A stocktake is a full physical count of everything you hold, checked against what your records say you hold. Known in the United States as a physical inventory count, it verifies stock quantities for the accounts and reveals discrepancies caused by mis-picks, damage, theft and unrecorded movements.
It is the moment the system stops being taken on trust. Everything else in inventory — reorder points, forecasts, valuations, delivery promises — rests on stock figures that are assumed correct. A stocktake is the only thing that establishes whether that assumption holds.
Most operations run one annually because an auditor asks for it, treat it as a compliance chore, and extract almost none of the value available from it. That is a missed opportunity, because the discrepancies a stocktake reveals are a map of where your processes are failing.
What a stocktake is
The defining feature is completeness. A stocktake counts everything, at a single point in time, so the result is a verified position across the whole operation rather than a sample.
The terms vary and mean roughly the same thing:
- Stocktake — the common term in the UK, Australia and much of the Commonwealth.
- Physical inventory or physical count — the usual American term.
- Wall-to-wall count — a full count of the entire site, emphasising that nothing is excluded.
- Annual count — a stocktake performed for the financial year end.
A stocktake differs from cycle counting, which counts a portion continuously so the whole site is covered over a period without ever stopping. More on that comparison below.
Why stocktaking matters
It establishes the true position. Records drift. Every unrecorded movement, mis-pick, damage not written off and transcription error accumulates, and only a count reveals the total.
It supports the accounts. Closing stock is usually a substantial asset on the balance sheet, and it feeds directly into cost of goods sold and therefore reported profit. Auditors need evidence that the figure is real.
It exposes shrinkage. Theft, damage and loss are invisible until counted. Knowing the scale is the first step to addressing it.
It reveals process faults. This is the part most often wasted. The pattern in your discrepancies — which products, which locations, which shifts — tells you where your operation is failing. Adjusting the numbers without reading the pattern throws that away.
It finds stock you did not know you had. Operations with poor records frequently discover significant value in forgotten locations, and stop reordering things they already own.
How to do a stocktake, step by step
Before
1. Plan the count and choose the timing. The quietest period you have, and ideally when stock is at its lowest — fewer units means less counting and fewer errors. Year-end is common for accounting reasons; it is rarely the operationally easiest moment.
2. Tidy and organise. Consolidate part-pallets, put away anything in the wrong place, clear the receiving bay, and resolve open transactions. Counting a disorganised warehouse takes several times longer and produces worse results. This preparation typically does more for accuracy than anything on the day.
3. Freeze movement. No receiving, no picking, no dispatch during the count. Where a full stop is impossible, segregate anything that must move and count it separately, recording the transaction clearly.
4. Process all paperwork. Every receipt booked, every dispatch confirmed, every adjustment posted. Stock physically present but not yet in the system, or vice versa, becomes a discrepancy that is not real.
5. Prepare count sheets or devices. Organised by location, in the sequence someone will walk. Counting in a logical route is faster and misses fewer locations than hunting for items on a list sorted by product code.
6. Assign roles and brief the team. Somebody owns the count, somebody records, somebody adjudicates the awkward locations. Then brief everyone: how to count, what to do with damaged goods, what to do when a location holds something unexpected, and who to ask. Ten minutes of briefing prevents hours of reconciliation.
6b. Check the tools. Count sheets printed or scanners charged and tested, pens that work, a ladder where you need one, and a means of marking a location as done. A count that stops for twenty minutes because a device has no battery loses more than the twenty minutes.
During
7. Count blind. Do not show counters the expected quantity. People unconsciously confirm what they are shown, and a blind count is the single most effective accuracy measure available.
8. Count by location, not by product. Walk the building systematically, counting what is in each location. A product held in three places gets counted three times, and the sum is compared afterwards. Counting by product invites people to stop looking once they reach the expected total.
9. Mark locations as counted. A visible marker on each completed bay prevents double-counting and missed locations, which are the two most common failures in a large count.
10. Record damaged and unsellable stock separately. It exists physically and should not be valued as sellable.
11. Count in pairs where practical. One counts, one records. Slower per location and materially more accurate.
After
12. Reconcile against system records and produce a variance report. This is the comparison the whole exercise exists to produce, so give it the time it needs rather than treating it as paperwork after the real work.
13. Recount the significant variances before adjusting anything. The second count disagrees with the first surprisingly often, and adjusting on a single count writes errors into the record.
14. Investigate before you adjust. Check adjacent locations, the returns area, quarantine, and the transaction history. A large share of apparent losses are misplacements that resolve by looking.
15. Categorise the causes using a fixed list — mis-pick, receiving error, unrecorded movement, damage, theft, transcription error. Free text cannot be counted.
