is the only major balance-sheet item where all three of those things can be wrong independently. Cash has one failure mode and a third-party record to check it against. Receivables can be aged and confirmed. Stock can be miscounted, mis-valued and mis-recorded, and the three errors do not offset in any predictable way.
This article is about the third one — the recording — because it is the largest share of most differences and the only share that can be fixed at a desk. It is also the share that gets the least attention, for a reason worth stating plainly.
Who gets blamed, and why that is convenient
When stock is short, the explanation reached for first is loss: breakage, wastage, theft. It is an intuitive story, it involves no arithmetic, and it has the considerable advantage of locating the problem somewhere other than the finance function.
It is also, in most reconciliations that are actually performed, wrong about the majority of the amount. The pattern is consistent enough to be worth stating as a rule of thumb: when a business reconciles its purchase side properly for the first time, most of the variance turns out to have a document attached to it.
That has a cost beyond the money. A warehouse team told annually that stock is short by an amount nobody can explain learns that the number is arbitrary, and stops treating receiving accuracy as something that matters. The documentary errors then get worse, because the one control that would catch them — counting at the door — is the one being quietly abandoned.
Two kinds of difference, and only one is interesting
It helps to split the problem before naming causes. A stock difference is either quantity — you hold a different number of things than you thought — or value — you hold the right number of things at the wrong cost.
The two feel similar and behave completely differently. A quantity difference means something physical happened, or a number was recorded wrong. A value difference means the physical world is fine and the arithmetic is not, which is why value differences survive counts: you can count perfectly and still be wrong about what the pile is worth.
Of the eight causes below, three are quantity and five are value. That ratio is itself informative, because the count only tests one of the two — which is a large part of why the annual count is not the reassurance people take it for.
1 · Unit of measure — the silent multiplier
The biggest cause, and the one that does the most damage per occurrence, because the error is multiplicative and completely silent. A product set up in your system as singles when the supplier invoices in cases of six is wrong by a factor of six on every purchase, forever, until somebody notices.
Nothing about the resulting figure looks unusual. The invoice is right, the payment is right, the stock system accepted the number without complaint. The only symptom is that stock of that product is persistently and inexplicably short — and because it is one product among hundreds, it hides comfortably inside a total.
It happens most on new products, because that is when the mapping is created, usually in a hurry, often by whoever was available. And it compounds with pack sizes buried in descriptions: Widget 6x500ml is readable by a person and is prose, not data, which means every system that reads it is interpreting rather than extracting.
Worth knowing the signature: a difference that is exactly six, twelve or twenty-four times what you expected is not a delivery problem. Our quantity extraction page covers why the stated unit should be its own field and why the conversion should never be guessed.
2 · Freight — not wrong, just inconsistent
Delivery and handling charges do not usually cause errors. They cause incomparability, which is worse in a specific way: an error can be corrected, while incomparable data quietly misinforms every decision made from it.
Three suppliers present carriage three ways. One builds it into the unit price. One shows it as a separate line. One invoices it a week later on its own document referencing nothing. Capitalise it where it is itemised and expense it where it is buried, and the same physical product carries different costs depending on who delivered it.
The consequence lands where it hurts most: product-level margin, and supplier comparison. A buyer asking which of two suppliers is cheaper is asking a question the data cannot answer, and will get an answer anyway. Decisions get made on it.
The fix is not choosing the right treatment — both are defensible — but choosing one and writing it down. It is one line in a note and it removes a recurring inconsistency permanently.
3 · Credit notes that were filed rather than applied
A credit note is a purchase in reverse and should reduce both quantity and value. It is also the document most likely to be missed, for a structural reason: an invoice generates pressure because somebody wants paying. A credit note generates none. Nobody chases you to accept money.
So it arrives, detached from the invoice it corrects, sometimes weeks later, gets filed as evidence that the matter was dealt with, and never reaches the stock system. The goods went back to the supplier in March and the ledger still holds them in December.
What makes this the best cause to look for is that the evidence exists. Unlike shrinkage, a missed credit has a document sitting in a folder proving it, which makes it the cheapest correction available and the most satisfying to find. Checking specifically for credits at period end is worth doing as its own step rather than trusting them to turn up.
4 · Cut-off — guaranteed to exist every period
Goods arrive on the 28th. The invoice arrives on the 4th. The stock is physically yours at the period end and no cost has been recorded anywhere. This is not an error, it is arithmetic — and it is structurally guaranteed to happen every single period.
Left alone it flatters margin: stock is right, cost of sales is understated, gross profit looks better than it was. Then the invoice lands in the following period and appears as a cost overrun nobody can explain. It is the single most common reason a monthly gross margin figure bounces around for no operational reason at all.
