Counting stock without closing the shop
A stocktake fails for boring reasons: it takes a day you don't have, and the till keeps selling while you count. Both are solvable.
Most repair shops know their stock figure is wrong. They also know roughly when it went wrong — around the time somebody took a screen off the shelf for a job and meant to write it down.
The reason it stays wrong is not laziness. It is that a stocktake, as most shops imagine it, means closing on a Sunday, counting everything, and typing numbers into a spreadsheet that is out of date by Tuesday. Nobody has a spare Sunday, so the figure drifts for another year.
What actually makes a count hard
Three things, and only one of them is the counting:
You are racing the till. If you count the accessory wall at 10am and sell two cases at 11am, your count is wrong by the time you enter it — but you have no way to know whether the difference is a sale or a discrepancy. So the whole count becomes untrustworthy, and an untrustworthy count is worse than none, because you will act on it.
It is all or nothing. A count that must cover the entire shop can only ever happen when the entire shop is closed. That is why it never happens.
Nothing separates the finding from the fixing. In a spreadsheet, correcting the figure and discovering the variance are the same keystroke. So the variance — the actually interesting bit, the thing that tells you where stock is going — is destroyed at the moment it is found.
Snapshot, count, approve, apply
The fix is to make those four things four separate steps rather than one.
Snapshot. When the count opens, the system records what it currently believes is on the shelf, right then. That number is frozen. Everything you count is compared against the snapshot, not against a figure that keeps moving — so the two cases you sell at 11am are a known movement rather than a mystery variance. You are no longer racing the till.
Count. Enter what is physically there, line by line. Lines you have not reached should stay unreached, not zero. This sounds obvious and is the single most common way a count destroys good data: an empty box treated as "none in stock" will happily write off a shelf you simply did not get to.
Approve. The variances appear — short, over, and what the difference costs. Nothing has changed yet. This is the step that is usually missing, and it is the one that makes a count useful: a manager looks at a list of differences and asks why, before the numbers are overwritten.
Apply. Only now does stock actually move, and each correction posts as its own movement with a reason attached. Next month, "why did we write off four screens in August" has an answer.
Count a shelf, not a shop
Once a count can be scoped, the Sunday problem disappears.
Count the accessory wall on a quiet Tuesday. Count one supplier's parts the week their statement arrives. Count the used handsets monthly, because that is where the money is, and the cases twice a year, because it is not.
A small count you actually do beats a full count you keep postponing. It also narrows the investigation: when twelve lines are out, you can go and find out why. When four hundred lines are out, you shrug and accept the number.
The variance is the point
It is tempting to treat a stocktake as an accounting chore — get the number right, move on. But the corrected number is the least valuable thing it produces.
The valuable thing is the pattern. Screens short and cases fine means something different from everything short by a bit. A line that is over means something was booked in twice or a return went back on the shelf without a record. Repeated shortages on one category, month after month, is a process problem with a location, and you now know where to look.
That pattern only exists if variances are recorded as variances. If the count silently overwrites the figure, you have bought yourself a correct number today and learned nothing.
The practical minimum
If you do nothing else:
- Count one category a month rather than the shop once a year.
- Freeze what you are comparing against before you start.
- Leave uncounted lines uncounted.
- Have someone other than the counter approve the differences.
- Keep the variance, not just the correction.
None of that needs a closed shop. It needs about forty minutes and a decision to stop treating the stock figure as something that will sort itself out.
How a scoped count works in SlickCell Pro — snapshot on open, submit, manager approval, then apply — is written up step by step in Run a stocktake. The wider model, where devices are tracked one unit at a time and parts by quantity, is on Inventory & device units.