16. Post the adjustments and update the system, with authorisation recorded. Until the records are updated the count has changed nothing, and a variance report sitting unposted for a fortnight is worse than not counting, because everyone now knows the figures are wrong and is still working from them.
17. Review the patterns and act. Which products, which locations, which reasons. This is where the value is, and it is the step that gets skipped when everyone is tired and the count is technically finished.
18. Plan the next one. Note what went wrong with the process itself — the locations nobody could reach, the products with no label, the paperwork that arrived late — and fix those before the next count rather than rediscovering them on the day.
Stocktake vs cycle counting
A stocktake counts everything at once. It gives a complete verified position, which is what auditors want, and it requires stopping the operation. Its weakness is timing: an error found at year end happened at some point in the preceding twelve months and is untraceable.
Cycle counting counts a small portion continuously — a few locations a day — so the whole site is covered over a period without ever stopping. Count high-value and fast-moving items often, slow-moving items rarely.
The decisive advantage of cycle counting is not accuracy but proximity to cause. An error found three days after it occurred can be traced through the transaction history to a specific movement, and the process fault behind it can be fixed. The same error found eleven months later is just a number to adjust.
Most well-run operations do both: cycle counting continuously as the working control, and a full stocktake where an auditor requires one or where accuracy has degraded enough that a clean baseline is needed.
If you can only do one, cycle counting delivers more operational value. If your auditor requires a full count, you may not have the choice.
Reading the variance report
The variance report is the output, and most operations glance at the total and file it. The total is the least useful number on it.
Look at direction, not just size. Variances that net to near zero can conceal large offsetting errors — a thousand units missing in one place and a thousand extra in another is a location discipline problem, not a loss. A report showing consistent negative variance across the site is shrinkage. One showing scattered positives and negatives is a recording problem.
Sort by value, not by unit count. A variance of two units on a high-value item matters more than two hundred on a low-value consumable. Investigating in unit order wastes the investigation effort on the cheapest problems.
Group by location. Variances clustering in one aisle, one level or one zone point at something specific — a labelling fault, a picker working that area, a rack that gets knocked by traffic. Variances spread evenly point at a systemic recording problem.
Group by product characteristics. Small, high-value, easily pocketed items behaving differently from bulky ones suggests theft. Items with similar packaging or adjacent part numbers behaving badly suggests pick confusion rather than loss.
Compare with the last count. The same products showing variance repeatedly is the single strongest signal available. A product wrong at three consecutive counts has a process fault attached to it, not bad luck.
Calculate accuracy properly. Inventory record accuracy is the share of counted locations matching exactly — not the net value variance, which flatters badly. An operation can show a 0.4% value variance whilst only 70% of its locations are correct, and the second figure is the one that predicts failed picks.
Three or four findings from a variance report, acted on, will do more for next year’s count than any change to how the counting itself is performed.
What a stocktake costs
Worth being explicit, because the cost is usually underestimated and it shapes how often one should be done.
Lost operating time. For most operations this is the largest cost by a distance. A day of no dispatch means a day of orders delayed, carrier collections missed, and a backlog to clear afterwards. In a business shipping several thousand orders a day, this dwarfs everything else on the list.
Labour. Counters, recorders, supervisors, and the reconciliation work afterwards, which routinely takes longer than the counting.
Preparation time. Tidying, consolidating, clearing paperwork. Real hours, and the ones that most affect the result.
Overtime or temporary staff, where the count runs outside normal hours.
The reconciliation tail. Investigating variances, recounting, posting adjustments, and the management time reviewing it all. This frequently continues for a week or more after the physical count finishes.
Against that, the benefits are a verified balance sheet figure, a quantified shrinkage number, and the process findings. The first two are usually required rather than optional. The third is where the return sits, and it is the one forfeited by treating the exercise as compliance.
This cost profile is the practical argument for cycle counting. Spreading the same counting effort across the year avoids the lost operating time almost entirely, which is the expensive part.
Stocktake best practices
- Preparation beats effort on the day. A tidy, consolidated warehouse with paperwork cleared counts faster and more accurately than a chaotic one counted by more people.
- Count blind, always.
- Do not let pickers count their own areas. The conflict is obvious when a discrepancy would reveal their own error.
- Use the same unit of measure as the system. Counting cases where the system holds eaches is the most common arithmetic error in a stocktake.
- Watch for stock in transit — dispatched but not yet collected, received but not yet booked in. Decide in advance whether it is counted and be consistent.
- Do not skip the quarantine and returns areas. They hold real stock and are routinely forgotten.
- Keep the count sheets. Auditors may ask, and you may want to re-examine a variance later.
