How to count stock without closing the shop
Cycle counting finds stock errors within days instead of once a year. How to choose what to count, and why a count must post the difference, not the total.
For anyone whose annual stocktake costs a weekend and still does not agree · 3 min read · updated
In short
- Cycle counting finds errors within days; an annual count finds a year of them at once, too late.
- Rank by consumption value: count the top group monthly, the tail annually.
- A count must post the difference, not the total, or a concurrent sale is erased.
- Do not print the expected quantity on the count sheet.
- Repeated variance on one item usually means a unit-of-measure mistake, not theft.
The annual count is the expensive way
A full physical count shuts the business, takes a weekend, and finds a year of errors at once — with no way to tell which of the 300 trading days each one happened on. It also arrives too late to fix anything.
Cycle counting spreads it: a handful of items every week, chosen so that fast movers and expensive items come round more often than the slow tail. Nobody closes, errors surface within days of being made, and the annual count becomes a confirmation rather than an excavation.
Which items, how often
The usual approach is ABC: rank items by annual consumption value. The top group is perhaps 20% of items and 80% of value — count those monthly. The middle quarterly. The long tail once a year.
Add a second rule that beats the classes: count anything that has gone negative, anything with an adjustment in the last cycle, and anything a person has complained about. Those are where the errors actually are.
The rule that makes counts stick
A count must post the DIFFERENCE, not the total.
It sounds like a distinction without one. It is not. If a count writes "set this item to 47", and a sale happens between the moment somebody counted the shelf and the moment they typed the number in, that sale is erased. The system says 47; the shelf has 46; and the correction created the error it was fixing.
Posting the difference — "found 2 fewer than the system thought" — survives concurrent movement, because it is a relative adjustment applied to whatever the balance is when it lands. Ask any system how it records a count. This is the single most useful question you can ask about one.
Counting blind
Do not print the system quantity on the count sheet. A person holding a sheet that says 47 counts to 47 — not through dishonesty, but because the brain stops at the expected answer.
Enter the counted figure and let the system show the variance afterwards. If your process requires a recount, trigger it on the variance, not on the counter’s judgement.
What to do with the difference
Write it off promptly and look at the pattern rather than the item. Consistent shortfalls on small high-value goods is shrinkage. Consistent shortfalls on one item is usually a unit-of-measure problem — somebody receiving in boxes and issuing in pieces.
A variance you investigate is worth having. A variance you simply post is a number you will post again next quarter.
Common questions
- How often should stock be counted?
- By value rather than by calendar. High-value fast movers monthly, mid-range quarterly, the long tail annually — plus anything that has gone negative or been adjusted recently, whichever class it belongs to.
- Does cycle counting satisfy an auditor?
- Generally yes, where the programme is documented, covers all items over the year, and variances are investigated and recorded. Auditors care that the system is reliable, not that everything was counted on one day.
- What is the difference between a cycle count and a stocktake?
- A stocktake counts everything at once, usually with the business closed. A cycle count covers a small set of items each week or month, so that across a year everything is counted without ever stopping trade.