The mirror case is less common and more damaging: an invoice received for goods not yet arrived. That is a prepayment, not stock, and treating it as stock overstates what you hold in a way that survives a physical count — because the count finds the right number of things and nobody thinks to ask whether one of the entries should be there at all.
The defence is a deliberate look at the few days either side of the period end, rather than trusting whichever date the accounting system asked for. It is a short list and it accounts for a disproportionate share of period-end differences.
5 · Price drift nobody was told about
A supplier raises a price. They invoice correctly. You pay correctly. Nothing in the transaction is wrong, which is exactly why nothing catches it: the approval process checks that the invoice is arithmetically right and it is.
What is now wrong is everything downstream that assumed the old price — the standard cost, the selling price, the margin in the pricing spreadsheet somebody built two years ago. The stock value may be perfectly correct while the business is losing money on every unit sold.
This is invisible without a view of unit price by product over time, which is the one view that almost no bookkeeping produces, because bookkeeping records invoice totals. Sort a year of purchase lines by product and date and every change appears with the date it took effect and the document it appeared on.
The question that follows is the valuable one: did our selling price move too? A cost increase passed on is a non-event. A cost increase absorbed for eight months is real money, and it is almost always discovered by accident.
6 · Customer returns handled three different ways
A returned item can go back on the shelf, into a damaged bin, or into a pile awaiting inspection. Each of those has a different correct treatment, and in most businesses which one happens depends on who was working that afternoon.
The result is not a single error but a drift, in both directions. Saleable returns not put back understate stock; damaged returns put back overstate it. Neither is a large amount and both recur, which is the profile of a difference that never gets investigated because it never looks like anything.
What makes this harder than the purchase-side causes is that there is often no document at all. A supplier credit note exists whether or not anyone acts on it; a customer return may leave no record beyond a note on a delivery sheet. That is a process problem rather than a data problem, and it cannot be reconciled retrospectively — only prevented.
7 · Samples, demos and staff purchases
Stock leaves for reasons that are not sales. A sample to a prospect. A unit opened for a demo. A staff purchase at cost. A damaged item written off informally. A product used internally rather than sold.
Each is trivial and each is one-way: the goods are off the shelf and often still in the ledger. Over a year in a business that does any amount of sampling, the total is not trivial, and it is indistinguishable from theft at the point where anyone looks.
Which is the reason to record them, and it is not primarily an accounting reason. A sample that is recorded is a marketing cost with a number attached; the same sample unrecorded becomes part of a shrinkage figure that gets attributed to the warehouse. One of those is information and the other is a grievance.
8 · The valuation method itself
The last cause is not an error at all, which is why it confuses people the most. FIFO, weighted average and standard costing give different values for identical physical stock. Change method, or apply different methods across product groups, and stock value differs from what a previous calculation produced — legitimately.
It becomes a problem when nobody knows which method is in force where. A business that grew by acquisition frequently has two, inherited from two systems, applied to overlapping product ranges. The resulting difference cannot be reconciled because there is nothing to reconcile: both numbers are correct answers to different questions.
| Cause | Quantity or value | Survives a count? | Has a document? |
|---|---|---|---|
| Unit of measure | Quantity and value | No — count exposes it | Yes, the invoice |
| Freight inconsistency | Value | Yes | Yes |
| Missed credit note | Both | No | Yes, in a folder |
| Cut-off | Value | Yes | Yes |
| Price drift | Value | Yes | Yes |
| Customer returns | Quantity | No | Often not |
| Samples and internal use | Quantity | No | Rarely |
| Valuation method | Value | Yes | It is policy, not error |
The third column is the one to read carefully. Five of the eight causes survive a physical count untouched, which means a business that counts diligently and never reconciles is testing three of eight possible problems.
What is actually shrinkage
Shrinkage is real. Things get broken, things go out of date, and in some sectors things get stolen at rates that are well documented. Nothing here argues otherwise.
The argument is about sequence. Shrinkage is what remains after the documentary causes have been removed, and until they have been, any figure described as shrinkage is really a figure described as «everything we cannot explain». Those are not the same thing, and only one of them justifies putting cameras in a warehouse.
The practical benefit of doing it in that order is that the residual becomes small enough to investigate properly. An unexplained variance across four thousand product lines is not actionable. The same variance, once the documentary causes are stripped out, is often concentrated in a handful of products — and a handful of products is something a warehouse manager can actually work with.
