A full count costs a Saturday, six people, and the goodwill of everyone who had plans. That is the price. The question worth asking before you pay it is what you are buying, because a full count answers one question very well and several others not at all.
This page covers what a physical inventory count actually produces, how it differs from the perpetual figure it is testing, when it earns the shutdown, and how to run one so the day is not wasted. The sheet and the day itself are covered in how to run a stocktake.
What is a physical inventory count?
A physical inventory count establishes what is actually on the shelf by counting all of it, usually with the operation stopped so nothing moves mid-count. It produces one dated figure everyone can sign, which is a different product from the continuous accuracy a rolling count programme gives you.
The word that matters is dated. The output is a position at a moment, and its value decays from that moment onward. A count on the 31st is authoritative on the 31st and merely indicative by the middle of February, which is why treating it as the annual accuracy programme rather than as a specific deliverable is where most of the disappointment comes from.
For a brand shipping to stockists the dated figure has a second use: it is the number a buyer's audit, a lender or an insurer will accept, because somebody stood in the room and counted it. The perpetual figure, however good, is a record of what you told the system. Keep the signed sheets with the date on them; that is the artefact the count was for.
Perpetual vs physical: what is the difference?
The perpetual figure is what your system believes, updated as things move. The physical count is what is there. Perpetual is a claim built from every recorded movement; physical is the test of that claim. Neither replaces the other, and the interesting number is the distance between them.
| Perpetual | Physical count | |
|---|---|---|
| Source | Recorded movements | Somebody counting |
| Available | Continuously | On the day |
| Fails when | A movement goes unrecorded | It is out of date, which is immediately |
| Costs | Discipline at each movement | A day of operations |
Put that way, the pairing is obvious: the perpetual figure is what you run the business on, and counting is how you keep earning the right to trust it. A business that only ever counts has no working figure between counts. A business that never counts has a figure nobody has tested.
When does a full count earn the shutdown?
When you need one dated number that everyone accepts at once: a year-end valuation, a warehouse move, a system migration, or the first count in an operation with no trustworthy history. Outside those, a rolling programme delivers better accuracy for a fraction of the disruption.
The first-count case is the one people underrate. An operation whose figures have drifted for two years does not have a variance problem it can cycle its way out of quickly; it has an unknown starting position. One full count to establish a baseline, then a rolling programme to hold it, is far cheaper than eighteen months of counting slices that keep disagreeing with a history nobody trusts.
Pick the date for the stock, not only for the accounts. The week after the holiday peak, when the room is emptiest, costs a fraction of what the same count costs in November with every bay full. If year-end falls in the busy season, agree with your accountant on a count a few weeks early and a roll-forward from recorded movements, which auditors accept when the records in between are clean.
How do you run one without wasting the day?
Freeze movement completely, count in pairs with one person counting and one recording, count blind by hiding the expected figure, and recount every variance above a value threshold before anyone adjusts anything. Most wasted counts are wasted because the counters could see what the answer was supposed to be.
Blind counting is the single highest-return rule and the one most often skipped, because printing the expected quantity beside each line feels helpful. It is not: it converts counting into confirming, and a tired person at hour five confirms. Hide the number, take the count, compare afterwards.
Then treat the output as the start rather than the end. Every variance gets a cause, exactly as a variance during the year does, and the day after the count is when you set the rolling cadence, ranking what to count often by value at risk per counting hour, so that next year's count is a formality rather than a discovery.
What should you do next?
Before
- Ask what the count is for: control, or one dated figure.
- Freeze movement, and mean it.
- Print sheets without the expected quantity.
On the day
- Count in pairs: one counts, one records.
- Recount every variance above your value threshold.
- Adjust nothing until the recounts are done.
After
- Give every surviving variance a cause.
- Set the rolling cadence the following week, while it still stings.
Frequently asked questions
When is a physical inventory count usually taken?
Most brands run one at financial year-end, and again whenever stock moves premises or systems. The date is chosen for the accounts rather than the warehouse, which is why it usually lands at the least convenient point in the trading year.
What is the difference between perpetual and physical inventory?
Perpetual is the running figure your system maintains as things move. Physical is what is actually on the shelf, established by counting. The perpetual figure is a claim; the physical count is the test of it, and the gap between them is the thing worth investigating.
Do I still need a full count if I cycle count all year?
Often not for control purposes, but sometimes yes for a specific event: a year-end valuation, a warehouse move, a system migration. Ask what the number is for. If it is to prove control, cycle-count records do it better. If it is one dated figure everyone signs, count.