- Debrief afterwards. What went wrong, what took longer than expected, what would be done differently. Stocktakes repeat, and the notes are worth more than anyone’s memory a year later.
Those are the habits that separate a count you can rely on from a count you have to repeat. None of them is difficult; they are simply the practices that get dropped first when a count is running late.
Counting methods and technology
How you capture the numbers affects both speed and accuracy more than most people expect.
Paper count sheets remain common and work perfectly well for smaller operations. Organised by location in walking sequence, they need no hardware and no training. The weakness is transcription: every figure gets written once and typed once, and each step is an opportunity for error. If you use paper, enter the data the same day whilst it can still be queried.
Handheld scanners confirm the location and the product by barcode, so the only thing typed is the quantity. This removes the most common error — counting the right number of the wrong product — and it is the single biggest improvement available to a manual count.
Mobile devices running the WMS go further, validating the count against the expected figure in real time and flagging variances whilst the counter is still standing in front of the location. That immediacy is valuable, though it conflicts with blind counting, so most operations configure it to accept the count without displaying the expected quantity and flag variances to a supervisor instead.
Weigh counting is useful for small identical items — fasteners, components, sachets. A calibrated scale counts thousands of units in seconds with better accuracy than a person. It depends on consistent unit weight, so it suits manufactured components and not variable natural products.
RFID reads many tagged items at once without line of sight, and where it is already deployed it transforms counting — a full count that took a night can take an hour. It is rarely worth deploying for counting alone, but where tags exist for other reasons the counting benefit is substantial.
Drones and automated scanning exist for high-bay storage where reaching top levels is slow and hazardous. Genuinely useful in very large operations, and disproportionate for most.
The pragmatic sequence for an operation improving its counting is barcode scanning first, because it addresses the largest error source for the least investment. Everything beyond that is optimisation.
Where to go deeper
- Cycle counting — the continuous alternative.
- What is inventory control — the discipline a stocktake verifies.
- What is inventory shrinkage — what the variance often reveals.
- Inventory valuation methods — how the counted stock gets valued.
- How to manage inventory — where counting fits in the wider process.
The one-line takeaway
A stocktake is a full physical count verifying that your records match reality — prepare the warehouse before you count, count blind and by location, investigate variances rather than adjusting them, and read the pattern in the causes, because that is where the value is.
Frequently asked questions
What is a stocktake? A full physical count of all inventory held, compared against system records to identify and correct discrepancies. Also called a physical inventory or wall-to-wall count.
How often should you do a stocktake? At minimum annually where an auditor requires it. Operations running a good cycle counting programme often need nothing more than that, because accuracy is being verified continuously. Where accuracy is poor or unknown, a full count establishes a baseline to improve from.
What is the difference between a stocktake and cycle counting? A stocktake counts everything at once and usually requires stopping operations. Cycle counting counts a portion continuously without stopping. Cycle counting finds errors close to when they happened, which makes the cause traceable; a stocktake finds them months later.
How long does a stocktake take? Anything from a few hours for a small operation to several days for a large multi-site one. Preparation has more effect on the duration than headcount does — a tidy, consolidated warehouse with cleared paperwork counts far faster, and adding more counters to a disorganised site tends to increase errors rather than reduce time. Budget separately for the reconciliation afterwards, which routinely takes longer than the counting itself.
Do I need to stop operations for a stocktake? Ideally yes, because counting a moving target produces numbers nobody trusts. Where a full stop is impossible, segregate anything that must move and record those transactions separately so they can be reconciled afterwards.
What is blind counting? Counting without being shown the expected quantity. It prevents the unconscious confirmation bias that occurs when someone counts towards a number they have already seen, and it is the most effective single accuracy measure in a stocktake.
Who should carry out the stocktake? Ideally not the people who normally pick and put away the areas being counted, because a discrepancy would reveal their own error. Rotating counters across unfamiliar zones, or bringing in staff from another function, removes that conflict. Some operations use external counting firms for the annual count, which also satisfies auditors looking for independence.
What should I do with discrepancies? Recount first, then investigate before adjusting — check adjacent locations, returns, quarantine and the transaction history. Categorise the cause using a fixed list, then post the adjustment. The pattern across all your discrepancies is more valuable than any individual correction.
Warehouse Wise is a free, independent library of guides on inventory management, warehousing and fulfilment, written for the people actually doing the work — new warehouse and inventory staff, small and growing retail and ecommerce businesses, and anyone who holds stock and wants to understand it properly. Every guide explains the term, shows a worked example with real numbers where it helps, and stops. We are independent: we do not sell software or equipment, we take no commission on referrals, and our guides are not ranked by who pays us.