Why the annual count is the worst place to find any of this
The count is treated as the moment of truth. In terms of information it is, and in terms of usefulness it is the worst possible timing, because every cause it reveals is by then a year old.
A credit note that was missed in February is unclaimable in January. A price increase from March is history rather than a decision — the selling price cannot be retroactively raised on eleven months of sales. A short delivery in April cannot be proven, because the person who signed for it does not remember and the supplier has no reason to agree.
The count tells you the size of the problem at precisely the moment nothing can be done about it. That is why it so reliably ends in an adjustment and a shrug: not because nobody cares, but because by then every available action has expired.
Which suggests the obvious rearrangement. Reconcile the purchase side monthly, where it is cheap and where actions are still available, and let the count do the one job only it can do — testing the physical quantity — against a purchase side that has already been settled.
What this does to reported margin
Every stock valuation error has a matching error in cost of sales, and therefore in gross margin. That is what elevates this above housekeeping: an overstated stock value understates cost of sales and flatters margin, and the correction eventually arrives as a write-down that was in nobody's forecast.
Notice that the causes on this page mostly push the same way. Missed credits overstate stock. Cut-off understates recorded cost. Unrecorded samples leave value on the books for goods that have gone. Individually small, consistently in one direction, and cumulatively enough that a monthly gross margin figure becomes something people stop quoting in meetings.
That loss of confidence is the real damage, and it is subtle. Once margin is known to be unreliable, it stops being used — and a business running on revenue and cash alone, without trustworthy margin, makes worse pricing and purchasing decisions for reasons nobody connects back to stock reconciliation.
Which way the errors point, and why that is not random
A detail worth sitting with: the causes on this page do not scatter evenly around zero. Most of them push stock value up and cost of sales down, which flatters margin now and produces a correction later.
The reason is structural rather than sinister. A missed credit note leaves goods on the books that went back to the supplier — value up. Goods received but not invoiced leave stock with no recorded cost — cost down. Samples and internal use leave value for goods that have gone — value up. In each case the omission is of a document that would have reduced something, and documents that reduce things are the ones nobody chases.
Errors that push the other way exist and are rarer, because they require someone to record something that was not there. Over-deliveries and duplicated receipts happen; they are simply outnumbered.
| Omission | Effect on stock value | Effect on reported margin |
|---|---|---|
| Credit note not applied | Overstated | Flattered |
| Goods received not invoiced | Correct, cost missing | Flattered |
| Samples and internal use | Overstated | Flattered |
| Customer return not booked back | Understated | Depressed |
| Over-delivery not invoiced | Correct, cost missing | Flattered |
| Price rise not reflected in cost | Understated | Flattered |
The practical consequence is that unreconciled stock tends to be optimistic, and the correction arrives as a write-down rather than a windfall. Anyone who has watched a stock adjustment appear in a December that was otherwise on plan has seen this mechanism, usually without it being named.
The order worth looking in
Cheapest to check first, because each step removes explanations for what remains.
| Order | Check | Signature to look for |
|---|---|---|
| 1 | Missing documents | A whole product line absent |
| 2 | Unit conversions | Difference is exactly 6× or 12× |
| 3 | Credit notes not applied | One product negative against expectation |
| 4 | Cut-off either side of period end | Stock present, no cost recorded |
| 5 | Unit price history by product | Quantity agrees, value does not |
| 6 | Freight treatment consistency | Value out by a round amount |
| 7 | Samples, returns, internal use | Steady one-way drift, no document |
| 8 | Whatever is left | This is the physical question |
Steps one to six are desk work on documents. Step seven is process. Only step eight requires anyone to walk around a warehouse — and by then the question is specific enough to be worth walking.
What actually fixes it
Count at the door and write the count down. Everything else depends on the received quantity being a real number. A signature meaning «a lorry came» is not a count, and no amount of downstream analysis can supply what was never recorded.
Keep one conversion table. One row per product, supplier unit to your unit. It is the single cheapest prevention for the single most damaging cause.
Extract purchase lines, not totals. A total is enough to pay and useless for stock. Every check on this page needs the lines to exist somewhere.
Look at price history monthly. Not as part of reconciliation but alongside it. Cost increases are only actionable while selling prices can still move.
Stop absorbing the residual. A difference forced to zero has destroyed the evidence that a question existed. Name it, size it, and carry it forward as an open item — it is the only version that ever gets resolved.
None of this requires new software or a project. It requires the purchase lines to be available as data rather than as PDFs, which is the bottleneck the stock purchase reconciliation page addresses, and a method to run through them — which is the guide.
Start with the messiest supplier
Extract one real invoice free — no registration — and see the lines before committing a month to the method.