The problem · The annual stock take
How to count your stock without closing the shop
Once or twice a year you close on a Sunday, bring everybody in, and count the whole shop with pens and printed sheets.
It takes a day you cannot trade, it costs you the staff time, somebody miscounts the back store, and the variance it produces covers six months — so it tells you something is wrong and nothing about when, where or who.
Then it is filed, and the shop goes back to not knowing its stock position until the next one.
Why the big annual count is the wrong tool
A full count is an audit instrument. It is designed to establish a position on a date, which is what a year-end set of accounts needs. It was never designed to find out what is going wrong in your shop, and it is bad at that job in a specific way: the longer the period it covers, the less actionable the result.
A variance over six months has had six months of deliveries, transfers, breakages, price changes and staff rotations mixed into it. Nothing in that number points anywhere. A variance over seven days points at a week, usually at a shift, and often at a single delivery that was short.
There is also a cost nobody counts: the day of trading you lost, plus the overtime, plus the errors that come from tired people counting thousands of lines at 8pm. Large counts done badly can introduce more inaccuracy than they resolve.
- A day closed is a day of takings and a day of wages against no revenue
- A six-month variance cannot be traced to a cause
- Long counts are error-prone precisely because they are long
- Between counts, the shop has no stock position at all
Why shops do it this way anyway
Because on paper, the alternative looks impossible. If the count is done on printed sheets, counting a fast-moving line while the shop is trading is genuinely meaningless — you count 40, three sell while you are walking to the office, and the figure you write down was never true.
That constraint is real, and it is the whole reason full closures exist. It also disappears the moment the system knows what sold during the count, because the count and the sales can be reconciled against the same clock.
The second reason is cultural: the annual count is what the accountant asks for, so it is treated as the stock process rather than as one output of it. A shop running weekly rolling counts can still produce a year-end position — a far more credible one, because it has been continuously checked.
- Paper counts cannot cope with a line selling while it is being counted
- No live figure to count against, so every count starts from nothing
- The annual count is mistaken for the whole process
- Nobody has decided which lines actually matter
Step by step
How to run counts in a working shop
Rank your lines by what a loss costs you
Value times movement. Airtime, spirits, phone accessories, baby formula, cooking oil, batteries — the high-value or high-velocity lines are where variance is both most likely and most expensive. Everything else can be counted far less often.
Count a small set, weekly, at the same time
Ten to twenty lines, early morning before the rush or right after close. Rotate so every significant line is counted monthly and your riskiest lines weekly. The discipline is the point: a count that happens is worth more than a thorough one that keeps being postponed.
Count against a live figure, not a blank sheet
Counting blind and comparing later doubles the work and loses the context. Counting against what the system says it should be turns a count into a yes-or-no question per line, which is dramatically faster and catches the miscount while the counter is still standing at the shelf.
Count one section at a time, while trading continues
Fence off a shelf, not a shop. If the system accounts for sales that happen during the count, a line selling while you count it is no longer a problem — which is the single constraint that made full closures necessary.
Investigate the variance the same day
While the week is still fresh: check deliveries against orders, look for transfers that left without arriving, check for voids and refunds in the period. A variance investigated within a day usually has an explanation; one investigated in six months never does.
Keep the full count, but make it a formality
Do one annually for the accounts. If the rolling counts have been running, it should confirm a position you already know rather than reveal one you did not — and that is exactly how a stock take is supposed to feel.
What you get
What sell.ke does about it
Partial counts, on a phone
Count a shelf or a category against the live figure without touching the rest of the shop. No printed sheets, no re-keying, no evening spent typing what somebody wrote down.
Blind counts when you want them
Hide the expected figure from the counter so the count is a genuine measurement rather than a confirmation of what the screen already said. Useful precisely where you most suspect a problem.
The shop keeps trading
Sales during the count are accounted for against the same clock, so a fast-moving line can be counted at 9am on a Tuesday instead of at 10pm on a Sunday.
A variance with a shape
The report names the product, the branch and the period, and sits next to the deliveries, transfers, voids and refunds from that same window — so investigating is reading, not excavating.
Counts per branch
Each location counts its own stock against its own figures, while the catalogue stays shared. A transfer that left one branch and never arrived at the other is visible as exactly that.
Adjustments that leave a name
Every correction records who made it and why. A stock figure that was changed is a decision in the audit trail rather than a number that quietly became different.
Questions
Stock takes that close the shop — questions
How can a count be accurate if the shop is still selling?
Because the sales during the count are recorded against the same clock as the count. The system knows that three units sold between the start of the count and its submission, so the comparison is made correctly. That is precisely what a paper count cannot do, and it is the only reason shops close to count in the first place.
What is a blind count and should I use one?
A blind count hides the expected quantity, so the person counting cannot anchor to it. Use it on your highest-risk lines and anywhere you suspect counts are being completed rather than performed. Use the non-blind version everywhere else, because it is faster and catches miscounts at the shelf.
How often should each line be counted?
Weekly for high-value or fast-moving lines, monthly for the main body of stock, quarterly or annually for the long tail. The aim is that every significant line has been checked recently enough that a variance can still be traced to something.
My variance is always negative. Is that theft?
Not necessarily, and the first place to look is receiving rather than selling. Short deliveries that were never checked against the order produce a permanent negative variance that looks exactly like theft, and unrecorded breakage and staff consumption do the same. Record those properly and the remaining gap is both much smaller and much more meaningful.
Do I still need an annual stock take?
For your accounts, usually yes. But it changes character: instead of being the one moment you discover your stock position, it becomes a confirmation of a position you have been maintaining all year — and it takes a fraction of the time, because the numbers are already close.
Can I use it for more than one shop?
Yes. Branches share one product catalogue but hold their own stock, so a transfer between them is a recorded movement rather than a re-count. Reports run per branch or across all of them, and staff permissions are set per branch — a Kisumu supervisor does not need to see Nairobi's margins.
How long does setup take?
A single-counter duka is usually selling the same day. The work is the product list, not the software: import a CSV or let Amina create products from a supplier invoice or a photo of your price list. A multi-branch business with thousands of SKUs, opening stock counts and an eTIMS enrolment should plan a week and run the old system in parallel for a fortnight.
Can I stop staff from giving discounts or deleting sales?
Yes. Discounts, voids and refunds sit behind a manager PIN, and every one of them records who authorised it and when. Staff accounts carry role permissions, so a cashier can sell without seeing cost prices, editing products or opening reports. A void that reverses stock is a movement in the audit trail, not a gap in it.
The problems that travel with this one
These rarely arrive alone. Fixing one usually exposes the next.
Stock going missing
Why shrinkage is almost never one big theft, and what actually closes the gap.
Dead stock and expiry
The shelves are full and the bank is empty. That is not a coincidence.
The second shop
The second branch is where most retail businesses discover their systems were held together by the owner.
The features that do the work
Stock you can trust
Batch and expiry, serial numbers, reorder levels, stock takes and the formulas behind them.
Buying, suppliers and approvals
Purchase orders, goods-received notes, supplier bills and the cost figure every margin depends on.
Branches, staff and control
Shared catalogue, separate stock, transfers that are movements, and permissions per branch.
Reports and real accounting
Double-entry books — trial balance, P&L, balance sheet — not a CSV export to somebody else's software.
If you are outside Kenya
The same problem, under the rails and the revenue authority you actually deal with.
Fix it on your own numbers
Fourteen days, no card. Import your product list or let Amina build it from a photo of your price list, and see what the reports say about your own shop.