Most of what follows is true whether you count with barcode scanners or a clipboard and a biro. It is the method, not the software — thirty-five years of watching counts go well and go badly, written down in the order you need it. What we sell comes in at the end, and only where it actually changes the answer.
Every bad stocktake we have seen went wrong at one of these four points, and in every case the decision was made accidentally rather than made.
A full count of every line, once a year, is what most businesses mean by a stocktake and what most accountants expect. It is also the version that takes the longest and gets the most tired towards the end.
The alternative is counting the valuable and fast-moving lines often — monthly, or quarterly — and the long tail once a year or not at all. That is a real strategy, not a shortcut, but it is a decision to take with whoever signs your accounts, because it changes what your stock figure is evidence of.
Most people freeze, and they are right to. Nothing in, nothing out, count it, open the doors. It costs you a day of trading and it removes every argument that would otherwise follow. If you are choosing between the two and have no strong reason, this is the one.
Counting while you trade is cheaper on the day and leaves you with a question nobody can answer afterwards: was that pallet counted before it shipped, or after?
If you do trade through it, decide what happens to stock that moves before it moves. The usual answer is the only one that works: nothing leaves a counted area until the count is signed off, and anything that must go out is written down separately and reconciled by one person. An agreement made at 11am on the day is not an agreement, it is an argument.
This is the uncomfortable one. The person who runs an area knows it best and counts it fastest, and is also the person least likely to find a discrepancy in it. Not through dishonesty, usually. They know what should be there, so that is what they see.
You do not need to make a drama of it. Swapping two people between two aisles costs nothing and removes the question entirely.
In a real warehouse the same part sits in a pick face, on a pallet up top, and in a box somebody put by the door. Three counts of that part are three correct answers, and the total is the sum. Decide now that a second count of a part is an addition rather than a correction, and make sure whoever is counting knows which one they are giving you.
Get this wrong and the total is quietly too low, which is the worst kind of wrong: it looks like a number.
A part list makes people hunt. A location tells them where to stand. It is the same stock and the same afternoon, and the difference in how long it takes is not small.
Work down a parts list and every line is a search of the whole building. Give somebody an aisle instead and they clear it once. The stock does not move; the walking does.
How they work down it is theirs. Top shelf first, one side and back up the other, whatever the racking suggests — people find their own way and it is quicker than any route you could hand them. You are assigning an area, not a sequence, and nothing should make them scan in a particular order.
It does not have to be racking. The floor by roller door two is a location. The back of the van is a location. If stock can be there at 5pm, it needs a name and a label, or it will not be counted.
"There is nothing in J14" and "nobody went to J14" look identical at the end of the day and mean completely different things. Record empty deliberately, so the difference survives.
This is the part that decides whether the rest works. An unlabelled bin is an uncounted bin, and no software fixes it from the office. Labelling racking before a count →
If the paper says the bin should hold twelve, you will be told twelve. People cheat. Not everybody and not out of malice, but hand somebody a number and a shortcut together, four hours into a long day, up a ladder, and a good proportion will take it. They stop at twelve. They decide the twelfth is behind something. They round the nearly-full box up to a full one.
The whole value of a stocktake is that it is independent of what the system believes. Hand the counter the expectation and you have not counted your stock — you have spent a day and a shift's wages confirming your own records.
And decide the largest number anyone can type. Not because people are careless. Because people invent conventions, and a convention outlives the system that made sense of it.
One customer has counted every year since 2010. Their operators type 9999 when a bin is empty. In the software they were using then, that was correct: 9999 meant I came, I looked, there is nothing here. The people who started doing it have long since left. Nobody ever wrote it down. The habit is fifteen years old and, from where the operator is standing, it is still the right answer.
With no ceiling on the count box, one empty shelf puts 9,999 units into your stock figure and nothing on the report looks unusual. With a ceiling, the scanner refuses it in front of the person who typed it — while they are still at the bin, and can tell you what they actually meant.
A stocktake that is exciting on the morning was prepared badly. Everything below is deliberately front-loaded, because every one of these problems is cheap to fix a week out and expensive to fix with twelve people standing around. If your count has to land on a fixed date, there is a separate run-up for 30 June.
