Most businesses count their stock once a year, because the accounts require it. The count takes a weekend, everyone is tired, and the differences it finds are a mystery: nobody remembers what happened to 40 units of a product eleven months ago.
Cycle counting is the alternative. Instead of counting everything once, you count a part of the stock every week or every day, so every product is counted several times a year and every error is found while someone can still explain it.
What each one is
An annual stocktake (a physical inventory) counts every product in every location on one date, usually at the end of the financial year. Receiving and shipping stop while it happens, so the figures are a clean snapshot.
Cycle counting counts a chosen slice of the stock on a regular rhythm: one aisle today, one supplier’s products tomorrow, the twenty most valuable products every month. Work carries on in the rest of the warehouse. Over the year every product is counted at least once, and the important ones many times.
| Annual stocktake | Cycle counting | |
|---|---|---|
| Frequency | Once a year | Weekly or daily, in slices |
| Disruption | Operations stop for a day or two | None, if planned around the slice being counted |
| When errors are found | Up to a year later | Within weeks |
| Cause of an error | Usually unknown | Often still traceable |
| Who counts | Everyone, including people who never handle stock | The people who work with that stock |
Cycle counting does not replace the year-end figure for your accountant. It makes that figure boring, because by December the stock is already right.
Planning the rhythm
A good cycle count plan counts the products where an error costs the most, most often. Two ways to slice the stock work well, and they can be combined.
By value and movement. Rank products by the value that passes through them in a year. The top 20% of products, which typically hold 80% of the value, get counted monthly. The middle group quarterly. The long tail of cheap, slow products once or twice a year. If you have already done an ABC analysis, the classes are your count frequencies.
By place. Count aisle by aisle, or shelf by shelf. This is easier to organise with a small team: one aisle a day gets through a 20-aisle warehouse in a month. It pairs well with shelf positions, because the count sheet is the shelf list.
Whichever slice you choose, write it down as a calendar. A cycle count that depends on someone remembering to do it is an annual stocktake in disguise.
Running a count that you can trust
Freeze the slice. Do not receive, ship or move stock in the aisle being counted until the count is approved. If a delivery lands between the moment you note the expected quantity and the moment you count, the difference you post will be wrong by exactly that delivery.
Count blind. The person counting should not see the expected quantity. People who see 148 on the sheet count 148.
Count units, not boxes. A box labelled 50 that holds 46 is where half the differences come from.
Recount before you post. Any difference above a threshold you set, in units or in value, gets a second count by a different person. Most large differences are counting errors, not stock errors.
Post the difference with a reason. Every correction should carry the count it came from and the date it applies to, so that six months later the movement history explains itself.
Reading the differences
The differences are the point. A count with no differences teaches you nothing. Group the differences and look for the pattern:
- Many small negatives on fast products usually mean stock leaving without a movement: samples, internal use, breakages nobody records.
- A positive on one product and a matching negative on a similar one is a picking mix-up, often two variants with similar codes.
- Repeated differences in one location point to a process, or a person, not to the stock.
Fix the cause, then watch the next count on the same slice. Accuracy improves count by count.
How StockBold handles stock counts
In StockBold a stock count is created for one location and can be narrowed to one supplier, one category or one department, which is how you count a slice rather than the whole warehouse. Products you leave out of the count are not touched.
When the count is created, StockBold freezes the expected quantity of every product in it. You enter what you found, and when the count is approved it posts a movement for every difference, dated on the count’s operation date and named after the count, so the movement history shows which count produced each correction. Movements at that location between creating and approving the count will distort the result, so create the count when you start counting and approve it when you finish.
Dates inside a locked financial year are refused, so a closed year’s figures stay as reported.
Read more about inventory management in StockBold, or the help guide on running a stock count.