Bins named, labels up, at a height a person can reach and read. This is the job that always takes longer than anyone thinks, which is why it goes first and not last.
Export it and look at it. Duplicated codes, parts that were deleted in 2019, descriptions that are blank, barcodes in the part-number column. Every one of these is a five-minute fix now and a person standing in an aisle later.
Have the counters scan actual stock and actual labels during their normal week, with no count running. Bad data and unreadable labels surface while there is still time to do something about them. This is the step most people skip and the one that saves the most.
Every device and every spare battery. A scanner that dies at 2pm costs you an operator for the rest of the day, and the flat one is always the one somebody left out of the cradle.
Final import of parts and locations first thing, then begin. Reloading reference data halfway through a count is the one thing guaranteed to make the totals arguable.
Which areas are finished, what has stopped, and how much each person has actually counted. It is not a race — somebody in bulk storage will always out-count somebody in small parts, and the fast one is not the better one. But it does show you the person who is not counting at all.
Why the per-operator graph exists at all. We built it for a company counting across a weekend, with everyone on penalty rates. More than a dozen people, a system none of them had used before, and us on site the whole weekend in case it went wrong. It was expensive before a single item had been scanned.
Almost everybody did exactly what you would hope. Eight hours on their feet, hundreds of items each, working fairly and carefully. One man counted four items in two hours. He had not come to count. He was sent home before lunch, and what that saved paid for the day.
That is the whole purpose of it. Not to rank the people doing the job properly — they are all doing the job properly, in aisles that are not comparable. It is so the one person who is not shows up while there is still a shift left to fix it, instead of on Monday when the invoice arrives.
That graph, and the rest of what a supervisor watches while a count is running, is on the Dashboard — progress by aisle and level, who has stopped and for how long, and the counts worth a second look.
Not when everybody says they are done. This is the step that separates a stocktake you can sign from a pile of numbers.
A part your records expect and nobody scanned. Either it is gone, or nobody went and looked — and those need different phone calls.
The list of bins with no answer at all. It is usually short, and it is usually the awkward ones: the top shelf, the mezzanine, the container out the back.
Kept separate from the two above, because an empty bin somebody stood in front of is a finished job and an unvisited bin is not.
Only once those three lists are dealt with is the count worth comparing to anything. Do it the other way round — reconcile first, chase the gaps after — and you will spend a week investigating variances that were only ever somebody not having been up a ladder.
Put what you counted next to what your accounting or inventory system expected, line by line. Short, over, agreed, and the two lists that matter most: things the books expect that nobody found, and things on the shelf the books have never heard of.
The adjustment itself belongs in your accounting system, where the audit trail is, and it is a conversation for whoever signs off your accounts. Before anybody says the word shrinkage, work down the ordinary causes first — units of measure, a part in two bins, paperwork still in flight. What a variance usually turns out to be → The stocktake's job was to produce evidence, and it has done it.
If you are on MYOB AccountRight, the Inventory Count Sheet loads into a stocktake as it comes and the comparison is done for you. Counting against a MYOB file →
We are not going to pretend a clipboard cannot count a small store. It can. These are the four places where scanners stop being a luxury.
Most stocktake errors are transcription: a digit dropped out of a long code, a 0 for an O, the line above copied twice. A trigger pull removes that category of mistake rather than reducing it.
Unscanned parts and unvisited bins, live, at 3pm — when somebody can still walk back and fix it. On paper you find out the next morning, once everyone has gone home and the stock has moved.
The day after a paper count is somebody at a keyboard entering hundreds of lines, introducing a fresh set of errors into numbers that were correct when they were written down.
Not to police anybody — so that a figure that looks wrong can be asked about instead of argued about, and so a withdrawn count leaves a trace rather than vanishing.
Ready to do it with scanners? The setup checklist is this method with our portal in it, step by step. Counting once or twice a year? Hire the scanners rather than buying them. And the software is priced per stocktake, so a business that counts in June pays in June and owns nothing in July.
Tell us roughly what you hold, how many people you can put on it and when it has to be done. We have been doing this since 1991 and will tell you plainly if scanners are not worth it for your count.